diff --git a/Makefile b/Makefile
index 26084ae48d95a219f8f0ba824f8ca93db6192b69..8ff8135605e21ffa0e9a0418db3e80fa957efae7 100644
--- a/Makefile
+++ b/Makefile
@@ -5,7 +5,7 @@ image:
generate: image
@echo ">> Compiling assets and generating Kubernetes manifests"
- docker run --rm -v `pwd`:/go/src/github.com/coreos/prometheus-operator/contrib/kube-prometheus --workdir /go/src/github.com/coreos/prometheus-operator/contrib/kube-prometheus po-jsonnet make generate-raw
+ docker run --rm -u=$(shell id -u $(USER)):$(shell id -g $(USER)) -v `pwd`:/go/src/github.com/coreos/prometheus-operator/contrib/kube-prometheus --workdir /go/src/github.com/coreos/prometheus-operator/contrib/kube-prometheus po-jsonnet make generate-raw
generate-raw:
- ./hack/scripts/generate-manifests.sh
+ ./hack/scripts/build-jsonnet.sh example-dist/base/kube-prometheus.jsonnet manifests
diff --git a/README.md b/README.md
index 4ada050d3407663a4c160f752ac296a02cd70f95..7defae27272d6a860cb9448e74b74344bc28b387 100644
--- a/README.md
+++ b/README.md
@@ -1,5 +1,7 @@
# kube-prometheus
+> Note that everything in the `contrib/kube-prometheus/` directory is experimental and may change significantly at any time.
+
This repository collects Kubernetes manifests, [Grafana](http://grafana.com/) dashboards, and
[Prometheus rules](https://prometheus.io/docs/prometheus/latest/configuration/recording_rules/)
combined with documentation and scripts to provide single-command deployments of end-to-end
@@ -46,16 +48,15 @@ install
Simply run:
```bash
-export KUBECONFIG=<path> # defaults to "~/.kube/config"
cd contrib/kube-prometheus/
hack/cluster-monitoring/deploy
```
-After all pods are ready, you can reach:
+After all pods are ready, you can reach each of the UIs by port-forwarding:
-* Prometheus UI on node port `30900`
-* Alertmanager UI on node port `30903`
-* Grafana on node port `30902`
+* Prometheus UI on node port `kubectl -n monitoring port-forward prometheus-k8s-0 9090`
+* Alertmanager UI on node port `kubectl -n monitoring port-forward alertmanager-main-0 9093`
+* Grafana on node port `kubectl -n monitoring port-forward $(kubectl get pods -n monitoring -lapp=grafana -ojsonpath='{range .items[*]}{.metadata.name}{"\n"}{end}') 3000`
To tear it all down again, run:
@@ -63,9 +64,53 @@ To tear it all down again, run:
hack/cluster-monitoring/teardown
```
+## Customizing
+
+As everyone's infrastructure is slightly different, different organizations have different requirements. Thereby there may be modifications you want to do on kube-prometheus to fit your needs.
+
+The kube-prometheus stack is intended to be a jsonnet library for organizations to consume and use in their own infrastructure repository. Below is an example how it can be used to deploy the stack properly on minikube.
+
+The three "distribution" examples we have assembled can be found in:
+
+* `example-dist/base`: contains the plain kube-prometheus stack for organizations to build on.
+* `example-dist/kubeadm`: contains the kube-prometheus stack with slight modifications to work properly monitoring kubeadm clusters and exposes UIs on NodePorts for demonstration purposes.
+* `example-dist/bootkube`: contains the kube-prometheus stack with slight modifications to work properly on clusters created with bootkube.
+
+The examples in `example-dist/` are purely meant for demonstration purposes, the `kube-prometheus.jsonnet` file should live in your organizations infrastructure repository and use the kube-prometheus library provided here.
+
+Examples of additoinal modifications you may want to make could be adding an `Ingress` object for each of the UIs, but the point of this is that as opposed to other solutions out there, this library does not need to yield all possible customization options, it's all up to the user to customize!
+
+### minikube kubeadm example
+
+See `example-dist/kubeadm` for an example for deploying on minikube, using the minikube kubeadm bootstrapper. The `example-dist/kubeadm/kube-prometheus.jsonnet` file renders the kube-prometheus manifests using jsonnet and then merges the result with kubeadm specifics, such as information on how to monitor kube-controller-manager and kube-scheduler as created by kubeadm. In addition for demonstration purposes, it converts the services selecting Prometheus, Alertmanager and Grafana to NodePort services.
+
+Let's give that a try, and create a minikube cluster:
+
+```
+minikube delete && minikube start --kubernetes-version=v1.9.6 --memory=4096 --bootstrapper=kubeadm --extra-config=kubelet.authentication-token-webhook=true --extra-config=kubelet.authorization-mode=Webhook --extra-config=scheduler.address=0.0.0.0 --extra-config=controller-manager.address=0.0.0.0
+```
+
+Then we can render the manifests for kubeadm (because we are using the minikube kubeadm bootstrapper):
+
+```
+docker run --rm \
+ -v `pwd`:/go/src/github.com/coreos/prometheus-operator/contrib/kube-prometheus \
+ --workdir /go/src/github.com/coreos/prometheus-operator/contrib/kube-prometheus \
+ po-jsonnet \
+ ./hack/scripts/build-jsonnet.sh example-dist/kubeadm/kube-prometheus.jsonnet example-dist/kubeadm/manifests
+```
+
+> Note the `po-jsonnet` docker image is built using [this Dockerfile](/scripts/jsonnet/Dockerfile), you can also build it using `make image` from the `contrib/kube-prometheus` folder.
+
+Then the stack can be deployed using
+
+```
+hack/cluster-monitoring/deploy example-dist/kubeadm
+```
+
## Monitoring custom services
-The example manifests in [manifests/examples/example-app](/contrib/kube-prometheus/manifests/examples/example-app)
+The example manifests in [examples/example-app](/contrib/kube-prometheus/examples/example-app)
deploy a fake service exposing Prometheus metrics. They additionally define a new Prometheus
server and a [`ServiceMonitor`](https://github.com/coreos/prometheus-operator/blob/master/Documentation/design.md#servicemonitor),
which specifies how the example service should be monitored.
@@ -76,10 +121,13 @@ manage its life cycle.
hack/example-service-monitoring/deploy
```
-After all pods are ready you can reach the Prometheus server on node port `30100` and observe
-how it monitors the service as specified. Same as before, this Prometheus server automatically
-discovers the Alertmanager cluster deployed in the [Monitoring Kubernetes](#Monitoring-Kubernetes)
-section.
+After all pods are ready you can reach the Prometheus server similar to the Prometheus server above:
+
+```bash
+kubectl port-forward prometheus-frontend-0 9090
+```
+
+Then you can access Prometheus through `http://localhost:9090/`.
Teardown:
diff --git a/build.sh b/build.sh
deleted file mode 100755
index a42a6bb76b6ac8eee63469d2be28351f18054da3..0000000000000000000000000000000000000000
--- a/build.sh
+++ /dev/null
@@ -1,19 +0,0 @@
-#!/usr/bin/env bash
-set -e
-set -x
-
-prefix="tmp/manifests"
-json="tmp/manifests.json"
-
-rm -rf ${prefix}
-mkdir -p $(dirname "${json}")
-jsonnet -J /home/brancz/.jsonnet-bundler/src/git/git@github.com-ksonnet-ksonnet-lib/master jsonnet/kube-prometheus.jsonnet > ${json}
-
-files=$(jq -r 'keys[]' ${json})
-
-for file in ${files}; do
- dir=$(dirname "${file}")
- path="${prefix}/${dir}"
- mkdir -p ${path}
- jq -r ".[\"${file}\"]" ${json} | yaml2json | json2yaml > "${prefix}/${file}"
-done
diff --git a/example-dist/base/kube-prometheus.jsonnet b/example-dist/base/kube-prometheus.jsonnet
new file mode 100644
index 0000000000000000000000000000000000000000..01760e6545710034cbcb09a403d539f51aa0aca9
--- /dev/null
+++ b/example-dist/base/kube-prometheus.jsonnet
@@ -0,0 +1,6 @@
+local kubePrometheus = import "kube-prometheus.libsonnet";
+
+local namespace = "monitoring";
+local objects = kubePrometheus.new(namespace);
+
+{[path]: std.manifestYamlDoc(objects[path]) for path in std.objectFields(objects)}
diff --git a/example-dist/bootkube/.gitignore b/example-dist/bootkube/.gitignore
new file mode 100644
index 0000000000000000000000000000000000000000..4ea90de6f06de4520a0fc0bb782f154778860149
--- /dev/null
+++ b/example-dist/bootkube/.gitignore
@@ -0,0 +1,2 @@
+tmp/
+manifests/
diff --git a/example-dist/bootkube/kube-prometheus.jsonnet b/example-dist/bootkube/kube-prometheus.jsonnet
new file mode 100644
index 0000000000000000000000000000000000000000..fa731106ba1d9083e96cecbbd80b4ebb0b6853bd
--- /dev/null
+++ b/example-dist/bootkube/kube-prometheus.jsonnet
@@ -0,0 +1,36 @@
+local k = import "ksonnet.beta.3/k.libsonnet";
+local service = k.core.v1.service;
+local servicePort = k.core.v1.service.mixin.spec.portsType;
+local kubePrometheus = import "kube-prometheus.libsonnet";
+
+local namespace = "monitoring";
+
+local controllerManagerService = service.new("kube-controller-manager-prometheus-discovery", {"k8s-app": "kube-controller-manager"}, servicePort.newNamed("http-metrics", 10252, 10252)) +
+ service.mixin.metadata.withNamespace("kube-system") +
+ service.mixin.metadata.withLabels({"k8s-app": "kube-controller-manager"});
+
+local schedulerService = service.new("kube-scheduler-prometheus-discovery", {"k8s-app": "kube-scheduler"}, servicePort.newNamed("http-metrics", 10251, 10251)) +
+ service.mixin.metadata.withNamespace("kube-system") +
+ service.mixin.metadata.withLabels({"k8s-app": "kube-scheduler"});
+
+local kubeDNSService = service.new("kube-dns-prometheus-discovery", {"k8s-app": "kube-dns"}, [servicePort.newNamed("http-metrics-skydns", 10055, 10055), servicePort.newNamed("http-metrics-dnsmasq", 10054, 10054)]) +
+ service.mixin.metadata.withNamespace("kube-system") +
+ service.mixin.metadata.withLabels({"k8s-app": "kube-dns"});
+
+local objects = kubePrometheus.new(namespace) +
+ {
+ "prometheus-k8s/prometheus-k8s-service.yaml"+:
+ service.mixin.spec.withPorts(servicePort.newNamed("web", 9090, "web") + servicePort.withNodePort(30900)) +
+ service.mixin.spec.withType("NodePort"),
+ "alertmanager-main/alertmanager-main-service.yaml"+:
+ service.mixin.spec.withPorts(servicePort.newNamed("web", 9093, "web") + servicePort.withNodePort(30903)) +
+ service.mixin.spec.withType("NodePort"),
+ "grafana/grafana-service.yaml"+:
+ service.mixin.spec.withPorts(servicePort.newNamed("http", 3000, "http") + servicePort.withNodePort(30902)) +
+ service.mixin.spec.withType("NodePort"),
+ "prometheus-k8s/kube-controller-manager-prometheus-discovery-service.yaml": controllerManagerService,
+ "prometheus-k8s/kube-scheduler-prometheus-discovery-service.yaml": schedulerService,
+ "prometheus-k8s/kube-dns-prometheus-discovery-service.yaml": kubeDNSService,
+ };
+
+{[path]: std.manifestYamlDoc(objects[path]) for path in std.objectFields(objects)}
diff --git a/example-dist/kubeadm/.gitignore b/example-dist/kubeadm/.gitignore
new file mode 100644
index 0000000000000000000000000000000000000000..4ea90de6f06de4520a0fc0bb782f154778860149
--- /dev/null
+++ b/example-dist/kubeadm/.gitignore
@@ -0,0 +1,2 @@
+tmp/
+manifests/
diff --git a/example-dist/kubeadm/kube-prometheus.jsonnet b/example-dist/kubeadm/kube-prometheus.jsonnet
new file mode 100644
index 0000000000000000000000000000000000000000..50ce102092917b97a80bffd79a19af98a369687b
--- /dev/null
+++ b/example-dist/kubeadm/kube-prometheus.jsonnet
@@ -0,0 +1,31 @@
+local k = import "ksonnet.beta.3/k.libsonnet";
+local service = k.core.v1.service;
+local servicePort = k.core.v1.service.mixin.spec.portsType;
+local kubePrometheus = import "kube-prometheus.libsonnet";
+
+local namespace = "monitoring";
+
+local controllerManagerService = service.new("kube-controller-manager-prometheus-discovery", {component: "kube-controller-manager"}, servicePort.newNamed("http-metrics", 10252, 10252)) +
+ service.mixin.metadata.withNamespace("kube-system") +
+ service.mixin.metadata.withLabels({"k8s-app": "kube-controller-manager"});
+
+local schedulerService = service.new("kube-scheduler-prometheus-discovery", {component: "kube-scheduler"}, servicePort.newNamed("http-metrics", 10251, 10251)) +
+ service.mixin.metadata.withNamespace("kube-system") +
+ service.mixin.metadata.withLabels({"k8s-app": "kube-scheduler"});
+
+local objects = kubePrometheus.new(namespace) +
+ {
+ "prometheus-k8s/prometheus-k8s-service.yaml"+:
+ service.mixin.spec.withPorts(servicePort.newNamed("web", 9090, "web") + servicePort.withNodePort(30900)) +
+ service.mixin.spec.withType("NodePort"),
+ "alertmanager-main/alertmanager-main-service.yaml"+:
+ service.mixin.spec.withPorts(servicePort.newNamed("web", 9093, "web") + servicePort.withNodePort(30903)) +
+ service.mixin.spec.withType("NodePort"),
+ "grafana/grafana-service.yaml"+:
+ service.mixin.spec.withPorts(servicePort.newNamed("http", 3000, "http") + servicePort.withNodePort(30902)) +
+ service.mixin.spec.withType("NodePort"),
+ "prometheus-k8s/kube-controller-manager-prometheus-discovery-service.yaml": controllerManagerService,
+ "prometheus-k8s/kube-scheduler-prometheus-discovery-service.yaml": schedulerService,
+ };
+
+{[path]: std.manifestYamlDoc(objects[path]) for path in std.objectFields(objects)}
diff --git a/manifests/examples/basic-auth/secrets.yaml b/examples/basic-auth/secrets.yaml
similarity index 100%
rename from manifests/examples/basic-auth/secrets.yaml
rename to examples/basic-auth/secrets.yaml
diff --git a/manifests/examples/basic-auth/service-monitor.yaml b/examples/basic-auth/service-monitor.yaml
similarity index 100%
rename from manifests/examples/basic-auth/service-monitor.yaml
rename to examples/basic-auth/service-monitor.yaml
diff --git a/manifests/examples/example-app/example-app.yaml b/examples/example-app/example-app.yaml
similarity index 100%
rename from manifests/examples/example-app/example-app.yaml
rename to examples/example-app/example-app.yaml
diff --git a/manifests/examples/example-app/prometheus-frontend-alertmanager-discovery-role-binding.yaml b/examples/example-app/prometheus-frontend-alertmanager-discovery-role-binding.yaml
similarity index 100%
rename from manifests/examples/example-app/prometheus-frontend-alertmanager-discovery-role-binding.yaml
rename to examples/example-app/prometheus-frontend-alertmanager-discovery-role-binding.yaml
diff --git a/manifests/examples/example-app/prometheus-frontend-alertmanager-discovery-role.yaml b/examples/example-app/prometheus-frontend-alertmanager-discovery-role.yaml
similarity index 100%
rename from manifests/examples/example-app/prometheus-frontend-alertmanager-discovery-role.yaml
rename to examples/example-app/prometheus-frontend-alertmanager-discovery-role.yaml
diff --git a/manifests/examples/example-app/prometheus-frontend-role-binding.yaml b/examples/example-app/prometheus-frontend-role-binding.yaml
similarity index 100%
rename from manifests/examples/example-app/prometheus-frontend-role-binding.yaml
rename to examples/example-app/prometheus-frontend-role-binding.yaml
diff --git a/manifests/examples/example-app/prometheus-frontend-role.yaml b/examples/example-app/prometheus-frontend-role.yaml
similarity index 100%
rename from manifests/examples/example-app/prometheus-frontend-role.yaml
rename to examples/example-app/prometheus-frontend-role.yaml
diff --git a/manifests/examples/example-app/prometheus-frontend-service-account.yaml b/examples/example-app/prometheus-frontend-service-account.yaml
similarity index 100%
rename from manifests/examples/example-app/prometheus-frontend-service-account.yaml
rename to examples/example-app/prometheus-frontend-service-account.yaml
diff --git a/manifests/examples/example-app/prometheus-frontend-svc.yaml b/examples/example-app/prometheus-frontend-svc.yaml
similarity index 100%
rename from manifests/examples/example-app/prometheus-frontend-svc.yaml
rename to examples/example-app/prometheus-frontend-svc.yaml
diff --git a/manifests/examples/example-app/prometheus-frontend.yaml b/examples/example-app/prometheus-frontend.yaml
similarity index 100%
rename from manifests/examples/example-app/prometheus-frontend.yaml
rename to examples/example-app/prometheus-frontend.yaml
diff --git a/manifests/examples/example-app/servicemonitor-frontend.yaml b/examples/example-app/servicemonitor-frontend.yaml
similarity index 100%
rename from manifests/examples/example-app/servicemonitor-frontend.yaml
rename to examples/example-app/servicemonitor-frontend.yaml
diff --git a/manifests/custom-metrics-api/.gitignore b/experimental/custom-metrics-api/.gitignore
similarity index 100%
rename from manifests/custom-metrics-api/.gitignore
rename to experimental/custom-metrics-api/.gitignore
diff --git a/manifests/custom-metrics-api/README.md b/experimental/custom-metrics-api/README.md
similarity index 100%
rename from manifests/custom-metrics-api/README.md
rename to experimental/custom-metrics-api/README.md
diff --git a/manifests/custom-metrics-api/custom-metrics-apiserver-auth-delegator-cluster-role-binding.yaml b/experimental/custom-metrics-api/custom-metrics-apiserver-auth-delegator-cluster-role-binding.yaml
similarity index 100%
rename from manifests/custom-metrics-api/custom-metrics-apiserver-auth-delegator-cluster-role-binding.yaml
rename to experimental/custom-metrics-api/custom-metrics-apiserver-auth-delegator-cluster-role-binding.yaml
diff --git a/manifests/custom-metrics-api/custom-metrics-apiserver-auth-reader-role-binding.yaml b/experimental/custom-metrics-api/custom-metrics-apiserver-auth-reader-role-binding.yaml
similarity index 100%
rename from manifests/custom-metrics-api/custom-metrics-apiserver-auth-reader-role-binding.yaml
rename to experimental/custom-metrics-api/custom-metrics-apiserver-auth-reader-role-binding.yaml
diff --git a/manifests/custom-metrics-api/custom-metrics-apiserver-deployment.yaml b/experimental/custom-metrics-api/custom-metrics-apiserver-deployment.yaml
similarity index 100%
rename from manifests/custom-metrics-api/custom-metrics-apiserver-deployment.yaml
rename to experimental/custom-metrics-api/custom-metrics-apiserver-deployment.yaml
diff --git a/manifests/custom-metrics-api/custom-metrics-apiserver-resource-reader-cluster-role-binding.yaml b/experimental/custom-metrics-api/custom-metrics-apiserver-resource-reader-cluster-role-binding.yaml
similarity index 100%
rename from manifests/custom-metrics-api/custom-metrics-apiserver-resource-reader-cluster-role-binding.yaml
rename to experimental/custom-metrics-api/custom-metrics-apiserver-resource-reader-cluster-role-binding.yaml
diff --git a/manifests/custom-metrics-api/custom-metrics-apiserver-service-account.yaml b/experimental/custom-metrics-api/custom-metrics-apiserver-service-account.yaml
similarity index 100%
rename from manifests/custom-metrics-api/custom-metrics-apiserver-service-account.yaml
rename to experimental/custom-metrics-api/custom-metrics-apiserver-service-account.yaml
diff --git a/manifests/custom-metrics-api/custom-metrics-apiserver-service.yaml b/experimental/custom-metrics-api/custom-metrics-apiserver-service.yaml
similarity index 100%
rename from manifests/custom-metrics-api/custom-metrics-apiserver-service.yaml
rename to experimental/custom-metrics-api/custom-metrics-apiserver-service.yaml
diff --git a/manifests/custom-metrics-api/custom-metrics-apiservice.yaml b/experimental/custom-metrics-api/custom-metrics-apiservice.yaml
similarity index 100%
rename from manifests/custom-metrics-api/custom-metrics-apiservice.yaml
rename to experimental/custom-metrics-api/custom-metrics-apiservice.yaml
diff --git a/manifests/custom-metrics-api/custom-metrics-cluster-role.yaml b/experimental/custom-metrics-api/custom-metrics-cluster-role.yaml
similarity index 100%
rename from manifests/custom-metrics-api/custom-metrics-cluster-role.yaml
rename to experimental/custom-metrics-api/custom-metrics-cluster-role.yaml
diff --git a/manifests/custom-metrics-api/custom-metrics-resource-reader-cluster-role.yaml b/experimental/custom-metrics-api/custom-metrics-resource-reader-cluster-role.yaml
similarity index 100%
rename from manifests/custom-metrics-api/custom-metrics-resource-reader-cluster-role.yaml
rename to experimental/custom-metrics-api/custom-metrics-resource-reader-cluster-role.yaml
diff --git a/manifests/custom-metrics-api/deploy.sh b/experimental/custom-metrics-api/deploy.sh
similarity index 100%
rename from manifests/custom-metrics-api/deploy.sh
rename to experimental/custom-metrics-api/deploy.sh
diff --git a/manifests/custom-metrics-api/gencerts.sh b/experimental/custom-metrics-api/gencerts.sh
similarity index 100%
rename from manifests/custom-metrics-api/gencerts.sh
rename to experimental/custom-metrics-api/gencerts.sh
diff --git a/manifests/custom-metrics-api/hpa-custom-metrics-cluster-role-binding.yaml b/experimental/custom-metrics-api/hpa-custom-metrics-cluster-role-binding.yaml
similarity index 100%
rename from manifests/custom-metrics-api/hpa-custom-metrics-cluster-role-binding.yaml
rename to experimental/custom-metrics-api/hpa-custom-metrics-cluster-role-binding.yaml
diff --git a/manifests/custom-metrics-api/teardown.sh b/experimental/custom-metrics-api/teardown.sh
similarity index 100%
rename from manifests/custom-metrics-api/teardown.sh
rename to experimental/custom-metrics-api/teardown.sh
diff --git a/manifests/metrics-server/auth-delegator.yaml b/experimental/metrics-server/auth-delegator.yaml
similarity index 100%
rename from manifests/metrics-server/auth-delegator.yaml
rename to experimental/metrics-server/auth-delegator.yaml
diff --git a/manifests/metrics-server/auth-reader.yaml b/experimental/metrics-server/auth-reader.yaml
similarity index 100%
rename from manifests/metrics-server/auth-reader.yaml
rename to experimental/metrics-server/auth-reader.yaml
diff --git a/manifests/metrics-server/metrics-apiservice.yaml b/experimental/metrics-server/metrics-apiservice.yaml
similarity index 100%
rename from manifests/metrics-server/metrics-apiservice.yaml
rename to experimental/metrics-server/metrics-apiservice.yaml
diff --git a/manifests/metrics-server/metrics-server-cluster-role-binding.yaml b/experimental/metrics-server/metrics-server-cluster-role-binding.yaml
similarity index 100%
rename from manifests/metrics-server/metrics-server-cluster-role-binding.yaml
rename to experimental/metrics-server/metrics-server-cluster-role-binding.yaml
diff --git a/manifests/metrics-server/metrics-server-cluster-role.yaml b/experimental/metrics-server/metrics-server-cluster-role.yaml
similarity index 100%
rename from manifests/metrics-server/metrics-server-cluster-role.yaml
rename to experimental/metrics-server/metrics-server-cluster-role.yaml
diff --git a/manifests/metrics-server/metrics-server-deployment.yaml b/experimental/metrics-server/metrics-server-deployment.yaml
similarity index 100%
rename from manifests/metrics-server/metrics-server-deployment.yaml
rename to experimental/metrics-server/metrics-server-deployment.yaml
diff --git a/manifests/metrics-server/metrics-server-service-account.yaml b/experimental/metrics-server/metrics-server-service-account.yaml
similarity index 100%
rename from manifests/metrics-server/metrics-server-service-account.yaml
rename to experimental/metrics-server/metrics-server-service-account.yaml
diff --git a/manifests/metrics-server/metrics-server-service.yaml b/experimental/metrics-server/metrics-server-service.yaml
similarity index 100%
rename from manifests/metrics-server/metrics-server-service.yaml
rename to experimental/metrics-server/metrics-server-service.yaml
diff --git a/hack/cluster-monitoring/deploy b/hack/cluster-monitoring/deploy
index a4f7c184403f2cd93569b3fd61443c15ecfaecde..41e051876498ccf22256c52b731d2a86be297d2d 100755
--- a/hack/cluster-monitoring/deploy
+++ b/hack/cluster-monitoring/deploy
@@ -1,40 +1,24 @@
#!/usr/bin/env bash
-if [ -z "${KUBECONFIG}" ]; then
- export KUBECONFIG=~/.kube/config
-fi
+manifest_prefix=${1-.}
-# CAUTION - setting NAMESPACE will deploy most components to the given namespace
-# however some are hardcoded to 'monitoring'. Only use if you have reviewed all manifests.
+kubectl create namespace monitoring
-if [ -z "${NAMESPACE}" ]; then
- NAMESPACE=monitoring
-fi
-
-kubectl create namespace "$NAMESPACE"
-
-kctl() {
- kubectl --namespace "$NAMESPACE" "$@"
-}
-
-kctl apply -f manifests/prometheus-operator
+kubectl apply -f ${manifest_prefix}/manifests/prometheus-operator/
# Wait for CRDs to be ready.
printf "Waiting for Operator to register custom resource definitions..."
-until kctl get customresourcedefinitions servicemonitors.monitoring.coreos.com > /dev/null 2>&1; do sleep 1; printf "."; done
-until kctl get customresourcedefinitions prometheuses.monitoring.coreos.com > /dev/null 2>&1; do sleep 1; printf "."; done
-until kctl get customresourcedefinitions alertmanagers.monitoring.coreos.com > /dev/null 2>&1; do sleep 1; printf "."; done
-until kctl get servicemonitors.monitoring.coreos.com > /dev/null 2>&1; do sleep 1; printf "."; done
-until kctl get prometheuses.monitoring.coreos.com > /dev/null 2>&1; do sleep 1; printf "."; done
-until kctl get alertmanagers.monitoring.coreos.com > /dev/null 2>&1; do sleep 1; printf "."; done
+until kubectl get customresourcedefinitions servicemonitors.monitoring.coreos.com > /dev/null 2>&1; do sleep 1; printf "."; done
+until kubectl get customresourcedefinitions prometheuses.monitoring.coreos.com > /dev/null 2>&1; do sleep 1; printf "."; done
+until kubectl get customresourcedefinitions alertmanagers.monitoring.coreos.com > /dev/null 2>&1; do sleep 1; printf "."; done
+until kubectl get servicemonitors.monitoring.coreos.com > /dev/null 2>&1; do sleep 1; printf "."; done
+until kubectl get prometheuses.monitoring.coreos.com > /dev/null 2>&1; do sleep 1; printf "."; done
+until kubectl get alertmanagers.monitoring.coreos.com > /dev/null 2>&1; do sleep 1; printf "."; done
echo "done!"
-kctl apply -f manifests/node-exporter
-kctl apply -f manifests/kube-state-metrics
-kctl apply -f manifests/grafana/grafana-credentials.yaml
-kctl apply -f manifests/grafana
-find manifests/prometheus -type f ! -name prometheus-k8s-roles.yaml ! -name prometheus-k8s-role-bindings.yaml -exec kubectl --namespace "$NAMESPACE" apply -f {} \;
-kubectl apply -f manifests/prometheus/prometheus-k8s-roles.yaml
-kubectl apply -f manifests/prometheus/prometheus-k8s-role-bindings.yaml
-kctl apply -f manifests/alertmanager/
+kubectl apply -f ${manifest_prefix}/manifests/node-exporter/
+kubectl apply -f ${manifest_prefix}/manifests/kube-state-metrics/
+kubectl apply -f ${manifest_prefix}/manifests/grafana/
+kubectl apply -f ${manifest_prefix}/manifests/prometheus-k8s/
+kubectl apply -f ${manifest_prefix}/manifests/alertmanager-main/
diff --git a/hack/cluster-monitoring/minikube-deploy b/hack/cluster-monitoring/minikube-deploy
deleted file mode 100755
index 64cb86be7e2ab34634741d5943d43a077ee9bc0d..0000000000000000000000000000000000000000
--- a/hack/cluster-monitoring/minikube-deploy
+++ /dev/null
@@ -1,17 +0,0 @@
-#!/usr/bin/env bash
-
-# We assume that the kubelet uses token authN and authZ, as otherwise
-# Prometheus needs a client certificate, which gives it full access to the
-# kubelet, rather than just the metrics. Token authN and authZ allows more fine
-# grained and easier access control. Simply start minikube with the following
-# command (you can of course adapt the version and memory to your needs):
-#
-# $ minikube delete && minikube start --kubernetes-version=v1.9.1 --memory=4096 --bootstrapper=kubeadm --extra-config=kubelet.authentication-token-webhook=true --extra-config=kubelet.authorization-mode=Webhook --extra-config=scheduler.address=0.0.0.0 --extra-config=controller-manager.address=0.0.0.0
-#
-# In future versions of minikube and kubeadm this will be the default, but for
-# the time being, we will have to configure it ourselves.
-
-hack/cluster-monitoring/deploy
-
-kubectl --namespace=kube-system apply -f manifests/k8s/kubeadm/
-
diff --git a/hack/cluster-monitoring/minikube-teardown b/hack/cluster-monitoring/minikube-teardown
deleted file mode 100755
index 3a4c986ed06f08edf4bd845c5223f90f4168ea43..0000000000000000000000000000000000000000
--- a/hack/cluster-monitoring/minikube-teardown
+++ /dev/null
@@ -1,6 +0,0 @@
-#!/usr/bin/env bash
-
-hack/cluster-monitoring/teardown
-
-kubectl --namespace=kube-system delete -f manifests/k8s/minikube
-
diff --git a/hack/cluster-monitoring/self-hosted-deploy b/hack/cluster-monitoring/self-hosted-deploy
deleted file mode 100755
index 7cbce37d5304346fe71dd8d69be17b9debeeb0ee..0000000000000000000000000000000000000000
--- a/hack/cluster-monitoring/self-hosted-deploy
+++ /dev/null
@@ -1,6 +0,0 @@
-#!/usr/bin/env bash
-
-hack/cluster-monitoring/deploy
-
-kubectl apply -f manifests/k8s/self-hosted
-
diff --git a/hack/cluster-monitoring/self-hosted-teardown b/hack/cluster-monitoring/self-hosted-teardown
deleted file mode 100755
index f9d7da9f7cb805344eedc997cb73b0fc0e5143cc..0000000000000000000000000000000000000000
--- a/hack/cluster-monitoring/self-hosted-teardown
+++ /dev/null
@@ -1,6 +0,0 @@
-#!/usr/bin/env bash
-
-hack/cluster-monitoring/teardown
-
-kubectl delete -f manifests/k8s/self-hosted
-
diff --git a/hack/cluster-monitoring/teardown b/hack/cluster-monitoring/teardown
index b2c4c544bbdf19bb7047d720b45f20bd5e668f81..0ef9a6b3d80e6ed7184694fc51fbc499e6609e10 100755
--- a/hack/cluster-monitoring/teardown
+++ b/hack/cluster-monitoring/teardown
@@ -1,30 +1,4 @@
#!/usr/bin/env bash
-if [ -z "${KUBECONFIG}" ]; then
- export KUBECONFIG=~/.kube/config
-fi
-
-# CAUTION - NAMESPACE must match its value when deploy script was run.
-# Some resources are always deployed to the monitoring namespace.
-
-if [ -z "${NAMESPACE}" ]; then
- NAMESPACE=monitoring
-fi
-
-kctl() {
- kubectl --namespace "$NAMESPACE" "$@"
-}
-
-kctl delete -f manifests/node-exporter
-kctl delete -f manifests/kube-state-metrics
-kctl delete -f manifests/grafana
-find manifests/prometheus -type f ! -name prometheus-k8s-roles.yaml ! -name prometheus-k8s-role-bindings.yaml -exec kubectl --namespace "$NAMESPACE" delete -f {} \;
-kubectl delete -f manifests/prometheus/prometheus-k8s-roles.yaml
-kubectl delete -f manifests/prometheus/prometheus-k8s-role-bindings.yaml
-kctl delete -f manifests/alertmanager
-
-# Hack: wait a bit to let the controller delete the deployed Prometheus server.
-sleep 5
-
-kctl delete -f manifests/prometheus-operator
+kubectl delete namespace monitoring
diff --git a/hack/example-service-monitoring/deploy b/hack/example-service-monitoring/deploy
index 18b0ef6a6120bb470bec1becc00128ec6de797f4..4912dd9658d18db27dba7e900a08edb0b03ed1b1 100755
--- a/hack/example-service-monitoring/deploy
+++ b/hack/example-service-monitoring/deploy
@@ -1,3 +1,3 @@
#!/usr/bin/env bash
-kubectl apply -f manifests/examples/example-app
+kubectl apply -f examples/example-app
diff --git a/hack/example-service-monitoring/teardown b/hack/example-service-monitoring/teardown
index a5fc176023d8ed3c33512afa26a4ce8a9f809f16..62b546dea28c9d8519059b1f49beff49fabc38f0 100755
--- a/hack/example-service-monitoring/teardown
+++ b/hack/example-service-monitoring/teardown
@@ -1,3 +1,3 @@
#!/usr/bin/env bash
-kubectl delete -f manifests/examples/example-app
+kubectl delete -f examples/example-app
diff --git a/hack/scripts/build-jsonnet.sh b/hack/scripts/build-jsonnet.sh
new file mode 100755
index 0000000000000000000000000000000000000000..7189962f7aecb5c45489c0214ba67e6fd9834425
--- /dev/null
+++ b/hack/scripts/build-jsonnet.sh
@@ -0,0 +1,25 @@
+#!/usr/bin/env bash
+set -e
+set -x
+
+jsonnet="${1-kube-prometheus.jsonnet}"
+prefix="${2-manifests}"
+json="tmp/manifests.json"
+
+rm -rf ${prefix}
+mkdir -p $(dirname "${json}")
+jsonnet \
+ -J $GOPATH/src/github.com/ksonnet/ksonnet-lib \
+ -J $GOPATH/src/github.com/grafana/grafonnet-lib \
+ -J $GOPATH/src/github.com/coreos/prometheus-operator/contrib/kube-prometheus/jsonnet \
+ -J $GOPATH/src/github.com/brancz/kubernetes-grafana/src/kubernetes-jsonnet \
+ ${jsonnet} > ${json}
+
+files=$(jq -r 'keys[]' ${json})
+
+for file in ${files}; do
+ dir=$(dirname "${file}")
+ path="${prefix}/${dir}"
+ mkdir -p ${path}
+ jq -r ".[\"${file}\"]" ${json} | gojsontoyaml -yamltojson | gojsontoyaml > "${prefix}/${file}"
+done
diff --git a/hack/scripts/generate-alertmanager-config-secret.sh b/hack/scripts/generate-alertmanager-config-secret.sh
deleted file mode 100755
index b0b4aaef77eb4467a51b2b979cd916b41ceed798..0000000000000000000000000000000000000000
--- a/hack/scripts/generate-alertmanager-config-secret.sh
+++ /dev/null
@@ -1,11 +0,0 @@
-#!/bin/bash
-
-cat <<-EOF
-apiVersion: v1
-kind: Secret
-metadata:
- name: alertmanager-main
-data:
- alertmanager.yaml: $(cat assets/alertmanager/alertmanager.yaml | base64 --wrap=0)
-EOF
-
diff --git a/hack/scripts/generate-dashboards-configmap.sh b/hack/scripts/generate-dashboards-configmap.sh
deleted file mode 100755
index 47ccfc127d876d85d7c46bf5e6e9d7f61f89dca6..0000000000000000000000000000000000000000
--- a/hack/scripts/generate-dashboards-configmap.sh
+++ /dev/null
@@ -1,39 +0,0 @@
-#!/bin/bash
-set -e
-set +x
-
-cat <<-EOF
-apiVersion: v1
-kind: ConfigMap
-metadata:
- name: grafana-dashboard-definitions-0
-data:
-EOF
-
-for f in assets/grafana/generated/*-dashboard.json
-do
- rm -rf $f
-done
-
-virtualenv -p python3 .env 2>&1 > /dev/null
-source .env/bin/activate 2>&1 > /dev/null
-pip install -Ur requirements.txt 2>&1 > /dev/null
-for f in assets/grafana/*.dashboard.py
-do
- basefilename=$(basename $f)
- JSON_FILENAME="assets/grafana/generated/${basefilename%%.*}-dashboard.json"
- generate-dashboard $f -o $JSON_FILENAME 2>&1 > /dev/null
-done
-
-cp assets/grafana/raw-json-dashboards/*-dashboard.json assets/grafana/generated/
-
-for f in assets/grafana/generated/*-dashboard.json
-do
- basefilename=$(basename $f)
- echo " $basefilename: |+"
- if [ "$basefilename" = "etcd-dashboard.json" ]; then
- hack/scripts/wrap-dashboard.sh $f prometheus-etcd | sed "s/^/ /g"
- else
- hack/scripts/wrap-dashboard.sh $f prometheus | sed "s/^/ /g"
- fi
-done
diff --git a/hack/scripts/generate-grafana-credentials-secret.sh b/hack/scripts/generate-grafana-credentials-secret.sh
deleted file mode 100755
index e877b080b1442d829c386e74a8c93c222a4138fe..0000000000000000000000000000000000000000
--- a/hack/scripts/generate-grafana-credentials-secret.sh
+++ /dev/null
@@ -1,20 +0,0 @@
-#!/bin/bash
-
-if [ "$#" -ne 2 ]; then
- echo "Usage: $0 user password"
- exit 1
-fi
-
-user=$1
-password=$2
-
-cat <<-EOF
-apiVersion: v1
-kind: Secret
-metadata:
- name: grafana-credentials
-data:
- user: $(echo -n ${user} | base64 --wrap=0)
- password: $(echo -n ${password} | base64 --wrap=0)
-EOF
-
diff --git a/hack/scripts/generate-manifests.sh b/hack/scripts/generate-manifests.sh
deleted file mode 100755
index 6f14056b9c69cf75ebcea27ce866eca7ce56521e..0000000000000000000000000000000000000000
--- a/hack/scripts/generate-manifests.sh
+++ /dev/null
@@ -1,26 +0,0 @@
-#!/bin/bash
-set -e
-set +x
-
-# Generate Alert Rules ConfigMap
-hack/scripts/generate-rules-configmap.sh > manifests/prometheus/prometheus-k8s-rules.yaml
-
-# Generate Dashboard ConfigMap
-hack/scripts/generate-dashboards-configmap.sh > manifests/grafana/grafana-dashboard-definitions.yaml
-
-# Generate Dashboard ConfigMap with configmap-generator tool
-# Max Size per ConfigMap: 240000
-# Input dir: assets/grafana
-# output file: manifests/grafana/grafana-dashboards.yaml
-# grafana deployment output file: manifests/grafana/grafana-deployment.yaml
-test -f manifests/grafana/grafana-dashboard-definitions.yaml && rm -f manifests/grafana/grafana-dashboard-definitions.yaml
-test -f manifests/grafana/grafana-deployment.yaml && rm -f manifests/grafana/grafana-deployment.yaml
-test -f manifests/grafana/grafana-dashboards.yaml && rm -f manifests/grafana/grafana-dashboards.yaml
-hack/grafana-dashboards-configmap-generator/bin/grafana_dashboards_generate.sh -s 240000 -i assets/grafana/generated -o manifests/grafana/grafana-dashboard-definitions.yaml -g manifests/grafana/grafana-deployment.yaml -d manifests/grafana/grafana-dashboards.yaml
-
-# Generate Grafana Credentials Secret
-hack/scripts/generate-grafana-credentials-secret.sh admin admin > manifests/grafana/grafana-credentials.yaml
-
-# Generate Secret for Alertmanager config
-hack/scripts/generate-alertmanager-config-secret.sh > manifests/alertmanager/alertmanager-config.yaml
-
diff --git a/hack/scripts/generate-rules-configmap.sh b/hack/scripts/generate-rules-configmap.sh
deleted file mode 100755
index 96c5433f18229ee71e645b864d7da44693842922..0000000000000000000000000000000000000000
--- a/hack/scripts/generate-rules-configmap.sh
+++ /dev/null
@@ -1,18 +0,0 @@
-#!/bin/bash
-
-cat <<-EOF
-apiVersion: v1
-kind: ConfigMap
-metadata:
- name: prometheus-k8s-rules
- labels:
- role: alert-rules
- prometheus: k8s
-data:
-EOF
-
-for f in assets/prometheus/rules/*.rules.y*ml
-do
- echo " $(basename "$f"): |+"
- cat $f | sed "s/^/ /g"
-done
diff --git a/hack/scripts/wrap-dashboard.sh b/hack/scripts/wrap-dashboard.sh
deleted file mode 100755
index d3b04085902aa87e8fefa548144a9f59ced9a2a7..0000000000000000000000000000000000000000
--- a/hack/scripts/wrap-dashboard.sh
+++ /dev/null
@@ -1,51 +0,0 @@
-#!/bin/bash -eu
-
-# Intended usage:
-# * Edit dashboard in Grafana (you need to login first with admin/admin
-# login/password).
-# * Save dashboard in Grafana to check is specification is correct.
-# Looks like this is the only way to check if dashboard specification
-# has errors.
-# * Download dashboard specification as JSON file in Grafana:
-# Share -> Export -> Save to file.
-# * Drop dashboard specification in assets folder:
-# mv Nodes-1488465802729.json assets/grafana/node-dashboard.json
-# * Regenerate Grafana configmap:
-# ./hack/scripts/generate-manifests.sh
-# * Apply new configmap:
-# kubectl -n monitoring apply -f manifests/grafana/grafana-cm.yaml
-
-if [ "$#" -ne 2 ]; then
- echo "Usage: $0 path-to-dashboard.json grafana-prometheus-datasource-name"
- exit 1
-fi
-
-dashboardjson=$1
-datasource_name=$2
-inputname="DS_PROMETHEUS"
-
-if [ "$datasource_name" = "prometheus-etcd" ]; then
- inputname="DS_PROMETHEUS-ETCD"
-fi
-
-cat <<EOF
-{
- "dashboard":
-EOF
-
-cat $dashboardjson
-
-cat <<EOF
-,
- "inputs": [
- {
- "name": "$inputname",
- "pluginId": "prometheus",
- "type": "datasource",
- "value": "$datasource_name"
- }
- ],
- "overwrite": true
-}
-EOF
-
diff --git a/jsonnet/alertmanager/alertmanager-main-secret.libsonnet b/jsonnet/alertmanager/alertmanager-main-secret.libsonnet
index fca165667692cf2cd3fdba34ef527f5e93951b09..a8f9011b3bd6ed062661fe8b9de3535d1d91de8d 100644
--- a/jsonnet/alertmanager/alertmanager-main-secret.libsonnet
+++ b/jsonnet/alertmanager/alertmanager-main-secret.libsonnet
@@ -1,25 +1,8 @@
local k = import "ksonnet.beta.3/k.libsonnet";
local secret = k.core.v1.secret;
-local plainConfig = "global:
- resolve_timeout: 5m
-route:
- group_by: ['job']
- group_wait: 30s
- group_interval: 5m
- repeat_interval: 12h
- receiver: 'null'
- routes:
- - match:
- alertname: DeadMansSwitch
- receiver: 'null'
-receivers:
-- name: 'null'";
-
-local config = std.base64(plainConfig);
-
{
- new(namespace)::
- secret.new("alertmanager-main", {"alertmanager.yaml": config}) +
+ new(namespace, plainConfig)::
+ secret.new("alertmanager-main", {"alertmanager.yaml": std.base64(plainConfig)}) +
secret.mixin.metadata.withNamespace(namespace)
}
diff --git a/jsonnet/kube-prometheus.jsonnet b/jsonnet/kube-prometheus.jsonnet
deleted file mode 100644
index 3a0ef2cfc477bc4db79eb6c852c54f77f0f8f582..0000000000000000000000000000000000000000
--- a/jsonnet/kube-prometheus.jsonnet
+++ /dev/null
@@ -1,62 +0,0 @@
-local k = import "ksonnet.beta.3/k.libsonnet";
-
-local alertmanager = import "alertmanager/alertmanager.libsonnet";
-local ksm = import "kube-state-metrics/kube-state-metrics.libsonnet";
-local nodeExporter = import "node-exporter/node-exporter.libsonnet";
-local po = import "prometheus-operator/prometheus-operator.libsonnet";
-local prometheus = import "prometheus/prometheus.libsonnet";
-
-local namespace = "monitoring";
-
-local objects = {
- "alertmanager-main/alertmanager-main-secret.yaml": alertmanager.config.new(namespace),
- "alertmanager-main/alertmanager-main-service-account.yaml": alertmanager.serviceAccount.new(namespace),
- "alertmanager-main/alertmanager-main-service.yaml": alertmanager.service.new(namespace),
- "alertmanager-main/alertmanager-main.yaml": alertmanager.alertmanager.new(namespace),
-
- "kube-state-metrics/kube-state-metrics-cluster-role-binding": ksm.clusterRoleBinding.new(namespace),
- "kube-state-metrics/kube-state-metrics-cluster-role.yaml": ksm.clusterRole.new(),
- "kube-state-metrics/kube-state-metrics-deployment.yaml": ksm.deployment.new(namespace),
- "kube-state-metrics/kube-state-metrics-role-binding.yaml": ksm.roleBinding.new(namespace),
- "kube-state-metrics/kube-state-metrics-role.yaml": ksm.role.new(namespace),
- "kube-state-metrics/kube-state-metrics-service-account.yaml": ksm.serviceAccount.new(namespace),
- "kube-state-metrics/kube-state-metrics-service.yaml": ksm.service.new(namespace),
-
- "node-exporter/node-exporter-cluster-role-binding.yaml": nodeExporter.clusterRoleBinding.new(namespace),
- "node-exporter/node-exporter-cluster-role.yaml": nodeExporter.clusterRole.new(),
- "node-exporter/node-exporter-daemonset.yaml": nodeExporter.daemonset.new(namespace),
- "node-exporter/node-exporter-service-account.yaml": nodeExporter.serviceAccount.new(namespace),
- "node-exporter/node-exporter-service.yaml": nodeExporter.service.new(namespace),
-
- "prometheus-operator/prometheus-operator-cluster-role-binding.yaml": po.clusterRoleBinding.new(namespace),
- "prometheus-operator/prometheus-operator-cluster-role.yaml": po.clusterRole.new(),
- "prometheus-operator/prometheus-operator-deployment.yaml": po.deployment.new(namespace),
- "prometheus-operator/prometheus-operator-service.yaml": po.service.new(namespace),
- "prometheus-operator/prometheus-operator-service-account.yaml": po.serviceAccount.new(namespace),
-
- "prometheus-k8s/prometheus-k8s-cluster-role-binding.yaml": prometheus.clusterRoleBinding.new(namespace),
- "prometheus-k8s/prometheus-k8s-cluster-role.yaml": prometheus.clusterRole.new(),
- "prometheus-k8s/prometheus-k8s-service-account.yaml": prometheus.serviceAccount.new(namespace),
- "prometheus-k8s/prometheus-k8s-service.yaml": prometheus.service.new(namespace),
- "prometheus-k8s/prometheus-k8s.yaml": prometheus.prometheus.new(namespace),
- "prometheus-k8s/prometheus-k8s-role-binding-config.yaml": prometheus.roleBindingConfig.new(namespace),
- "prometheus-k8s/prometheus-k8s-role-binding-namespace.yaml": prometheus.roleBindingNamespace.new(namespace),
- "prometheus-k8s/prometheus-k8s-role-binding-kube-system.yaml": prometheus.roleBindingKubeSystem.new(namespace),
- "prometheus-k8s/prometheus-k8s-role-binding-default.yaml": prometheus.roleBindingDefault.new(namespace),
- "prometheus-k8s/prometheus-k8s-role-config.yaml": prometheus.roleConfig.new(namespace),
- "prometheus-k8s/prometheus-k8s-role-namespace.yaml": prometheus.roleNamespace.new(namespace),
- "prometheus-k8s/prometheus-k8s-role-kube-system.yaml": prometheus.roleKubeSystem.new(),
- "prometheus-k8s/prometheus-k8s-role-default.yaml": prometheus.roleDefault.new(),
- "prometheus-k8s/prometheus-k8s-service-monitor-alertmanager.yaml": prometheus.serviceMonitorAlertmanager.new(namespace),
- "prometheus-k8s/prometheus-k8s-service-monitor-apiserver.yaml": prometheus.serviceMonitorApiserver.new(namespace),
- "prometheus-k8s/prometheus-k8s-service-monitor-coredns.yaml": prometheus.serviceMonitorCoreDNS.new(namespace),
- "prometheus-k8s/prometheus-k8s-service-monitor-kube-controller-manager.yaml": prometheus.serviceMonitorControllerManager.new(namespace),
- "prometheus-k8s/prometheus-k8s-service-monitor-kube-scheduler.yaml": prometheus.serviceMonitorScheduler.new(namespace),
- "prometheus-k8s/prometheus-k8s-service-monitor-kube-state-metrics.yaml": prometheus.serviceMonitorKubeStateMetrics.new(namespace),
- "prometheus-k8s/prometheus-k8s-service-monitor-kubelet.yaml": prometheus.serviceMonitorKubelet.new(namespace),
- "prometheus-k8s/prometheus-k8s-service-monitor-node-exporter.yaml": prometheus.serviceMonitorNodeExporter.new(namespace),
- "prometheus-k8s/prometheus-k8s-service-monitor-prometheus-operator.yaml": prometheus.serviceMonitorPrometheusOperator.new(namespace),
- "prometheus-k8s/prometheus-k8s-service-monitor-prometheus.yaml": prometheus.serviceMonitorPrometheus.new(namespace),
-};
-
-{[path]: std.manifestYamlDoc(objects[path]) for path in std.objectFields(objects)}
diff --git a/jsonnet/kube-prometheus.libsonnet b/jsonnet/kube-prometheus.libsonnet
new file mode 100644
index 0000000000000000000000000000000000000000..44ee06a5dfd41b0c2bd49b8b2be81d508914cb43
--- /dev/null
+++ b/jsonnet/kube-prometheus.libsonnet
@@ -0,0 +1,85 @@
+local k = import "ksonnet.beta.3/k.libsonnet";
+
+local alertmanager = import "alertmanager/alertmanager.libsonnet";
+local ksm = import "kube-state-metrics/kube-state-metrics.libsonnet";
+local nodeExporter = import "node-exporter/node-exporter.libsonnet";
+local po = import "prometheus-operator/prometheus-operator.libsonnet";
+local prometheus = import "prometheus/prometheus.libsonnet";
+local grafana = import "grafana/grafana.libsonnet";
+
+local alertmanagerConfig = importstr "../assets/alertmanager/alertmanager.yaml";
+
+local ruleFiles = {
+ "alertmanager.rules.yaml": importstr "../assets/prometheus/rules/alertmanager.rules.yaml",
+ "etcd3.rules.yaml": importstr "../assets/prometheus/rules/etcd3.rules.yaml",
+ "general.rules.yaml": importstr "../assets/prometheus/rules/general.rules.yaml",
+ "kube-controller-manager.rules.yaml": importstr "../assets/prometheus/rules/kube-controller-manager.rules.yaml",
+ "kube-scheduler.rules.yaml": importstr "../assets/prometheus/rules/kube-scheduler.rules.yaml",
+ "kube-state-metrics.rules.yaml": importstr "../assets/prometheus/rules/kube-state-metrics.rules.yaml",
+ "kubelet.rules.yaml": importstr "../assets/prometheus/rules/kubelet.rules.yaml",
+ "kubernetes.rules.yaml": importstr "../assets/prometheus/rules/kubernetes.rules.yaml",
+ "node.rules.yaml": importstr "../assets/prometheus/rules/node.rules.yaml",
+ "prometheus.rules.yaml": importstr "../assets/prometheus/rules/prometheus.rules.yaml",
+};
+
+{
+ new(namespace)::
+ {
+ "grafana/grafana-dashboard-definitions.yaml": grafana.dashboardDefinitions.new(namespace),
+ "grafana/grafana-dashboard-sources.yaml": grafana.dashboardSources.new(namespace),
+ "grafana/grafana-datasources.yaml": grafana.dashboardDatasources.new(namespace),
+ "grafana/grafana-deployment.yaml": grafana.deployment.new(namespace),
+ "grafana/grafana-service-account.yaml": grafana.serviceAccount.new(namespace),
+ "grafana/grafana-service.yaml": grafana.service.new(namespace),
+
+ "alertmanager-main/alertmanager-main-secret.yaml": alertmanager.config.new(namespace, alertmanagerConfig),
+ "alertmanager-main/alertmanager-main-service-account.yaml": alertmanager.serviceAccount.new(namespace),
+ "alertmanager-main/alertmanager-main-service.yaml": alertmanager.service.new(namespace),
+ "alertmanager-main/alertmanager-main.yaml": alertmanager.alertmanager.new(namespace),
+
+ "kube-state-metrics/kube-state-metrics-cluster-role-binding.yaml": ksm.clusterRoleBinding.new(namespace),
+ "kube-state-metrics/kube-state-metrics-cluster-role.yaml": ksm.clusterRole.new(),
+ "kube-state-metrics/kube-state-metrics-deployment.yaml": ksm.deployment.new(namespace),
+ "kube-state-metrics/kube-state-metrics-role-binding.yaml": ksm.roleBinding.new(namespace),
+ "kube-state-metrics/kube-state-metrics-role.yaml": ksm.role.new(namespace),
+ "kube-state-metrics/kube-state-metrics-service-account.yaml": ksm.serviceAccount.new(namespace),
+ "kube-state-metrics/kube-state-metrics-service.yaml": ksm.service.new(namespace),
+
+ "node-exporter/node-exporter-cluster-role-binding.yaml": nodeExporter.clusterRoleBinding.new(namespace),
+ "node-exporter/node-exporter-cluster-role.yaml": nodeExporter.clusterRole.new(),
+ "node-exporter/node-exporter-daemonset.yaml": nodeExporter.daemonset.new(namespace),
+ "node-exporter/node-exporter-service-account.yaml": nodeExporter.serviceAccount.new(namespace),
+ "node-exporter/node-exporter-service.yaml": nodeExporter.service.new(namespace),
+
+ "prometheus-operator/prometheus-operator-cluster-role-binding.yaml": po.clusterRoleBinding.new(namespace),
+ "prometheus-operator/prometheus-operator-cluster-role.yaml": po.clusterRole.new(),
+ "prometheus-operator/prometheus-operator-deployment.yaml": po.deployment.new(namespace),
+ "prometheus-operator/prometheus-operator-service.yaml": po.service.new(namespace),
+ "prometheus-operator/prometheus-operator-service-account.yaml": po.serviceAccount.new(namespace),
+
+ "prometheus-k8s/prometheus-k8s-cluster-role-binding.yaml": prometheus.clusterRoleBinding.new(namespace),
+ "prometheus-k8s/prometheus-k8s-cluster-role.yaml": prometheus.clusterRole.new(),
+ "prometheus-k8s/prometheus-k8s-service-account.yaml": prometheus.serviceAccount.new(namespace),
+ "prometheus-k8s/prometheus-k8s-service.yaml": prometheus.service.new(namespace),
+ "prometheus-k8s/prometheus-k8s.yaml": prometheus.prometheus.new(namespace),
+ "prometheus-k8s/prometheus-k8s-rules.yaml": prometheus.rules.new(namespace, ruleFiles),
+ "prometheus-k8s/prometheus-k8s-role-binding-config.yaml": prometheus.roleBindingConfig.new(namespace),
+ "prometheus-k8s/prometheus-k8s-role-binding-namespace.yaml": prometheus.roleBindingNamespace.new(namespace),
+ "prometheus-k8s/prometheus-k8s-role-binding-kube-system.yaml": prometheus.roleBindingKubeSystem.new(namespace),
+ "prometheus-k8s/prometheus-k8s-role-binding-default.yaml": prometheus.roleBindingDefault.new(namespace),
+ "prometheus-k8s/prometheus-k8s-role-config.yaml": prometheus.roleConfig.new(namespace),
+ "prometheus-k8s/prometheus-k8s-role-namespace.yaml": prometheus.roleNamespace.new(namespace),
+ "prometheus-k8s/prometheus-k8s-role-kube-system.yaml": prometheus.roleKubeSystem.new(),
+ "prometheus-k8s/prometheus-k8s-role-default.yaml": prometheus.roleDefault.new(),
+ "prometheus-k8s/prometheus-k8s-service-monitor-alertmanager.yaml": prometheus.serviceMonitorAlertmanager.new(namespace),
+ "prometheus-k8s/prometheus-k8s-service-monitor-apiserver.yaml": prometheus.serviceMonitorApiserver.new(namespace),
+ "prometheus-k8s/prometheus-k8s-service-monitor-coredns.yaml": prometheus.serviceMonitorCoreDNS.new(namespace),
+ "prometheus-k8s/prometheus-k8s-service-monitor-kube-controller-manager.yaml": prometheus.serviceMonitorControllerManager.new(namespace),
+ "prometheus-k8s/prometheus-k8s-service-monitor-kube-scheduler.yaml": prometheus.serviceMonitorScheduler.new(namespace),
+ "prometheus-k8s/prometheus-k8s-service-monitor-kube-state-metrics.yaml": prometheus.serviceMonitorKubeStateMetrics.new(namespace),
+ "prometheus-k8s/prometheus-k8s-service-monitor-kubelet.yaml": prometheus.serviceMonitorKubelet.new(namespace),
+ "prometheus-k8s/prometheus-k8s-service-monitor-node-exporter.yaml": prometheus.serviceMonitorNodeExporter.new(namespace),
+ "prometheus-k8s/prometheus-k8s-service-monitor-prometheus-operator.yaml": prometheus.serviceMonitorPrometheusOperator.new(namespace),
+ "prometheus-k8s/prometheus-k8s-service-monitor-prometheus.yaml": prometheus.serviceMonitorPrometheus.new(namespace),
+ }
+}
diff --git a/jsonnet/prometheus-operator/prometheus-operator-deployment.libsonnet b/jsonnet/prometheus-operator/prometheus-operator-deployment.libsonnet
index 2ad7f5260bebe19fdfd4a3bb3378be61f86d07c9..0212bc966e1e122b498717a56136b81bb7eefff9 100644
--- a/jsonnet/prometheus-operator/prometheus-operator-deployment.libsonnet
+++ b/jsonnet/prometheus-operator/prometheus-operator-deployment.libsonnet
@@ -2,7 +2,7 @@ local k = import "ksonnet.beta.3/k.libsonnet";
local rawVersion = importstr "../../../../VERSION";
local removeLineBreaks = function(str) std.join("", std.filter(function(c) c != "\n", std.stringChars(str)));
-local version = removeLineBreaks(rawVersion);
+local version = "v0.18.1";//removeLineBreaks(rawVersion);
local deployment = k.apps.v1beta2.deployment;
local container = k.apps.v1beta2.deployment.mixin.spec.template.spec.containersType;
@@ -12,7 +12,7 @@ local targetPort = 8080;
local podLabels = {"k8s-app": "prometheus-operator"};
local operatorContainer =
- container.new("prometheus-operator", "quay.io/coreos/prometheus-operator:v" + version) +
+ container.new("prometheus-operator", "quay.io/coreos/prometheus-operator:" + version) +
container.withPorts(containerPort.newNamed("http", targetPort)) +
container.withArgs(["--kubelet-service=kube-system/kubelet", "--config-reloader-image=quay.io/coreos/configmap-reload:v0.0.1"]) +
container.mixin.resources.withRequests({cpu: "100m", memory: "50Mi"}) +
diff --git a/jsonnet/prometheus/prometheus-k8s-rules.libsonnet b/jsonnet/prometheus/prometheus-k8s-rules.libsonnet
new file mode 100644
index 0000000000000000000000000000000000000000..abe98fa943dd632ff6fbda18a2ea7dd7af3f0183
--- /dev/null
+++ b/jsonnet/prometheus/prometheus-k8s-rules.libsonnet
@@ -0,0 +1,8 @@
+local k = import "ksonnet.beta.3/k.libsonnet";
+local configMap = k.core.v1.configMap;
+
+{
+ new(namespace, ruleFiles)::
+ configMap.new("prometheus-k8s-rules", ruleFiles) +
+ configMap.mixin.metadata.withNamespace(namespace)
+}
diff --git a/jsonnet/prometheus/prometheus-k8s-service.libsonnet b/jsonnet/prometheus/prometheus-k8s-service.libsonnet
index 96781d695ecc8e98d97e71733bb97a8057ee167b..add240ddd0ec37ee01f40299717c9f9e6f72fa7e 100644
--- a/jsonnet/prometheus/prometheus-k8s-service.libsonnet
+++ b/jsonnet/prometheus/prometheus-k8s-service.libsonnet
@@ -4,7 +4,6 @@ local servicePort = k.core.v1.service.mixin.spec.portsType;
local prometheusPort = servicePort.newNamed("web", 9090, "web");
-
{
new(namespace)::
service.new("prometheus-k8s", {app: "prometheus", prometheus: "k8s"}, prometheusPort) +
diff --git a/jsonnet/prometheus/prometheus.libsonnet b/jsonnet/prometheus/prometheus.libsonnet
index edc75c0893d5415c4c25ec7f78be4f1f20ef0a7d..99fb9f541ab5a6b8f7ccb72b6c1af96c3da48f0b 100644
--- a/jsonnet/prometheus/prometheus.libsonnet
+++ b/jsonnet/prometheus/prometheus.libsonnet
@@ -9,6 +9,7 @@
roleNamespace:: import "prometheus-k8s-role-namespace.libsonnet",
roleKubeSystem:: import "prometheus-k8s-role-kube-system.libsonnet",
roleDefault:: import "prometheus-k8s-role-default.libsonnet",
+ rules:: import "prometheus-k8s-rules.libsonnet",
serviceAccount:: import "prometheus-k8s-service-account.libsonnet",
serviceMonitorAlertmanager:: import "prometheus-k8s-service-monitor-alertmanager.libsonnet",
serviceMonitorApiserver:: import "prometheus-k8s-service-monitor-apiserver.libsonnet",
diff --git a/manifests/alertmanager/alertmanager-config.yaml b/manifests/alertmanager-main/alertmanager-main-secret.yaml
similarity index 91%
rename from manifests/alertmanager/alertmanager-config.yaml
rename to manifests/alertmanager-main/alertmanager-main-secret.yaml
index 62d39016214739e8f04a9158aad07df4a488013b..4a143fbb16883e5afd3375518d4abf12632398fe 100644
--- a/manifests/alertmanager/alertmanager-config.yaml
+++ b/manifests/alertmanager-main/alertmanager-main-secret.yaml
@@ -1,6 +1,8 @@
apiVersion: v1
+data:
+ alertmanager.yaml: Z2xvYmFsOgogIHJlc29sdmVfdGltZW91dDogNW0Kcm91dGU6CiAgZ3JvdXBfYnk6IFsnam9iJ10KICBncm91cF93YWl0OiAzMHMKICBncm91cF9pbnRlcnZhbDogNW0KICByZXBlYXRfaW50ZXJ2YWw6IDEyaAogIHJlY2VpdmVyOiAnbnVsbCcKICByb3V0ZXM6CiAgLSBtYXRjaDoKICAgICAgYWxlcnRuYW1lOiBEZWFkTWFuc1N3aXRjaAogICAgcmVjZWl2ZXI6ICdudWxsJwpyZWNlaXZlcnM6Ci0gbmFtZTogJ251bGwnCg==
kind: Secret
metadata:
name: alertmanager-main
-data:
- alertmanager.yaml: Z2xvYmFsOgogIHJlc29sdmVfdGltZW91dDogNW0Kcm91dGU6CiAgZ3JvdXBfYnk6IFsnam9iJ10KICBncm91cF93YWl0OiAzMHMKICBncm91cF9pbnRlcnZhbDogNW0KICByZXBlYXRfaW50ZXJ2YWw6IDEyaAogIHJlY2VpdmVyOiAnbnVsbCcKICByb3V0ZXM6CiAgLSBtYXRjaDoKICAgICAgYWxlcnRuYW1lOiBEZWFkTWFuc1N3aXRjaAogICAgcmVjZWl2ZXI6ICdudWxsJwpyZWNlaXZlcnM6Ci0gbmFtZTogJ251bGwnCg==
+ namespace: monitoring
+type: Opaque
diff --git a/manifests/alertmanager-main/alertmanager-main-service-account.yaml b/manifests/alertmanager-main/alertmanager-main-service-account.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..5c06d5e40cb4b3d99ad5388af1bb401458b63213
--- /dev/null
+++ b/manifests/alertmanager-main/alertmanager-main-service-account.yaml
@@ -0,0 +1,5 @@
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+ name: alertmanager-main
+ namespace: monitoring
diff --git a/manifests/alertmanager/alertmanager-service.yaml b/manifests/alertmanager-main/alertmanager-main-service.yaml
similarity index 78%
rename from manifests/alertmanager/alertmanager-service.yaml
rename to manifests/alertmanager-main/alertmanager-main-service.yaml
index a5413102dd60338d6dffffa29810d4abc0e5783b..0c7567932b707ebe7c30571f009ab7381b254ee4 100644
--- a/manifests/alertmanager/alertmanager-service.yaml
+++ b/manifests/alertmanager-main/alertmanager-main-service.yaml
@@ -4,13 +4,12 @@ metadata:
labels:
alertmanager: main
name: alertmanager-main
+ namespace: monitoring
spec:
- type: NodePort
ports:
- name: web
- nodePort: 30903
port: 9093
- protocol: TCP
targetPort: web
selector:
alertmanager: main
+ app: alertmanager
diff --git a/manifests/alertmanager/alertmanager.yaml b/manifests/alertmanager-main/alertmanager-main.yaml
similarity index 70%
rename from manifests/alertmanager/alertmanager.yaml
rename to manifests/alertmanager-main/alertmanager-main.yaml
index 3157d5874a5ec995fec3b9ebcaa44026ba5b27c9..84e72ec521f19dc69889b3ea36d9ab18db85b284 100644
--- a/manifests/alertmanager/alertmanager.yaml
+++ b/manifests/alertmanager-main/alertmanager-main.yaml
@@ -1,9 +1,11 @@
apiVersion: monitoring.coreos.com/v1
kind: Alertmanager
metadata:
- name: main
labels:
alertmanager: main
+ name: main
+ namespace: monitoring
spec:
replicas: 3
+ serviceAccountName: alertmanager-main
version: v0.14.0
diff --git a/manifests/etcd/etcd-bootkube-gce.yaml b/manifests/etcd/etcd-bootkube-gce.yaml
deleted file mode 100644
index 94e1c020b33f48906da12a50def47c441ab982cc..0000000000000000000000000000000000000000
--- a/manifests/etcd/etcd-bootkube-gce.yaml
+++ /dev/null
@@ -1,28 +0,0 @@
-apiVersion: v1
-kind: Service
-metadata:
- name: etcd-k8s
- labels:
- k8s-app: etcd
-spec:
- type: ClusterIP
- clusterIP: None
- ports:
- - name: api
- port: 2379
- protocol: TCP
----
-apiVersion: v1
-kind: Endpoints
-metadata:
- name: etcd-k8s
- labels:
- k8s-app: etcd
-subsets:
-- addresses:
- - ip: 10.142.0.2
- nodeName: 10.142.0.2
- ports:
- - name: api
- port: 2379
- protocol: TCP
diff --git a/manifests/etcd/etcd-bootkube-vagrant-multi.yaml b/manifests/etcd/etcd-bootkube-vagrant-multi.yaml
deleted file mode 100644
index 51293c60e96839b4586e11ad9c77033bb554d5b3..0000000000000000000000000000000000000000
--- a/manifests/etcd/etcd-bootkube-vagrant-multi.yaml
+++ /dev/null
@@ -1,28 +0,0 @@
-apiVersion: v1
-kind: Service
-metadata:
- name: etcd-k8s
- labels:
- k8s-app: etcd
-spec:
- type: ClusterIP
- clusterIP: None
- ports:
- - name: api
- port: 2379
- protocol: TCP
----
-apiVersion: v1
-kind: Endpoints
-metadata:
- name: etcd-k8s
- labels:
- k8s-app: etcd
-subsets:
-- addresses:
- - ip: 172.17.4.51
- nodeName: 172.17.4.51
- ports:
- - name: api
- port: 2379
- protocol: TCP
diff --git a/manifests/grafana/grafana-credentials.yaml b/manifests/grafana/grafana-credentials.yaml
deleted file mode 100644
index c3da1b631131dee0940720aa1f85aa62c7fa0f18..0000000000000000000000000000000000000000
--- a/manifests/grafana/grafana-credentials.yaml
+++ /dev/null
@@ -1,7 +0,0 @@
-apiVersion: v1
-kind: Secret
-metadata:
- name: grafana-credentials
-data:
- user: YWRtaW4=
- password: YWRtaW4=
diff --git a/manifests/grafana/grafana-dashboard-definitions.yaml b/manifests/grafana/grafana-dashboard-definitions.yaml
index dbd2a30d47eeb977dab15bdca78d9b95abe646f4..df4c5203b2537f72ed93f6d438ef9a188e7e2961 100644
--- a/manifests/grafana/grafana-dashboard-definitions.yaml
+++ b/manifests/grafana/grafana-dashboard-definitions.yaml
@@ -1,7360 +1,5403 @@
apiVersion: v1
-kind: ConfigMap
-metadata:
- name: grafana-dashboard-definitions-0
data:
- deployment-dashboard.json: |+
+ deployments-dashboard.json: |-
{
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "prometheus",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 1,
- "hideControls": false,
- "links": [],
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "200px",
- "panels": [
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
{
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 8,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "cores",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 4,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "sum(rate(container_cpu_usage_seconds_total{namespace=\"$deployment_namespace\",pod_name=~\"$deployment_name.*\"}[3m]))",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "CPU",
- "type": "singlestat",
- "valueFontSize": "110%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 2,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 4,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(rate(container_cpu_usage_seconds_total{namespace=\"$deployment_namespace\",pod_name=\u007e\"$deployment_name.*\"}[3m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "CPU",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 3,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 4,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(container_memory_usage_bytes{namespace=\"$deployment_namespace\",pod_name=\u007e\"$deployment_name.*\"}) / 1024^3",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Memory",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 4,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 4,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(rate(container_network_transmit_bytes_total{namespace=\"$deployment_namespace\",pod_name=\u007e\"$deployment_name.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{namespace=\"$deployment_namespace\",pod_name=\u007e\"$deployment_name.*\"}[3m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Network",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 9,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "GB",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "80%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 4,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "sum(container_memory_usage_bytes{namespace=\"$deployment_namespace\",pod_name=~\"$deployment_name.*\"}) / 1024^3",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Memory",
- "type": "singlestat",
- "valueFontSize": "110%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
+ "collapse": false,
+ "collapsed": false,
+ "height": "100px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 5,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "max(kube_deployment_spec_replicas{namespace=\"$deployment_namespace\",deployment=\"$deployment_name\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Desired Replicas",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 6,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "min(kube_deployment_status_replicas_available{namespace=\"$deployment_namespace\",deployment=\"$deployment_name\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Available Replicas",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 7,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "max(kube_deployment_status_observed_generation{namespace=\"$deployment_namespace\",deployment=\"$deployment_name\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Observed Generation",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 8,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "max(kube_deployment_metadata_generation{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Metadata Generation",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "Bps",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": false
- },
- "id": 7,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 9,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "max(kube_deployment_status_replicas{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "current replicas",
+ "refId": "A"
+ },
+ {
+ "expr": "min(kube_deployment_status_replicas_available{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "available",
+ "refId": "B"
+ },
+ {
+ "expr": "max(kube_deployment_status_replicas_unavailable{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "unavailable",
+ "refId": "C"
+ },
+ {
+ "expr": "min(kube_deployment_status_replicas_updated{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "updated",
+ "refId": "D"
+ },
+ {
+ "expr": "max(kube_deployment_spec_replicas{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "desired",
+ "refId": "E"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Replicas",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+
+ ],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Namespace",
+ "multi": false,
+ "name": "deployment_namespace",
+ "options": [
+
+ ],
+ "query": "label_values(kube_deployment_metadata_generation, namespace)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
},
{
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 4,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "sum(rate(container_network_transmit_bytes_total{namespace=\"$deployment_namespace\",pod_name=~\"$deployment_name.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{namespace=\"$deployment_namespace\",pod_name=~\"$deployment_name.*\"}[3m]))",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Network",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Name",
+ "multi": false,
+ "name": "deployment_name",
+ "options": [
+
+ ],
+ "query": "label_values(kube_deployment_metadata_generation{namespace=\"$deployment_namespace\"}, deployment)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Deployments",
+ "version": 0
+ }
+ kubernetes-capacity-planning-dashboard.json: |-
+ {
+ "annotations": {
+ "list": [
+
+ ]
},
- {
- "collapse": false,
- "editable": false,
- "height": "100px",
- "panels": [
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
{
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": false
- },
- "id": 5,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(kube_deployment_spec_replicas{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "metric": "kube_deployment_spec_replicas",
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Desired Replicas",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 2,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(node_cpu{mode=\"idle\"}[2m])) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{cpu}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Idle CPU",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 3,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(node_load1)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "load 1m",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(node_load5)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "load 5m",
+ "refId": "B"
+ },
+ {
+ "expr": "sum(node_load15)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "load 15m",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "System Load",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 6,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "min(kube_deployment_status_replicas_available{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Available Replicas",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(node_memory_MemTotal) - sum(node_memory_MemFree) - sum(node_memory_Buffers) - sum(node_memory_Cached)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory used",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(node_memory_Buffers)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory buffers",
+ "refId": "B"
+ },
+ {
+ "expr": "sum(node_memory_Cached)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory cached",
+ "refId": "C"
+ },
+ {
+ "expr": "sum(node_memory_MemFree)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory free",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory Usage",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 5,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "((sum(node_memory_MemTotal) - sum(node_memory_MemFree) - sum(node_memory_Buffers) - sum(node_memory_Cached)) / sum(node_memory_MemTotal)) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Memory Usage",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 3,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(kube_deployment_status_observed_generation{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Observed Generation",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 6,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(node_disk_bytes_read[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "read",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(rate(node_disk_bytes_written[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "written",
+ "refId": "B"
+ },
+ {
+ "expr": "sum(rate(node_disk_io_time_ms[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "io time",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Disk I/O",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 7,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "(sum(node_filesystem_size{device!=\"rootfs\"}) - sum(node_filesystem_free{device!=\"rootfs\"})) / sum(node_filesystem_size{device!=\"rootfs\"}) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Disk Space Usage",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 2,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(kube_deployment_metadata_generation{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Metadata Generation",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "350px",
- "panels": [
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 8,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(node_network_receive_bytes{device!\u007e\"lo\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network Received",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 9,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(node_network_transmit_bytes{device!\u007e\"lo\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network Transmitted",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 1,
- "isNew": true,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 12,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "max(kube_deployment_status_replicas{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "current replicas",
- "refId": "A",
- "step": 30
- },
- {
- "expr": "min(kube_deployment_status_replicas_available{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "available",
- "refId": "B",
- "step": 30
- },
- {
- "expr": "max(kube_deployment_status_replicas_unavailable{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "unavailable",
- "refId": "C",
- "step": 30
- },
- {
- "expr": "min(kube_deployment_status_replicas_updated{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "updated",
- "refId": "D",
- "step": 30
- },
- {
- "expr": "max(kube_deployment_spec_replicas{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "desired",
- "refId": "E",
- "step": 30
- }
- ],
- "title": "Replicas",
- "tooltip": {
- "msResolution": true,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "none",
- "label": "",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "label": "",
- "logBase": 1,
- "show": false
- }
- ]
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 10,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(kube_pod_info)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Current Number of Pods",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(kube_node_status_capacity_pods)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Maximum Capacity of Pods",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Cluster Pod Utilization",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 11,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "100 - (sum(kube_node_status_capacity_pods) - sum(kube_pod_info)) / sum(kube_node_status_capacity_pods) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Pod Utilization",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
}
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": [
- {
- "allValue": ".*",
- "current": {},
- "datasource": "prometheus",
- "hide": 0,
- "includeAll": false,
- "label": "Namespace",
- "multi": false,
- "name": "deployment_namespace",
- "options": [],
- "query": "label_values(kube_deployment_metadata_generation, namespace)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": null,
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {},
- "datasource": "prometheus",
- "hide": 0,
- "includeAll": false,
- "label": "Deployment",
- "multi": false,
- "name": "deployment_name",
- "options": [],
- "query": "label_values(kube_deployment_metadata_generation{namespace=\"$deployment_namespace\"}, deployment)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "deployment",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-6h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Deployment",
- "version": 1
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+
+ ],
+ "templating": {
+ "list": [
+
+ ]
+ },
+ "time": {
+ "from": "now-24h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Kubernetes Capacity Planning",
+ "version": 0
}
- etcd-dashboard.json: |+
+ kubernetes-cluster-health-dashboard.json: |-
{
- "__inputs": [
- {
- "name": "prometheus",
- "label": "prometheus",
- "description": "",
- "type": "datasource",
- "pluginId": "prometheus",
- "pluginName": "Prometheus"
- }
- ],
- "__requires": [
- {
- "type": "grafana",
- "id": "grafana",
- "name": "Grafana",
- "version": "4.5.2"
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "10s",
+ "rows": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 2,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(up{job=\u007e\"apiserver|kube-scheduler|kube-controller-manager\"} == 0)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Control Plane Components Down",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "Everything UP and healthy",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 3,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(ALERTS{alertstate=\"firing\",alertname!=\"DeadMansSwitch\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Alerts Firing",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 4,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(ALERTS{alertstate=\"pending\",alertname!=\"DeadMansSwitch\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Alerts Pending",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 5,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "count(increase(kube_pod_container_status_restarts[1h]) > 5)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Crashlooping Pods",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 6,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(kube_node_status_condition{condition=\"Ready\",status!=\"true\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Node Not Ready",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 7,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(kube_node_status_condition{condition=\"DiskPressure\",status=\"true\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Node Disk Pressure",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 8,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(kube_node_status_condition{condition=\"MemoryPressure\",status=\"true\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Node Memory Pressure",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 9,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(kube_node_spec_unschedulable)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Node Unschedulable",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+
+ ],
+ "templating": {
+ "list": [
+
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
},
- {
- "type": "panel",
- "id": "graph",
- "name": "Graph",
- "version": ""
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
},
- {
- "type": "datasource",
- "id": "prometheus",
- "name": "Prometheus",
- "version": "1.0.0"
+ "timezone": "browser",
+ "title": "Kubernetes Cluster Health",
+ "version": 0
+ }
+ kubernetes-cluster-status-dashboard.json: |-
+ {
+ "annotations": {
+ "list": [
+
+ ]
},
- {
- "type": "panel",
- "id": "singlestat",
- "name": "Singlestat",
- "version": ""
- }
- ],
- "annotations": {
- "list": []
- },
- "description": "etcd sample Grafana dashboard with Prometheus",
- "editable": false,
- "gnetId": null,
- "graphTooltip": 0,
- "hideControls": false,
- "id": null,
- "links": [],
- "refresh": false,
- "rows": [
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "10s",
+ "rows": [
{
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 28,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum(etcd_server_has_leader)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "metric": "etcd_server_has_leader",
- "refId": "A",
- "step": 20
- }
- ],
- "thresholds": "",
- "title": "Up",
- "type": "singlestat",
- "valueFontSize": "200%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 2,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 6,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(up{job=\u007e\"apiserver|kube-scheduler|kube-controller-manager\"} == 0)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Control Plane UP",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 3,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 6,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(ALERTS{alertstate=\"firing\",alertname!=\"DeadMansSwitch\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Alerts Firing",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 0,
- "id": 23,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 5,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(grpc_server_started_total{grpc_type=\"unary\"}[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "RPC Rate",
- "metric": "grpc_server_started_total",
- "refId": "A",
- "step": 4
- },
- {
- "expr": "sum(rate(grpc_server_handled_total{grpc_type=\"unary\",grpc_code!=\"OK\"}[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "RPC Failed Rate",
- "metric": "grpc_server_handled_total",
- "refId": "B",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "RPC Rate",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "ops",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 4,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "(sum(up{job=\"apiserver\"} == 1) / count(up{job=\"apiserver\"})) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "50, 80",
+ "title": "API Servers UP",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 5,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "(sum(up{job=\"kube-controller-manager\"} == 1) / count(up{job=\"kube-controller-manager\"})) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "50, 80",
+ "title": "Controller Managers UP",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 6,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "(sum(up{job=\"kube-scheduler\"} == 1) / count(up{job=\"kube-scheduler\"})) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "50, 80",
+ "title": "Schedulers Up",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 7,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "count(increase(kube_pod_container_status_restarts{namespace=\u007e\"kube-system|tectonic-system\"}[1h]) > 5)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Crashlooping Control Plane Pods",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 0,
- "id": 41,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 4,
- "stack": true,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(grpc_server_started_total{grpc_service=\"etcdserverpb.Watch\",grpc_type=\"bidi_stream\"}) - sum(grpc_server_handled_total{grpc_service=\"etcdserverpb.Watch\",grpc_type=\"bidi_stream\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Watch Streams",
- "metric": "grpc_server_handled_total",
- "refId": "A",
- "step": 4
- },
- {
- "expr": "sum(grpc_server_started_total{grpc_service=\"etcdserverpb.Lease\",grpc_type=\"bidi_stream\"}) - sum(grpc_server_handled_total{grpc_service=\"etcdserverpb.Lease\",grpc_type=\"bidi_stream\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Lease Streams",
- "metric": "grpc_server_handled_total",
- "refId": "B",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Active Streams",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "label": "",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 8,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(100 - (avg by (instance) (rate(node_cpu{job=\"node-exporter\",mode=\"idle\"}[5m])) * 100)) / count(node_cpu{job=\"node-exporter\",mode=\"idle\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "CPU Utilization",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 9,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "((sum(node_memory_MemTotal) - sum(node_memory_MemFree) - sum(node_memory_Buffers) - sum(node_memory_Cached)) / sum(node_memory_MemTotal)) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Memory Utilization",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 10,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "(sum(node_filesystem_size{device!=\"rootfs\"}) - sum(node_filesystem_free{device!=\"rootfs\"})) / sum(node_filesystem_size{device!=\"rootfs\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Filesystem Utilization",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 11,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "100 - (sum(kube_node_status_capacity_pods) - sum(kube_pod_info)) / sum(kube_node_status_capacity_pods) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Pod Utilization",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
}
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "Row",
- "titleSize": "h6"
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+
+ ],
+ "templating": {
+ "list": [
+
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
},
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Kubernetes Cluster Status",
+ "version": 0
+ }
+ kubernetes-kubelet-dashboard.json: |-
+ {
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "decimals": null,
- "editable": false,
- "error": false,
- "fill": 0,
- "grid": {},
- "id": 1,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 4,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "etcd_debugging_mvcc_db_total_size_in_bytes",
- "format": "time_series",
- "hide": false,
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} DB Size",
- "metric": "",
- "refId": "A",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "DB Size",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ]
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 2,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(kubelet_running_pod_count{instance=\u007e\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Count",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 3,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 10,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "kubelet_running_pod_count{instance=\u007e\"$instance\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{ instance }}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Count",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Pods",
+ "titleSize": "h4",
+ "type": "row"
},
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 0,
- "grid": {},
- "id": 3,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 1,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 4,
- "stack": false,
- "steppedLine": true,
- "targets": [
- {
- "expr": "histogram_quantile(0.99, sum(rate(etcd_disk_wal_fsync_duration_seconds_bucket[5m])) by (instance, le))",
- "format": "time_series",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "{{instance}} WAL fsync",
- "metric": "etcd_disk_wal_fsync_duration_seconds_bucket",
- "refId": "A",
- "step": 4
- },
- {
- "expr": "histogram_quantile(0.99, sum(rate(etcd_disk_backend_commit_duration_seconds_bucket[5m])) by (instance, le))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} DB fsync",
- "metric": "etcd_disk_backend_commit_duration_seconds_bucket",
- "refId": "B",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Disk Sync Duration",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "s",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ]
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "prometheus",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 4,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(kubelet_running_container_count{instance=\u007e\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "",
+ "title": "Count",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 5,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 10,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "kubelet_running_container_count{instance=\u007e\"$instance\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{ instance }}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Count",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Containers",
+ "titleSize": "h4",
+ "type": "row"
},
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 0,
- "id": 29,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 4,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "process_resident_memory_bytes",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} Resident Memory",
- "metric": "process_resident_memory_bytes",
- "refId": "A",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Memory",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "description": "Rate of Kubelet Operations in 5min",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 6,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(kubelet_runtime_operations{instance=\u007e\"$instance\"}[5m])) by (instance)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{ instance }}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Operations",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Kubelet",
+ "titleSize": "h4",
+ "type": "row"
}
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "New row",
- "titleSize": "h6"
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+
+ ],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": true,
+ "label": null,
+ "multi": false,
+ "name": "instance",
+ "options": [
+
+ ],
+ "query": "label_values(kubelet_running_pod_count,instance)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
},
- {
- "collapse": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 5,
- "id": 22,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 3,
- "stack": true,
- "steppedLine": false,
- "targets": [
- {
- "expr": "rate(etcd_network_client_grpc_received_bytes_total[5m])",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} Client Traffic In",
- "metric": "etcd_network_client_grpc_received_bytes_total",
- "refId": "A",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Client Traffic In",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "Bps",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Kubelet",
+ "version": 0
+ }
+ nodes.json: |-
+ {
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 5,
- "id": 21,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 3,
- "stack": true,
- "steppedLine": false,
- "targets": [
- {
- "expr": "rate(etcd_network_client_grpc_sent_bytes_total[5m])",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} Client Traffic Out",
- "metric": "etcd_network_client_grpc_sent_bytes_total",
- "refId": "A",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Client Traffic Out",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "Bps",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 2,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "100 - (avg by (cpu) (irate(node_cpu{mode=\"idle\", instance=\"$server\"}[5m])) * 100)",
+ "format": "time_series",
+ "intervalFactor": 10,
+ "legendFormat": "{{cpu}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Idle CPU",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": 100,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": 100,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 3,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node_load1{instance=\"$server\"} * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "load 1m",
+ "refId": "A"
+ },
+ {
+ "expr": "node_load5{instance=\"$server\"} * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "load 5m",
+ "refId": "B"
+ },
+ {
+ "expr": "node_load15{instance=\"$server\"} * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "load 15m",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "System load",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 0,
- "id": 20,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 3,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(etcd_network_peer_received_bytes_total[5m])) by (instance)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} Peer Traffic In",
- "metric": "etcd_network_peer_received_bytes_total",
- "refId": "A",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Peer Traffic In",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "Bps",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node_memory_MemTotal{instance=\"$server\"} - node_memory_MemFree{instance=\"$server\"} - node_memory_Buffers{instance=\"$server\"} - node_memory_Cached{instance=\"$server\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory used",
+ "refId": "A"
+ },
+ {
+ "expr": "node_memory_Buffers{instance=\"$server\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory buffers",
+ "refId": "B"
+ },
+ {
+ "expr": "node_memory_Cached{instance=\"$server\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory cached",
+ "refId": "C"
+ },
+ {
+ "expr": "node_memory_MemFree{instance=\"$server\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "memory free",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory Usage",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 5,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "((node_memory_MemTotal{instance=\"$server\"} - node_memory_MemFree{instance=\"$server\"} - node_memory_Buffers{instance=\"$server\"} - node_memory_Cached{instance=\"$server\"}) / node_memory_MemTotal{instance=\"$server\"}) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Memory Usage",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "decimals": null,
- "editable": false,
- "error": false,
- "fill": 0,
- "grid": {},
- "id": 16,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 3,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(etcd_network_peer_sent_bytes_total[5m])) by (instance)",
- "format": "time_series",
- "hide": false,
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} Peer Traffic Out",
- "metric": "etcd_network_peer_sent_bytes_total",
- "refId": "A",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Peer Traffic Out",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "Bps",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "New row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 0,
- "id": 40,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(etcd_server_proposals_failed_total[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Proposal Failure Rate",
- "metric": "etcd_server_proposals_failed_total",
- "refId": "A",
- "step": 2
- },
- {
- "expr": "sum(etcd_server_proposals_pending)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Proposal Pending Total",
- "metric": "etcd_server_proposals_pending",
- "refId": "B",
- "step": 2
- },
- {
- "expr": "sum(rate(etcd_server_proposals_committed_total[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Proposal Commit Rate",
- "metric": "etcd_server_proposals_committed_total",
- "refId": "C",
- "step": 2
- },
- {
- "expr": "sum(rate(etcd_server_proposals_applied_total[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Proposal Apply Rate",
- "refId": "D",
- "step": 2
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Raft Proposals",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "label": "",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 6,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "read",
+ "yaxis": 1
+ },
+ {
+ "alias": "io time",
+ "yaxis": 2
+ }
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by (instance) (rate(node_disk_bytes_read{instance=\"$server\"}[2m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "read",
+ "refId": "A"
+ },
+ {
+ "expr": "sum by (instance) (rate(node_disk_bytes_written{instance=\"$server\"}[2m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "written",
+ "refId": "B"
+ },
+ {
+ "expr": "sum by (instance) (rate(node_disk_io_time_ms{instance=\"$server\"}[2m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "io time",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Disk I/O",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "ms",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+
+ },
+ "id": 7,
+ "interval": null,
+ "links": [
+
+ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "(sum(node_filesystem_size{device!=\"rootfs\",instance=\"$server\"}) - sum(node_filesystem_free{device!=\"rootfs\",instance=\"$server\"})) / sum(node_filesystem_size{device!=\"rootfs\",instance=\"$server\"}) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Disk Space Usage",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "decimals": 0,
- "editable": false,
- "error": false,
- "fill": 0,
- "id": 19,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "changes(etcd_server_leader_changes_seen_total[1d])",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} Total Leader Elections Per Day",
- "metric": "etcd_server_leader_changes_seen_total",
- "refId": "A",
- "step": 2
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Total Leader Elections Per Day",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 8,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(node_network_receive_bytes{instance=\"$server\",device!\u007e\"lo\"}[5m])",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{device}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network Received",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 9,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(node_network_transmit_bytes{instance=\"$server\",device!\u007e\"lo\"}[5m])",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{device}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network Transmitted",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
}
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "New row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": []
- },
- "time": {
- "from": "now-15m",
- "to": "now"
- },
- "timepicker": {
- "now": true,
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "etcd",
- "version": 4
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+
+ ],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": null,
+ "multi": false,
+ "name": "server",
+ "options": [
+
+ ],
+ "query": "label_values(node_boot_time, instance)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Nodes",
+ "version": 0
}
- kubernetes-capacity-planning-dashboard.json: |+
+ pods-dashboard.json: |-
{
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "prometheus",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "gnetId": 22,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [],
- "refresh": false,
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
+ "annotations": {
+ "list": [
+
+ ]
+ },
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [
+
+ ],
+ "refresh": "",
+ "rows": [
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 3,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(node_cpu{mode=\"idle\"}[2m])) * 100",
- "hide": false,
- "intervalFactor": 10,
- "legendFormat": "",
- "refId": "A",
- "step": 50
- }
- ],
- "title": "Idle CPU",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "percent",
- "label": "cpu usage",
- "logBase": 1,
- "min": 0,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 2,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by(container_name) (container_memory_usage_bytes{namespace=\"$namespace\", pod_name=\"$pod\", container_name=\u007e\"$container\", container_name!=\"POD\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Current: {{ container_name }}",
+ "refId": "A"
+ },
+ {
+ "expr": "sum by(container) (kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\", pod=\"$pod\", container=\u007e\"$container\", container!=\"POD\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Requested: {{ container }}",
+ "refId": "B"
+ },
+ {
+ "expr": "sum by(container) (kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\", pod=\"$pod\", container=\u007e\"$container\", container!=\"POD\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Limit: {{ container }}",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory Usage",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
},
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 9,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(node_load1)",
- "intervalFactor": 4,
- "legendFormat": "load 1m",
- "refId": "A",
- "step": 20,
- "target": ""
- },
- {
- "expr": "sum(node_load5)",
- "intervalFactor": 4,
- "legendFormat": "load 5m",
- "refId": "B",
- "step": 20,
- "target": ""
- },
- {
- "expr": "sum(node_load15)",
- "intervalFactor": 4,
- "legendFormat": "load 15m",
- "refId": "C",
- "step": 20,
- "target": ""
- }
- ],
- "title": "System Load",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "percentunit",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 3,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by (container_name) (rate(container_cpu_usage_seconds_total{image!=\"\",container_name!=\"POD\",pod_name=\"$pod\"}[1m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{ container_name }}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU Usage",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ },
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 4,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "node_memory_SwapFree{instance=\"172.17.0.1:9100\",job=\"prometheus\"}",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 9,
- "stack": true,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(node_memory_MemTotal) - sum(node_memory_MemFree) - sum(node_memory_Buffers) - sum(node_memory_Cached)",
- "intervalFactor": 2,
- "legendFormat": "memory usage",
- "metric": "memo",
- "refId": "A",
- "step": 10,
- "target": ""
+ "collapse": false,
+ "collapsed": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {
+
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "fill": 1,
+ "gridPos": {
+
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+
+ ],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sort_desc(sum by (pod_name) (rate(container_network_receive_bytes_total{pod_name=\"$pod\"}[1m])))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{ pod_name }}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [
+
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network I/O",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [
+
+ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6",
+ "type": "row"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [
+
+ ],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Namespace",
+ "multi": false,
+ "name": "namespace",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info, namespace)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
},
{
- "expr": "sum(node_memory_Buffers)",
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "memory buffers",
- "metric": "memo",
- "refId": "B",
- "step": 10,
- "target": ""
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Pod",
+ "multi": false,
+ "name": "pod",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info{namespace=\u007e\"$namespace\"}, pod)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
},
{
- "expr": "sum(node_memory_Cached)",
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "memory cached",
- "metric": "memo",
- "refId": "C",
- "step": 10,
- "target": ""
- },
- {
- "expr": "sum(node_memory_MemFree)",
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "memory free",
- "metric": "memo",
- "refId": "D",
- "step": 10,
- "target": ""
- }
- ],
- "title": "Memory Usage",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "min": "0",
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 5,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "((sum(node_memory_MemTotal) - sum(node_memory_MemFree) - sum(node_memory_Buffers) - sum(node_memory_Cached)) / sum(node_memory_MemTotal)) * 100",
- "intervalFactor": 2,
- "metric": "",
- "refId": "A",
- "step": 60,
- "target": ""
- }
- ],
- "thresholds": "80, 90",
- "title": "Memory Usage",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "246px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 6,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "read",
- "yaxis": 1
- },
- {
- "alias": "{instance=\"172.17.0.1:9100\"}",
- "yaxis": 2
- },
- {
- "alias": "io time",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 9,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(node_disk_bytes_read[5m]))",
- "hide": false,
- "intervalFactor": 4,
- "legendFormat": "read",
- "refId": "A",
- "step": 20,
- "target": ""
- },
- {
- "expr": "sum(rate(node_disk_bytes_written[5m]))",
- "intervalFactor": 4,
- "legendFormat": "written",
- "refId": "B",
- "step": 20
- },
- {
- "expr": "sum(rate(node_disk_io_time_ms[5m]))",
- "intervalFactor": 4,
- "legendFormat": "io time",
- "refId": "C",
- "step": 20
- }
- ],
- "title": "Disk I/O",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "ms",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percentunit",
- "gauge": {
- "maxValue": 1,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 12,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(node_filesystem_size{device!=\"rootfs\"}) - sum(node_filesystem_free{device!=\"rootfs\"})) / sum(node_filesystem_size{device!=\"rootfs\"})",
- "intervalFactor": 2,
- "refId": "A",
- "step": 60,
- "target": ""
- }
- ],
- "thresholds": "0.75, 0.9",
- "title": "Disk Space Usage",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 8,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "transmitted",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(node_network_receive_bytes{device!~\"lo\"}[5m]))",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 10,
- "target": ""
- }
- ],
- "title": "Network Received",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 10,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "transmitted",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(node_network_transmit_bytes{device!~\"lo\"}[5m]))",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "B",
- "step": 10,
- "target": ""
- }
- ],
- "title": "Network Transmitted",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "276px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 11,
- "isNew": true,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 11,
- "span": 9,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(kube_pod_info)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Current number of Pods",
- "refId": "A",
- "step": 10
- },
- {
- "expr": "sum(kube_node_status_capacity_pods)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Maximum capacity of pods",
- "refId": "B",
- "step": 10
- }
- ],
- "title": "Cluster Pod Utilization",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 7,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "100 - (sum(kube_node_status_capacity_pods) - sum(kube_pod_info)) / sum(kube_node_status_capacity_pods) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 60,
- "target": ""
- }
- ],
- "thresholds": "80, 90",
- "title": "Pod Utilization",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": []
- },
- "time": {
- "from": "now-1h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Kubernetes Capacity Planning",
- "version": 4
- }
- kubernetes-cluster-health-dashboard.json: |+
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "prometheus",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [],
- "refresh": "10s",
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "254px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 1,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(up{job=~\"apiserver|kube-scheduler|kube-controller-manager\"} == 0)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Control Plane Components Down",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "Everything UP and healthy",
- "value": "null"
- },
- {
- "op": "=",
- "text": "",
- "value": ""
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 2,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(ALERTS{alertstate=\"firing\",alertname!=\"DeadMansSwitch\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Alerts Firing",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 3,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(ALERTS{alertstate=\"pending\",alertname!=\"DeadMansSwitch\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "3, 5",
- "title": "Alerts Pending",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 4,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "count(increase(kube_pod_container_status_restarts[1h]) > 5)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Crashlooping Pods",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": false,
- "title": "Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 5,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(kube_node_status_condition{condition=\"Ready\",status!=\"true\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Node Not Ready",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 6,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(kube_node_status_condition{condition=\"DiskPressure\",status=\"true\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Node Disk Pressure",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 7,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(kube_node_status_condition{condition=\"MemoryPressure\",status=\"true\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Node Memory Pressure",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 8,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(kube_node_spec_unschedulable)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Nodes Unschedulable",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": false,
- "title": "Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": []
- },
- "time": {
- "from": "now-6h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Kubernetes Cluster Health",
- "version": 9
- }
- kubernetes-cluster-status-dashboard.json: |+
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "prometheus",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [],
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "129px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 5,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 6,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(up{job=~\"apiserver|kube-scheduler|kube-controller-manager\"} == 0)",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Control Plane UP",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "UP",
- "value": "null"
- }
- ],
- "valueName": "total"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 6,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 6,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(ALERTS{alertstate=\"firing\",alertname!=\"DeadMansSwitch\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "3, 5",
- "title": "Alerts Firing",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": true,
- "title": "Cluster Health",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "168px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 1,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(up{job=\"apiserver\"} == 1) / count(up{job=\"apiserver\"})) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "50, 80",
- "title": "API Servers UP",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 2,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(up{job=\"kube-controller-manager\"} == 1) / count(up{job=\"kube-controller-manager\"})) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "50, 80",
- "title": "Controller Managers UP",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 3,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(up{job=\"kube-scheduler\"} == 1) / count(up{job=\"kube-scheduler\"})) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "50, 80",
- "title": "Schedulers UP",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 4,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "count(increase(kube_pod_container_status_restarts{namespace=~\"kube-system|tectonic-system\"}[1h]) > 5)",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Crashlooping Control Plane Pods",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": true,
- "title": "Control Plane Status",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "158px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 8,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(100 - (avg by (instance) (rate(node_cpu{job=\"node-exporter\",mode=\"idle\"}[5m])) * 100)) / count(node_cpu{job=\"node-exporter\",mode=\"idle\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "80, 90",
- "title": "CPU Utilization",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 7,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "((sum(node_memory_MemTotal) - sum(node_memory_MemFree) - sum(node_memory_Buffers) - sum(node_memory_Cached)) / sum(node_memory_MemTotal)) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "80, 90",
- "title": "Memory Utilization",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 9,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(node_filesystem_size{device!=\"rootfs\"}) - sum(node_filesystem_free{device!=\"rootfs\"})) / sum(node_filesystem_size{device!=\"rootfs\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "80, 90",
- "title": "Filesystem Utilization",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 10,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "100 - (sum(kube_node_status_capacity_pods) - sum(kube_pod_info)) / sum(kube_node_status_capacity_pods) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "80, 90",
- "title": "Pod Utilization",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": true,
- "title": "Capacity Planning",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": []
- },
- "time": {
- "from": "now-6h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Kubernetes Cluster Status",
- "version": 3
- }
- kubernetes-control-plane-status-dashboard.json: |+
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "prometheus",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [],
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 1,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(up{job=\"apiserver\"} == 1) / sum(up{job=\"apiserver\"})) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "50, 80",
- "title": "API Servers UP",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 2,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(up{job=\"kube-controller-manager\"} == 1) / sum(up{job=\"kube-controller-manager\"})) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "50, 80",
- "title": "Controller Managers UP",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 3,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(up{job=\"kube-scheduler\"} == 1) / sum(up{job=\"kube-scheduler\"})) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "50, 80",
- "title": "Schedulers UP",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 4,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(sum by(instance) (rate(apiserver_request_count{code=~\"5..\"}[5m])) / sum by(instance) (rate(apiserver_request_count[5m]))) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "5, 10",
- "title": "API Server Request Error Rate",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 7,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 12,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum by(verb) (rate(apiserver_latency_seconds:quantile[5m]) >= 0)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 30
- }
- ],
- "title": "API Server Request Latency",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 5,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "cluster:scheduler_e2e_scheduling_latency_seconds:quantile",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 60
- }
- ],
- "title": "End to End Scheduling Latency",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "logBase": 1,
- "show": true
- },
- {
- "format": "dtdurations",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 6,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum by(instance) (rate(apiserver_request_count{code!~\"2..\"}[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Error Rate",
- "refId": "A",
- "step": 60
- },
- {
- "expr": "sum by(instance) (rate(apiserver_request_count[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Request Rate",
- "refId": "B",
- "step": 60
- }
- ],
- "title": "API Server Request Rates",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": []
- },
- "time": {
- "from": "now-6h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Kubernetes Control Plane Status",
- "version": 3
- }
- kubernetes-resource-requests-dashboard.json: |+
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "prometheus",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [],
- "refresh": false,
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "300px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "description": "This represents the total [CPU resource requests](https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/#meaning-of-cpu) in the cluster.\nFor comparison the total [allocatable CPU cores](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/node-allocatable.md) is also shown.",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 1,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 9,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "min(sum(kube_node_status_allocatable_cpu_cores) by (instance))",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "Allocatable CPU Cores",
- "refId": "A",
- "step": 20
- },
- {
- "expr": "max(sum(kube_pod_container_resource_requests_cpu_cores) by (instance))",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "Requested CPU Cores",
- "refId": "B",
- "step": 20
- }
- ],
- "title": "CPU Cores",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "label": "CPU Cores",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 2,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "max(sum(kube_pod_container_resource_requests_cpu_cores) by (instance)) / min(sum(kube_node_status_allocatable_cpu_cores) by (instance)) * 100",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 240
- }
- ],
- "thresholds": "80, 90",
- "title": "CPU Cores",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "110%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "CPU Cores",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "300px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "description": "This represents the total [memory resource requests](https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/#meaning-of-memory) in the cluster.\nFor comparison the total [allocatable memory](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/node-allocatable.md) is also shown.",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 3,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 9,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "min(sum(kube_node_status_allocatable_memory_bytes) by (instance))",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "Allocatable Memory",
- "refId": "A",
- "step": 20
- },
- {
- "expr": "max(sum(kube_pod_container_resource_requests_memory_bytes) by (instance))",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "Requested Memory",
- "refId": "B",
- "step": 20
- }
- ],
- "title": "Memory",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "label": "Memory",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 4,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "max(sum(kube_pod_container_resource_requests_memory_bytes) by (instance)) / min(sum(kube_node_status_allocatable_memory_bytes) by (instance)) * 100",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 240
- }
- ],
- "thresholds": "80, 90",
- "title": "Memory",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "110%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "Memory",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": []
- },
- "time": {
- "from": "now-3h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Kubernetes Resource Requests",
- "version": 2
- }
- nodes-dashboard.json: |+
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "prometheus",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "description": "Dashboard to get an overview of one server",
- "editable": false,
- "gnetId": 22,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [],
- "refresh": false,
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 3,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "100 - (avg by (cpu) (irate(node_cpu{mode=\"idle\", instance=\"$server\"}[5m])) * 100)",
- "hide": false,
- "intervalFactor": 10,
- "legendFormat": "{{cpu}}",
- "refId": "A",
- "step": 50
- }
- ],
- "title": "Idle CPU",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "percent",
- "label": "cpu usage",
- "logBase": 1,
- "max": 100,
- "min": 0,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 9,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "node_load1{instance=\"$server\"}",
- "intervalFactor": 4,
- "legendFormat": "load 1m",
- "refId": "A",
- "step": 20,
- "target": ""
- },
- {
- "expr": "node_load5{instance=\"$server\"}",
- "intervalFactor": 4,
- "legendFormat": "load 5m",
- "refId": "B",
- "step": 20,
- "target": ""
- },
- {
- "expr": "node_load15{instance=\"$server\"}",
- "intervalFactor": 4,
- "legendFormat": "load 15m",
- "refId": "C",
- "step": 20,
- "target": ""
- }
- ],
- "title": "System Load",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "percentunit",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 4,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "node_memory_SwapFree{instance=\"172.17.0.1:9100\",job=\"prometheus\"}",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 9,
- "stack": true,
- "steppedLine": false,
- "targets": [
- {
- "expr": "node_memory_MemTotal{instance=\"$server\"} - node_memory_MemFree{instance=\"$server\"} - node_memory_Buffers{instance=\"$server\"} - node_memory_Cached{instance=\"$server\"}",
- "hide": false,
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "memory used",
- "metric": "",
- "refId": "C",
- "step": 10
- },
- {
- "expr": "node_memory_Buffers{instance=\"$server\"}",
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "memory buffers",
- "metric": "",
- "refId": "E",
- "step": 10
- },
- {
- "expr": "node_memory_Cached{instance=\"$server\"}",
- "intervalFactor": 2,
- "legendFormat": "memory cached",
- "metric": "",
- "refId": "F",
- "step": 10
- },
- {
- "expr": "node_memory_MemFree{instance=\"$server\"}",
- "intervalFactor": 2,
- "legendFormat": "memory free",
- "metric": "",
- "refId": "D",
- "step": 10
- }
- ],
- "title": "Memory Usage",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "min": "0",
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 5,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "((node_memory_MemTotal{instance=\"$server\"} - node_memory_MemFree{instance=\"$server\"} - node_memory_Buffers{instance=\"$server\"} - node_memory_Cached{instance=\"$server\"}) / node_memory_MemTotal{instance=\"$server\"}) * 100",
- "intervalFactor": 2,
- "refId": "A",
- "step": 60,
- "target": ""
- }
- ],
- "thresholds": "80, 90",
- "title": "Memory Usage",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 6,
- "isNew": true,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "read",
- "yaxis": 1
- },
- {
- "alias": "{instance=\"172.17.0.1:9100\"}",
- "yaxis": 2
- },
- {
- "alias": "io time",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 9,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum by (instance) (rate(node_disk_bytes_read{instance=\"$server\"}[2m]))",
- "hide": false,
- "intervalFactor": 4,
- "legendFormat": "read",
- "refId": "A",
- "step": 20,
- "target": ""
- },
- {
- "expr": "sum by (instance) (rate(node_disk_bytes_written{instance=\"$server\"}[2m]))",
- "intervalFactor": 4,
- "legendFormat": "written",
- "refId": "B",
- "step": 20
- },
- {
- "expr": "sum by (instance) (rate(node_disk_io_time_ms{instance=\"$server\"}[2m]))",
- "intervalFactor": 4,
- "legendFormat": "io time",
- "refId": "C",
- "step": 20
- }
- ],
- "title": "Disk I/O",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "ms",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "percentunit",
- "gauge": {
- "maxValue": 1,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 7,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(node_filesystem_size{device!=\"rootfs\",instance=\"$server\"}) - sum(node_filesystem_free{device!=\"rootfs\",instance=\"$server\"})) / sum(node_filesystem_size{device!=\"rootfs\",instance=\"$server\"})",
- "intervalFactor": 2,
- "refId": "A",
- "step": 60,
- "target": ""
- }
- ],
- "thresholds": "0.75, 0.9",
- "title": "Disk Space Usage",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 8,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "transmitted",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "rate(node_network_receive_bytes{instance=\"$server\",device!~\"lo\"}[5m])",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "{{device}}",
- "refId": "A",
- "step": 10,
- "target": ""
- }
- ],
- "title": "Network Received",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 10,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "transmitted",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "rate(node_network_transmit_bytes{instance=\"$server\",device!~\"lo\"}[5m])",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "{{device}}",
- "refId": "B",
- "step": 10,
- "target": ""
- }
- ],
- "title": "Network Transmitted",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": [
- {
- "allValue": null,
- "current": {},
- "datasource": "prometheus",
- "hide": 0,
- "includeAll": false,
- "label": null,
- "multi": false,
- "name": "server",
- "options": [],
- "query": "label_values(node_boot_time, instance)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-1h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Nodes",
- "version": 2
- }
- pods-dashboard.json: |+
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "prometheus",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 1,
- "hideControls": false,
- "links": [],
- "refresh": false,
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 1,
- "isNew": false,
- "legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": true
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 12,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum by(container_name) (container_memory_usage_bytes{pod_name=\"$pod\", container_name=~\"$container\", container_name!=\"POD\"})",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "Current: {{ container_name }}",
- "metric": "container_memory_usage_bytes",
- "refId": "A",
- "step": 15
- },
- {
- "expr": "kube_pod_container_resource_requests_memory_bytes{pod=\"$pod\", container=~\"$container\"}",
- "interval": "10s",
- "intervalFactor": 2,
- "legendFormat": "Requested: {{ container }}",
- "metric": "kube_pod_container_resource_requests_memory_bytes",
- "refId": "B",
- "step": 20
- },
- {
- "expr": "kube_pod_container_resource_limits_memory_bytes{pod=\"$pod\", container=~\"$container\"}",
- "interval": "10s",
- "intervalFactor": 2,
- "legendFormat": "Limit: {{ container }}",
- "metric": "kube_pod_container_resource_limits_memory_bytes",
- "refId": "C",
- "step": 20
- }
- ],
- "title": "Memory Usage",
- "tooltip": {
- "msResolution": true,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 2,
- "isNew": false,
- "legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": true
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 12,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum by (container_name)(rate(container_cpu_usage_seconds_total{image!=\"\",container_name!=\"POD\",pod_name=\"$pod\"}[1m]))",
- "intervalFactor": 2,
- "legendFormat": "{{ container_name }}",
- "refId": "A",
- "step": 30
- },
- {
- "expr": "kube_pod_container_resource_requests_cpu_cores{pod=\"$pod\", container=~\"$container\"}",
- "interval": "10s",
- "intervalFactor": 2,
- "legendFormat": "Requested: {{ container }}",
- "metric": "kube_pod_container_resource_requests_cpu_cores",
- "refId": "B",
- "step": 20
- },
- {
- "expr": "kube_pod_container_resource_limits_cpu_cores{pod=\"$pod\", container=~\"$container\"}",
- "interval": "10s",
- "intervalFactor": 2,
- "legendFormat": "Limit: {{ container }}",
- "metric": "kube_pod_container_resource_limits_memory_bytes",
- "refId": "C",
- "step": 20
- }
- ],
- "title": "CPU Usage",
- "tooltip": {
- "msResolution": true,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "Row",
- "titleSize": "h6"
+ "allValue": null,
+ "current": {
+
+ },
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": true,
+ "label": "Container",
+ "multi": false,
+ "name": "container",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_container_info{namespace=\"$namespace\", pod=\"$pod\"}, container)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
},
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 3,
- "isNew": false,
- "legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": true
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 12,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sort_desc(sum by (pod_name) (rate(container_network_receive_bytes_total{pod_name=\"$pod\"}[1m])))",
- "intervalFactor": 2,
- "legendFormat": "{{ pod_name }}",
- "refId": "A",
- "step": 30
- }
- ],
- "title": "Network I/O",
- "tooltip": {
- "msResolution": true,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": [
- {
- "allValue": ".*",
- "current": {},
- "datasource": "prometheus",
- "hide": 0,
- "includeAll": true,
- "label": "Namespace",
- "multi": false,
- "name": "namespace",
- "options": [],
- "query": "label_values(kube_pod_info, namespace)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {},
- "datasource": "prometheus",
- "hide": 0,
- "includeAll": false,
- "label": "Pod",
- "multi": false,
- "name": "pod",
- "options": [],
- "query": "label_values(kube_pod_info{namespace=~\"$namespace\"}, pod)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": ".*",
- "current": {},
- "datasource": "prometheus",
- "hide": 0,
- "includeAll": true,
- "label": "Container",
- "multi": false,
- "name": "container",
- "options": [],
- "query": "label_values(kube_pod_container_info{namespace=\"$namespace\", pod=\"$pod\"}, container)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-6h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Pods",
- "version": 1
- }
- statefulset-dashboard.json: |+
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "prometheus",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 1,
- "hideControls": false,
- "links": [],
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "200px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 8,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "cores",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 4,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "sum(rate(container_cpu_usage_seconds_total{namespace=\"$statefulset_namespace\",pod_name=~\"$statefulset_name.*\"}[3m]))",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "CPU",
- "type": "singlestat",
- "valueFontSize": "110%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 9,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "GB",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "80%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 4,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "sum(container_memory_usage_bytes{namespace=\"$statefulset_namespace\",pod_name=~\"$statefulset_name.*\"}) / 1024^3",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Memory",
- "type": "singlestat",
- "valueFontSize": "110%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "Bps",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": false
- },
- "id": 7,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 4,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "sum(rate(container_network_transmit_bytes_total{namespace=\"$statefulset_namespace\",pod_name=~\"$statefulset_name.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{namespace=\"$statefulset_namespace\",pod_name=~\"$statefulset_name.*\"}[3m]))",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Network",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
+ "time": {
+ "from": "now-1h",
+ "to": "now"
},
- {
- "collapse": false,
- "editable": false,
- "height": "100px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": false
- },
- "id": 5,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(kube_statefulset_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "metric": "kube_statefulset_replicas",
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Desired Replicas",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 6,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "min(kube_statefulset_status_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Available Replicas",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 3,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(kube_statefulset_status_observed_generation{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Observed Generation",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "prometheus",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 2,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(kube_statefulset_metadata_generation{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Metadata Generation",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
},
- {
- "collapse": false,
- "editable": false,
- "height": "350px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "prometheus",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 1,
- "isNew": true,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 12,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "min(kube_statefulset_status_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "available",
- "refId": "B",
- "step": 30
- },
- {
- "expr": "max(kube_statefulset_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "desired",
- "refId": "E",
- "step": 30
- }
- ],
- "title": "Replicas",
- "tooltip": {
- "msResolution": true,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "none",
- "label": "",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "label": "",
- "logBase": 1,
- "show": false
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": [
- {
- "allValue": ".*",
- "current": {},
- "datasource": "prometheus",
- "hide": 0,
- "includeAll": false,
- "label": "Namespace",
- "multi": false,
- "name": "statefulset_namespace",
- "options": [],
- "query": "label_values(kube_statefulset_metadata_generation, namespace)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": null,
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {},
- "datasource": "prometheus",
- "hide": 0,
- "includeAll": false,
- "label": "StatefulSet",
- "multi": false,
- "name": "statefulset_name",
- "options": [],
- "query": "label_values(kube_statefulset_metadata_generation{namespace=\"$statefulset_namespace\"}, statefulset)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "statefulset",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-6h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "StatefulSet",
- "version": 1
+ "timezone": "browser",
+ "title": "Pods",
+ "version": 0
}
----
+kind: ConfigMap
+metadata:
+ name: grafana-dashboard-definitions
+ namespace: monitoring
diff --git a/manifests/grafana/grafana-dashboard-sources.yaml b/manifests/grafana/grafana-dashboard-sources.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..61fdcf611110388a60708078701e6cfddfb54ebd
--- /dev/null
+++ b/manifests/grafana/grafana-dashboard-sources.yaml
@@ -0,0 +1,18 @@
+apiVersion: v1
+data:
+ dashboards.yaml: |-
+ [
+ {
+ "folder": "",
+ "name": "0",
+ "options": {
+ "path": "/grafana-dashboard-definitions/0"
+ },
+ "org_id": 1,
+ "type": "file"
+ }
+ ]
+kind: ConfigMap
+metadata:
+ name: grafana-dashboards
+ namespace: monitoring
diff --git a/manifests/grafana/grafana-dashboards.yaml b/manifests/grafana/grafana-dashboards.yaml
deleted file mode 100644
index 772d3f64902f652365e4b7b2714f483cdde42b6e..0000000000000000000000000000000000000000
--- a/manifests/grafana/grafana-dashboards.yaml
+++ /dev/null
@@ -1,12 +0,0 @@
-apiVersion: v1
-kind: ConfigMap
-metadata:
- name: grafana-dashboards
-data:
- dashboards.yaml: |+
- - name: '0'
- org_id: 1
- folder: ''
- type: file
- options:
- folder: /grafana-dashboard-definitions/0
diff --git a/manifests/grafana/grafana-datasources.yaml b/manifests/grafana/grafana-datasources.yaml
index 33c3081f7497e91daba85b737908c8e30314adec..5ed25a025c4f3247985dac8b4736321e5ddebcb8 100644
--- a/manifests/grafana/grafana-datasources.yaml
+++ b/manifests/grafana/grafana-datasources.yaml
@@ -1,15 +1,20 @@
apiVersion: v1
+data:
+ prometheus.yaml: |-
+ {
+ "datasources": [
+ {
+ "access": "proxy",
+ "etitable": false,
+ "name": "prometheus",
+ "org_id": 1,
+ "type": "prometheus",
+ "url": "http://prometheus-k8s.monitoring.svc:9090",
+ "version": 1
+ }
+ ]
+ }
kind: ConfigMap
metadata:
name: grafana-datasources
-data:
- prometheus.yaml: |+
- datasources:
- - name: prometheus
- type: prometheus
- access: proxy
- org_id: 1
- url: http://prometheus-k8s.monitoring.svc:9090
- version: 1
- editable: false
-
+ namespace: monitoring
diff --git a/manifests/grafana/grafana-deployment.yaml b/manifests/grafana/grafana-deployment.yaml
index 9eb8750fa024ac446b633066bbf8c2643c7756d9..9d7ae88f09fd855cc40f13fb031539adaa54df40 100644
--- a/manifests/grafana/grafana-deployment.yaml
+++ b/manifests/grafana/grafana-deployment.yaml
@@ -1,48 +1,59 @@
-apiVersion: apps/v1beta1
+apiVersion: apps/v1beta2
kind: Deployment
metadata:
+ labels:
+ app: grafana
name: grafana
+ namespace: monitoring
spec:
replicas: 1
+ selector:
+ matchLabels:
+ app: grafana
template:
metadata:
labels:
app: grafana
spec:
- securityContext:
- runAsNonRoot: true
- runAsUser: 65534
containers:
- - name: grafana
- image: quay.io/coreos/monitoring-grafana:5.0.3
- volumeMounts:
- - name: grafana-storage
- mountPath: /data
- - name: grafana-datasources
- mountPath: /grafana/conf/provisioning/datasources
- - name: grafana-dashboards
- mountPath: /grafana/conf/provisioning/dashboards
- - name: grafana-dashboard-definitions-0
- mountPath: /grafana-dashboard-definitions/0
+ - image: quay.io/coreos/monitoring-grafana:5.0.3
+ name: grafana
ports:
- - name: web
- containerPort: 3000
+ - containerPort: 3000
+ name: http
resources:
- requests:
- memory: 100Mi
- cpu: 100m
limits:
- memory: 200Mi
cpu: 200m
+ memory: 200Mi
+ requests:
+ cpu: 100m
+ memory: 100Mi
+ volumeMounts:
+ - mountPath: /data
+ name: grafana-storage
+ readOnly: false
+ - mountPath: /grafana/conf/provisioning/datasources
+ name: grafana-datasources
+ readOnly: false
+ - mountPath: /grafana/conf/provisioning/dashboards
+ name: grafana-dashboards
+ readOnly: false
+ - mountPath: /grafana-dashboard-definitions/0
+ name: grafana-dashboard-definitions
+ readOnly: false
+ securityContext:
+ runAsNonRoot: true
+ runAsUser: 65534
+ serviceAccountName: grafana
volumes:
- - name: grafana-storage
- emptyDir: {}
- - name: grafana-datasources
- configMap:
+ - emptyDir: {}
+ name: grafana-storage
+ - configMap:
name: grafana-datasources
- - name: grafana-dashboards
- configMap:
+ name: grafana-datasources
+ - configMap:
name: grafana-dashboards
- - name: grafana-dashboard-definitions-0
- configMap:
- name: grafana-dashboard-definitions-0
+ name: grafana-dashboards
+ - configMap:
+ name: grafana-dashboard-definitions
+ name: grafana-dashboard-definitions
diff --git a/manifests/grafana/grafana-service-account.yaml b/manifests/grafana/grafana-service-account.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..3ed3e031e7445fc347e27f8f7134d70c384f0be3
--- /dev/null
+++ b/manifests/grafana/grafana-service-account.yaml
@@ -0,0 +1,5 @@
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+ name: grafana
+ namespace: monitoring
diff --git a/manifests/grafana/grafana-service.yaml b/manifests/grafana/grafana-service.yaml
index fbcee40dcb2b0ca0020c410c3bfdabd4ec21911a..45f77a0dca1d20f1051fc4c19053b7afc9fb5b19 100644
--- a/manifests/grafana/grafana-service.yaml
+++ b/manifests/grafana/grafana-service.yaml
@@ -2,14 +2,11 @@ apiVersion: v1
kind: Service
metadata:
name: grafana
- labels:
- app: grafana
+ namespace: monitoring
spec:
- type: NodePort
ports:
- - port: 3000
- protocol: TCP
- nodePort: 30902
- targetPort: web
+ - name: http
+ port: 3000
+ targetPort: http
selector:
app: grafana
diff --git a/manifests/k8s/kubeadm/kube-controller-manager.yaml b/manifests/k8s/kubeadm/kube-controller-manager.yaml
deleted file mode 100644
index bd8d7cb5a975a8d709f068f551a65360b97b0631..0000000000000000000000000000000000000000
--- a/manifests/k8s/kubeadm/kube-controller-manager.yaml
+++ /dev/null
@@ -1,17 +0,0 @@
-apiVersion: v1
-kind: Service
-metadata:
- namespace: kube-system
- name: kube-controller-manager-prometheus-discovery
- labels:
- k8s-app: kube-controller-manager
-spec:
- selector:
- component: kube-controller-manager
- type: ClusterIP
- clusterIP: None
- ports:
- - name: http-metrics
- port: 10252
- targetPort: 10252
- protocol: TCP
diff --git a/manifests/k8s/kubeadm/kube-scheduler.yaml b/manifests/k8s/kubeadm/kube-scheduler.yaml
deleted file mode 100644
index 2d90097a9ba63a45ac36a2bdf8d38bbb748e87ba..0000000000000000000000000000000000000000
--- a/manifests/k8s/kubeadm/kube-scheduler.yaml
+++ /dev/null
@@ -1,17 +0,0 @@
-apiVersion: v1
-kind: Service
-metadata:
- namespace: kube-system
- name: kube-scheduler-prometheus-discovery
- labels:
- k8s-app: kube-scheduler
-spec:
- selector:
- component: kube-scheduler
- type: ClusterIP
- clusterIP: None
- ports:
- - name: http-metrics
- port: 10251
- targetPort: 10251
- protocol: TCP
diff --git a/manifests/k8s/self-hosted/kube-controller-manager.yaml b/manifests/k8s/self-hosted/kube-controller-manager.yaml
deleted file mode 100644
index a2983101ab87a56a1aad5a3e72304b8ae544b23f..0000000000000000000000000000000000000000
--- a/manifests/k8s/self-hosted/kube-controller-manager.yaml
+++ /dev/null
@@ -1,17 +0,0 @@
-apiVersion: v1
-kind: Service
-metadata:
- namespace: kube-system
- name: kube-controller-manager-prometheus-discovery
- labels:
- k8s-app: kube-controller-manager
-spec:
- selector:
- k8s-app: kube-controller-manager
- type: ClusterIP
- clusterIP: None
- ports:
- - name: http-metrics
- port: 10252
- targetPort: 10252
- protocol: TCP
diff --git a/manifests/k8s/self-hosted/kube-dns.yaml b/manifests/k8s/self-hosted/kube-dns.yaml
deleted file mode 100644
index e032771489b144e695aefe9a4b1b26dd39bdc91d..0000000000000000000000000000000000000000
--- a/manifests/k8s/self-hosted/kube-dns.yaml
+++ /dev/null
@@ -1,21 +0,0 @@
-apiVersion: v1
-kind: Service
-metadata:
- namespace: kube-system
- name: kube-dns-prometheus-discovery
- labels:
- k8s-app: kube-dns
-spec:
- selector:
- k8s-app: kube-dns
- type: ClusterIP
- clusterIP: None
- ports:
- - name: http-metrics-skydns
- port: 10055
- targetPort: 10055
- protocol: TCP
- - name: http-metrics-dnsmasq
- port: 10054
- targetPort: 10054
- protocol: TCP
diff --git a/manifests/k8s/self-hosted/kube-scheduler.yaml b/manifests/k8s/self-hosted/kube-scheduler.yaml
deleted file mode 100644
index 0fe05dd74a339f4d152e3b15cf7035322f42e975..0000000000000000000000000000000000000000
--- a/manifests/k8s/self-hosted/kube-scheduler.yaml
+++ /dev/null
@@ -1,17 +0,0 @@
-apiVersion: v1
-kind: Service
-metadata:
- namespace: kube-system
- name: kube-scheduler-prometheus-discovery
- labels:
- k8s-app: kube-scheduler
-spec:
- selector:
- k8s-app: kube-scheduler
- type: ClusterIP
- clusterIP: None
- ports:
- - name: http-metrics
- port: 10251
- targetPort: 10251
- protocol: TCP
diff --git a/manifests/kube-state-metrics/kube-state-metrics-cluster-role-binding.yaml b/manifests/kube-state-metrics/kube-state-metrics-cluster-role-binding.yaml
index 8284fc151317205c95baf4c62add202ebd8e22ea..9a8f3111abb8f0f418960ff0cdd56aaf95037076 100644
--- a/manifests/kube-state-metrics/kube-state-metrics-cluster-role-binding.yaml
+++ b/manifests/kube-state-metrics/kube-state-metrics-cluster-role-binding.yaml
@@ -1,4 +1,4 @@
-apiVersion: rbac.authorization.k8s.io/v1beta1
+apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
name: kube-state-metrics
diff --git a/manifests/kube-state-metrics/kube-state-metrics-cluster-role.yaml b/manifests/kube-state-metrics/kube-state-metrics-cluster-role.yaml
index ef5e91ac61b225f148413fca0151bdc51e123286..cae184834cfff42306b4d56fff6cff4004267ba8 100644
--- a/manifests/kube-state-metrics/kube-state-metrics-cluster-role.yaml
+++ b/manifests/kube-state-metrics/kube-state-metrics-cluster-role.yaml
@@ -1,10 +1,13 @@
-apiVersion: rbac.authorization.k8s.io/v1beta1
+apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
name: kube-state-metrics
rules:
-- apiGroups: [""]
+- apiGroups:
+ - ""
resources:
+ - configmaps
+ - secrets
- nodes
- pods
- services
@@ -15,31 +18,49 @@ rules:
- persistentvolumes
- namespaces
- endpoints
- verbs: ["list", "watch"]
-- apiGroups: ["extensions"]
+ verbs:
+ - list
+ - watch
+- apiGroups:
+ - extensions
resources:
- daemonsets
- deployments
- replicasets
- verbs: ["list", "watch"]
-- apiGroups: ["apps"]
+ verbs:
+ - list
+ - watch
+- apiGroups:
+ - apps
resources:
- statefulsets
- verbs: ["list", "watch"]
-- apiGroups: ["batch"]
+ verbs:
+ - list
+ - watch
+- apiGroups:
+ - batch
resources:
- cronjobs
- jobs
- verbs: ["list", "watch"]
-- apiGroups: ["autoscaling"]
+ verbs:
+ - list
+ - watch
+- apiGroups:
+ - autoscaling
resources:
- horizontalpodautoscalers
- verbs: ["list", "watch"]
-- apiGroups: ["authentication.k8s.io"]
+ verbs:
+ - list
+ - watch
+- apiGroups:
+ - authentication.k8s.io
resources:
- tokenreviews
- verbs: ["create"]
-- apiGroups: ["authorization.k8s.io"]
+ verbs:
+ - create
+- apiGroups:
+ - authorization.k8s.io
resources:
- subjectaccessreviews
- verbs: ["create"]
\ No newline at end of file
+ verbs:
+ - create
diff --git a/manifests/kube-state-metrics/kube-state-metrics-deployment.yaml b/manifests/kube-state-metrics/kube-state-metrics-deployment.yaml
index 61f918eb57d638de82466d7fbb9b64fdc1adb85d..bd6d9475db3006b31c5cbab509f49469ced3d3fc 100644
--- a/manifests/kube-state-metrics/kube-state-metrics-deployment.yaml
+++ b/manifests/kube-state-metrics/kube-state-metrics-deployment.yaml
@@ -1,80 +1,95 @@
-apiVersion: extensions/v1beta1
+apiVersion: apps/v1beta2
kind: Deployment
metadata:
+ labels:
+ app: kube-state-metrics
name: kube-state-metrics
+ namespace: monitoring
spec:
replicas: 1
+ selector:
+ matchLabels:
+ app: kube-state-metrics
template:
metadata:
labels:
app: kube-state-metrics
spec:
- serviceAccountName: kube-state-metrics
- securityContext:
- runAsNonRoot: true
- runAsUser: 65534
containers:
- - name: kube-rbac-proxy-main
- image: quay.io/brancz/kube-rbac-proxy:v0.2.0
- args:
- - "--secure-listen-address=:8443"
- - "--upstream=http://127.0.0.1:8081/"
+ - args:
+ - --secure-listen-address=:8443
+ - --upstream=http://127.0.0.1:8081/
+ image: quay.io/coreos/kube-rbac-proxy:v0.3.0
+ name: kube-rbac-proxy-main
ports:
- - name: https-main
- containerPort: 8443
+ - containerPort: 8443
+ name: https-main
resources:
- requests:
- memory: 20Mi
- cpu: 10m
limits:
- memory: 40Mi
cpu: 20m
- - name: kube-rbac-proxy-self
- image: quay.io/brancz/kube-rbac-proxy:v0.2.0
- args:
- - "--secure-listen-address=:9443"
- - "--upstream=http://127.0.0.1:8082/"
+ memory: 40Mi
+ requests:
+ cpu: 10m
+ memory: 20Mi
+ - args:
+ - --secure-listen-address=:9443
+ - --upstream=http://127.0.0.1:8082/
+ image: quay.io/coreos/kube-rbac-proxy:v0.3.0
+ name: kube-rbac-proxy-self
ports:
- - name: https-self
- containerPort: 9443
+ - containerPort: 9443
+ name: https-self
resources:
+ limits:
+ cpu: 20m
+ memory: 40Mi
requests:
- memory: 20Mi
cpu: 10m
+ memory: 20Mi
+ - args:
+ - --host=127.0.0.1
+ - --port=8081
+ - --telemetry-host=127.0.0.1
+ - --telemetry-port=8082
+ image: quay.io/coreos/kube-state-metrics:v1.3.0
+ name: kube-state-metrics
+ resources:
limits:
- memory: 40Mi
- cpu: 20m
- - name: kube-state-metrics
- image: quay.io/coreos/kube-state-metrics:v1.2.0
- args:
- - "--host=127.0.0.1"
- - "--port=8081"
- - "--telemetry-host=127.0.0.1"
- - "--telemetry-port=8082"
- - name: addon-resizer
- image: gcr.io/google_containers/addon-resizer:1.0
+ cpu: 102m
+ memory: 180Mi
+ requests:
+ cpu: 102m
+ memory: 180Mi
+ - command:
+ - /pod_nanny
+ - --container=kube-state-metrics
+ - --cpu=100m
+ - --extra-cpu=2m
+ - --memory=150Mi
+ - --extra-memory=30Mi
+ - --threshold=5
+ - --deployment=kube-state-metrics
+ env:
+ - name: MY_POD_NAME
+ valueFrom:
+ fieldRef:
+ apiVersion: v1
+ fieldPath: metadata.name
+ - name: MY_POD_NAMESPACE
+ valueFrom:
+ fieldRef:
+ apiVersion: v1
+ fieldPath: metadata.namespace
+ image: quay.io/coreos/addon-resizer:1.0
+ name: addon-resizer
resources:
limits:
- cpu: 100m
+ cpu: 10m
memory: 30Mi
requests:
- cpu: 100m
+ cpu: 10m
memory: 30Mi
- env:
- - name: MY_POD_NAME
- valueFrom:
- fieldRef:
- fieldPath: metadata.name
- - name: MY_POD_NAMESPACE
- valueFrom:
- fieldRef:
- fieldPath: metadata.namespace
- command:
- - /pod_nanny
- - --container=kube-state-metrics
- - --cpu=100m
- - --extra-cpu=2m
- - --memory=150Mi
- - --extra-memory=30Mi
- - --threshold=5
- - --deployment=kube-state-metrics
+ securityContext:
+ runAsNonRoot: true
+ runAsUser: 65534
+ serviceAccountName: kube-state-metrics
diff --git a/manifests/kube-state-metrics/kube-state-metrics-role-binding.yaml b/manifests/kube-state-metrics/kube-state-metrics-role-binding.yaml
index a93c396505f49e6b9b1ca545d9342a767b294b79..dcad8055aba90e8ad0f12ecf2ac6d8de7e39ff01 100644
--- a/manifests/kube-state-metrics/kube-state-metrics-role-binding.yaml
+++ b/manifests/kube-state-metrics/kube-state-metrics-role-binding.yaml
@@ -1,12 +1,12 @@
-apiVersion: rbac.authorization.k8s.io/v1beta1
+apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
name: kube-state-metrics
+ namespace: monitoring
roleRef:
apiGroup: rbac.authorization.k8s.io
kind: Role
- name: kube-state-metrics-resizer
+ name: kube-state-metrics-addon-resizer
subjects:
- kind: ServiceAccount
name: kube-state-metrics
-
diff --git a/manifests/kube-state-metrics/kube-state-metrics-role.yaml b/manifests/kube-state-metrics/kube-state-metrics-role.yaml
index 6bf21fb882e15683cc57654f2efc33d82f8d3eff..0063ffb453a8d1f254755e7209a73f74ec1a5195 100644
--- a/manifests/kube-state-metrics/kube-state-metrics-role.yaml
+++ b/manifests/kube-state-metrics/kube-state-metrics-role.yaml
@@ -1,15 +1,21 @@
-apiVersion: rbac.authorization.k8s.io/v1beta1
+apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
- name: kube-state-metrics-resizer
+ name: kube-state-metrics
+ namespace: monitoring
rules:
-- apiGroups: [""]
+- apiGroups:
+ - ""
resources:
- pods
- verbs: ["get"]
-- apiGroups: ["extensions"]
+ verbs:
+ - get
+- apiGroups:
+ - extensions
+ resourceNames:
+ - kube-state-metrics
resources:
- deployments
- resourceNames: ["kube-state-metrics"]
- verbs: ["get", "update"]
-
+ verbs:
+ - get
+ - update
diff --git a/manifests/kube-state-metrics/kube-state-metrics-service-account.yaml b/manifests/kube-state-metrics/kube-state-metrics-service-account.yaml
index 997793528ad8b9970a4b5100cc644ea1ceafcb41..fff1028b442c69109cb8fa8e5f808f2a856838f8 100644
--- a/manifests/kube-state-metrics/kube-state-metrics-service-account.yaml
+++ b/manifests/kube-state-metrics/kube-state-metrics-service-account.yaml
@@ -2,3 +2,4 @@ apiVersion: v1
kind: ServiceAccount
metadata:
name: kube-state-metrics
+ namespace: monitoring
diff --git a/manifests/kube-state-metrics/kube-state-metrics-service.yaml b/manifests/kube-state-metrics/kube-state-metrics-service.yaml
index b4422685c270f2ad45556ba1e8c3b2a09fe4431c..3e88b562b615e5ff0a3f9441f377c418c0449b9a 100644
--- a/manifests/kube-state-metrics/kube-state-metrics-service.yaml
+++ b/manifests/kube-state-metrics/kube-state-metrics-service.yaml
@@ -2,20 +2,16 @@ apiVersion: v1
kind: Service
metadata:
labels:
- app: kube-state-metrics
k8s-app: kube-state-metrics
name: kube-state-metrics
+ namespace: monitoring
spec:
- clusterIP: None
ports:
- name: https-main
port: 8443
targetPort: https-main
- protocol: TCP
- name: https-self
port: 9443
targetPort: https-self
- protocol: TCP
selector:
app: kube-state-metrics
-
diff --git a/manifests/node-exporter/node-exporter-cluster-role.yaml b/manifests/node-exporter/node-exporter-cluster-role.yaml
index 932b7762c43051937e9ead2a721e9b2414eecdf9..ad783ae9bfdb3eb2c57e8d7bc9be580e75880bbf 100644
--- a/manifests/node-exporter/node-exporter-cluster-role.yaml
+++ b/manifests/node-exporter/node-exporter-cluster-role.yaml
@@ -3,11 +3,15 @@ kind: ClusterRole
metadata:
name: node-exporter
rules:
-- apiGroups: ["authentication.k8s.io"]
+- apiGroups:
+ - authentication.k8s.io
resources:
- tokenreviews
- verbs: ["create"]
-- apiGroups: ["authorization.k8s.io"]
+ verbs:
+ - create
+- apiGroups:
+ - authorization.k8s.io
resources:
- subjectaccessreviews
- verbs: ["create"]
+ verbs:
+ - create
diff --git a/manifests/node-exporter/node-exporter-daemonset.yaml b/manifests/node-exporter/node-exporter-daemonset.yaml
index f92113e87ccb0afd13b9a65ad2166a05b2911384..1284e93dd260a3ad37adcfd65ed771dcc67752de 100644
--- a/manifests/node-exporter/node-exporter-daemonset.yaml
+++ b/manifests/node-exporter/node-exporter-daemonset.yaml
@@ -1,69 +1,63 @@
-apiVersion: extensions/v1beta1
+apiVersion: apps/v1beta2
kind: DaemonSet
metadata:
+ labels:
+ app: node-exporter
name: node-exporter
+ namespace: monitoring
spec:
- updateStrategy:
- rollingUpdate:
- maxUnavailable: 1
- type: RollingUpdate
+ selector:
+ matchLabels:
+ app: node-exporter
template:
metadata:
labels:
app: node-exporter
- name: node-exporter
spec:
- serviceAccountName: node-exporter
- securityContext:
- runAsNonRoot: true
- runAsUser: 65534
- hostNetwork: true
- hostPID: true
containers:
- - image: quay.io/prometheus/node-exporter:v0.15.2
- args:
- - "--web.listen-address=127.0.0.1:9101"
- - "--path.procfs=/host/proc"
- - "--path.sysfs=/host/sys"
+ - args:
+ - --web.listen-address=127.0.0.1:9101
+ - --path.procfs=/host/proc
+ - --path.sysfs=/host/sys
+ image: quay.io/prometheus/node-exporter:v0.15.2
name: node-exporter
resources:
- requests:
- memory: 30Mi
- cpu: 100m
limits:
- memory: 50Mi
- cpu: 200m
+ cpu: 102m
+ memory: 180Mi
+ requests:
+ cpu: 102m
+ memory: 180Mi
volumeMounts:
- - name: proc
- readOnly: true
- mountPath: /host/proc
- - name: sys
- readOnly: true
- mountPath: /host/sys
- - name: kube-rbac-proxy
- image: quay.io/brancz/kube-rbac-proxy:v0.2.0
- args:
- - "--secure-listen-address=:9100"
- - "--upstream=http://127.0.0.1:9101/"
+ - mountPath: /host/proc
+ name: proc
+ readOnly: false
+ - mountPath: /host/sys
+ name: sys
+ readOnly: false
+ - args:
+ - --secure-listen-address=:9100
+ - --upstream=http://127.0.0.1:9101/
+ image: quay.io/coreos/kube-rbac-proxy:v0.3.0
+ name: kube-rbac-proxy
ports:
- containerPort: 9100
- hostPort: 9100
name: https
resources:
- requests:
- memory: 20Mi
- cpu: 10m
limits:
- memory: 40Mi
cpu: 20m
- tolerations:
- - effect: NoSchedule
- operator: Exists
+ memory: 40Mi
+ requests:
+ cpu: 10m
+ memory: 20Mi
+ securityContext:
+ runAsNonRoot: true
+ runAsUser: 65534
+ serviceAccountName: node-exporter
volumes:
- - name: proc
- hostPath:
+ - hostPath:
path: /proc
- - name: sys
- hostPath:
+ name: proc
+ - hostPath:
path: /sys
-
+ name: sys
diff --git a/manifests/node-exporter/node-exporter-service-account.yaml b/manifests/node-exporter/node-exporter-service-account.yaml
index 703a274882355461607aeea10d00f2127186810a..8a03ac1641f68cbca54468aae86c79a20156058a 100644
--- a/manifests/node-exporter/node-exporter-service-account.yaml
+++ b/manifests/node-exporter/node-exporter-service-account.yaml
@@ -2,3 +2,4 @@ apiVersion: v1
kind: ServiceAccount
metadata:
name: node-exporter
+ namespace: monitoring
diff --git a/manifests/node-exporter/node-exporter-service.yaml b/manifests/node-exporter/node-exporter-service.yaml
index 8aa3774792925629be50c08eed611247de866fd7..101a9769daff5d003f9970c58c04233cc4914095 100644
--- a/manifests/node-exporter/node-exporter-service.yaml
+++ b/manifests/node-exporter/node-exporter-service.yaml
@@ -2,16 +2,13 @@ apiVersion: v1
kind: Service
metadata:
labels:
- app: node-exporter
k8s-app: node-exporter
name: node-exporter
+ namespace: monitoring
spec:
- type: ClusterIP
- clusterIP: None
ports:
- name: https
port: 9100
- protocol: TCP
+ targetPort: https
selector:
app: node-exporter
-
diff --git a/manifests/prometheus-k8s/prometheus-k8s-cluster-role-binding.yaml b/manifests/prometheus-k8s/prometheus-k8s-cluster-role-binding.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..554bb6f8767c0a8a346614c685649321a176b640
--- /dev/null
+++ b/manifests/prometheus-k8s/prometheus-k8s-cluster-role-binding.yaml
@@ -0,0 +1,12 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+ name: prometheus-k8s
+roleRef:
+ apiGroup: rbac.authorization.k8s.io
+ kind: ClusterRole
+ name: prometheus-k8s
+subjects:
+- kind: ServiceAccount
+ name: prometheus-k8s
+ namespace: monitoring
diff --git a/manifests/prometheus-k8s/prometheus-k8s-cluster-role.yaml b/manifests/prometheus-k8s/prometheus-k8s-cluster-role.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..d5c4598304eaf7fe0c183e34969ca734f793aa75
--- /dev/null
+++ b/manifests/prometheus-k8s/prometheus-k8s-cluster-role.yaml
@@ -0,0 +1,15 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+ name: prometheus-k8s
+rules:
+- apiGroups:
+ - ""
+ resources:
+ - nodes/metrics
+ verbs:
+ - get
+- nonResourceURLs:
+ - /metrics
+ verbs:
+ - get
diff --git a/manifests/prometheus-k8s/prometheus-k8s-role-binding-config.yaml b/manifests/prometheus-k8s/prometheus-k8s-role-binding-config.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..f21ff3c5978a18636dd43473d13eb3e62a697f8b
--- /dev/null
+++ b/manifests/prometheus-k8s/prometheus-k8s-role-binding-config.yaml
@@ -0,0 +1,13 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+ name: prometheus-k8s-config
+ namespace: monitoring
+roleRef:
+ apiGroup: rbac.authorization.k8s.io
+ kind: Role
+ name: prometheus-k8s-config
+subjects:
+- kind: ServiceAccount
+ name: prometheus-k8s-config
+ namespace: monitoring
diff --git a/manifests/prometheus-k8s/prometheus-k8s-role-binding-default.yaml b/manifests/prometheus-k8s/prometheus-k8s-role-binding-default.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..c40397108012dde1390d3e5e5301c77f2d41b1fb
--- /dev/null
+++ b/manifests/prometheus-k8s/prometheus-k8s-role-binding-default.yaml
@@ -0,0 +1,13 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+ name: prometheus-k8s
+ namespace: default
+roleRef:
+ apiGroup: rbac.authorization.k8s.io
+ kind: Role
+ name: prometheus-k8s
+subjects:
+- kind: ServiceAccount
+ name: prometheus-k8s
+ namespace: monitoring
diff --git a/manifests/prometheus-k8s/prometheus-k8s-role-binding-kube-system.yaml b/manifests/prometheus-k8s/prometheus-k8s-role-binding-kube-system.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..250c7307540671cb11c18185cdb7a80d38c42647
--- /dev/null
+++ b/manifests/prometheus-k8s/prometheus-k8s-role-binding-kube-system.yaml
@@ -0,0 +1,13 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+ name: prometheus-k8s
+ namespace: kube-system
+roleRef:
+ apiGroup: rbac.authorization.k8s.io
+ kind: Role
+ name: prometheus-k8s
+subjects:
+- kind: ServiceAccount
+ name: prometheus-k8s
+ namespace: monitoring
diff --git a/manifests/prometheus-k8s/prometheus-k8s-role-binding-namespace.yaml b/manifests/prometheus-k8s/prometheus-k8s-role-binding-namespace.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..068c77d37d9b1c54b6216994b01efd260b64a35a
--- /dev/null
+++ b/manifests/prometheus-k8s/prometheus-k8s-role-binding-namespace.yaml
@@ -0,0 +1,13 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+ name: prometheus-k8s
+ namespace: monitoring
+roleRef:
+ apiGroup: rbac.authorization.k8s.io
+ kind: Role
+ name: prometheus-k8s
+subjects:
+- kind: ServiceAccount
+ name: prometheus-k8s
+ namespace: monitoring
diff --git a/manifests/prometheus-k8s/prometheus-k8s-role-config.yaml b/manifests/prometheus-k8s/prometheus-k8s-role-config.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..5f1cd043919a18141c2331c05836d5ef45d099eb
--- /dev/null
+++ b/manifests/prometheus-k8s/prometheus-k8s-role-config.yaml
@@ -0,0 +1,12 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+ name: prometheus-k8s-config
+ namespace: monitoring
+rules:
+- apiGroups:
+ - ""
+ resources:
+ - configmaps
+ verbs:
+ - get
diff --git a/manifests/prometheus-k8s/prometheus-k8s-role-default.yaml b/manifests/prometheus-k8s/prometheus-k8s-role-default.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..1c336117a26c5efa32dca57a1b73c2a8b4245162
--- /dev/null
+++ b/manifests/prometheus-k8s/prometheus-k8s-role-default.yaml
@@ -0,0 +1,17 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+ name: prometheus-k8s
+ namespace: default
+rules:
+- apiGroups:
+ - ""
+ resources:
+ - nodes
+ - services
+ - endpoints
+ - pods
+ verbs:
+ - get
+ - list
+ - watch
diff --git a/manifests/prometheus-k8s/prometheus-k8s-role-kube-system.yaml b/manifests/prometheus-k8s/prometheus-k8s-role-kube-system.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..d82fe3ab8b2d12352dcf390a4dc8cf4d47277b1c
--- /dev/null
+++ b/manifests/prometheus-k8s/prometheus-k8s-role-kube-system.yaml
@@ -0,0 +1,17 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+ name: prometheus-k8s
+ namespace: kube-system
+rules:
+- apiGroups:
+ - ""
+ resources:
+ - nodes
+ - services
+ - endpoints
+ - pods
+ verbs:
+ - get
+ - list
+ - watch
diff --git a/manifests/prometheus-k8s/prometheus-k8s-role-namespace.yaml b/manifests/prometheus-k8s/prometheus-k8s-role-namespace.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..343cfc6dff7e41ff59ced5a55f093da3032aecd7
--- /dev/null
+++ b/manifests/prometheus-k8s/prometheus-k8s-role-namespace.yaml
@@ -0,0 +1,17 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+ name: prometheus-k8s
+ namespace: monitoring
+rules:
+- apiGroups:
+ - ""
+ resources:
+ - nodes
+ - services
+ - endpoints
+ - pods
+ verbs:
+ - get
+ - list
+ - watch
diff --git a/manifests/prometheus/prometheus-k8s-rules.yaml b/manifests/prometheus-k8s/prometheus-k8s-rules.yaml
similarity index 83%
rename from manifests/prometheus/prometheus-k8s-rules.yaml
rename to manifests/prometheus-k8s/prometheus-k8s-rules.yaml
index c45bbbbc88bd8d3d204120cca5dca48a3ccdff3d..90cb3f3e165c6a07d0206fa98682b00814bc02ae 100644
--- a/manifests/prometheus/prometheus-k8s-rules.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-rules.yaml
@@ -1,12 +1,6 @@
apiVersion: v1
-kind: ConfigMap
-metadata:
- name: prometheus-k8s-rules
- labels:
- role: alert-rules
- prometheus: k8s
data:
- alertmanager.rules.yaml: |+
+ alertmanager.rules.yaml: |
groups:
- name: alertmanager.rules
rules:
@@ -40,7 +34,7 @@ data:
description: Reloading Alertmanager's configuration has failed for {{ $labels.namespace
}}/{{ $labels.pod}}.
summary: Alertmanager's configuration reload failed
- etcd3.rules.yaml: |+
+ etcd3.rules.yaml: |
groups:
- name: ./etcd3.rules
rules:
@@ -164,7 +158,7 @@ data:
annotations:
description: etcd instance {{ $labels.instance }} commit durations are high
summary: high commit durations
- general.rules.yaml: |+
+ general.rules.yaml: |
groups:
- name: general.rules
rules:
@@ -204,7 +198,7 @@ data:
description: '{{ $labels.job }}: {{ $labels.namespace }}/{{ $labels.pod }} instance
will exhaust in file/socket descriptors within the next hour'
summary: file descriptors soon exhausted
- kube-controller-manager.rules.yaml: |+
+ kube-controller-manager.rules.yaml: |
groups:
- name: kube-controller-manager.rules
rules:
@@ -218,7 +212,7 @@ data:
controllers are not making progress.
runbook: https://coreos.com/tectonic/docs/latest/troubleshooting/controller-recovery.html#recovering-a-controller-manager
summary: Controller manager is down
- kube-scheduler.rules.yaml: |+
+ kube-scheduler.rules.yaml: |
groups:
- name: kube-scheduler.rules
rules:
@@ -277,7 +271,7 @@ data:
to nodes.
runbook: https://coreos.com/tectonic/docs/latest/troubleshooting/controller-recovery.html#recovering-a-scheduler
summary: Scheduler is down
- kube-state-metrics.rules.yaml: |+
+ kube-state-metrics.rules.yaml: |
groups:
- name: kube-state-metrics.rules
rules:
@@ -337,7 +331,7 @@ data:
description: Pod {{$labels.namespace}}/{{$labels.pod}} was restarted {{$value}}
times within the last hour
summary: Pod is restarting frequently
- kubelet.rules.yaml: |+
+ kubelet.rules.yaml: |
groups:
- name: kubelet.rules
rules:
@@ -386,7 +380,7 @@ data:
description: Kubelet {{$labels.instance}} is running {{$value}} pods, close
to the limit of 110
summary: Kubelet is close to pod limit
- kubernetes.rules.yaml: |+
+ kubernetes.rules.yaml: |
groups:
- name: kubernetes.rules
rules:
@@ -477,7 +471,7 @@ data:
description: No API servers are reachable or all have disappeared from service
discovery
summary: No API servers are reachable
-
+
- alert: K8sCertificateExpirationNotice
labels:
severity: warning
@@ -485,7 +479,7 @@ data:
description: Kubernetes API Certificate is expiring soon (less than 7 days)
summary: Kubernetes API Certificate is expiering soon
expr: sum(apiserver_client_certificate_expiration_seconds_bucket{le="604800"}) > 0
-
+
- alert: K8sCertificateExpirationNotice
labels:
severity: critical
@@ -493,7 +487,7 @@ data:
description: Kubernetes API Certificate is expiring in less than 1 day
summary: Kubernetes API Certificate is expiering
expr: sum(apiserver_client_certificate_expiration_seconds_bucket{le="86400"}) > 0
- node.rules.yaml: |+
+ node.rules.yaml: |
groups:
- name: node.rules
rules:
@@ -541,105 +535,53 @@ data:
description: device {{$labels.device}} on node {{$labels.instance}} is running
full within the next 2 hours (mounted at {{$labels.mountpoint}})
summary: Node disk is running full within 2 hours
- prometheus.rules.yaml: |+
- groups:
- - name: prometheus.rules
- rules:
- - alert: PrometheusConfigReloadFailed
- expr: prometheus_config_last_reload_successful == 0
- for: 10m
- labels:
- severity: warning
- annotations:
- description: Reloading Prometheus' configuration has failed for {{$labels.namespace}}/{{$labels.pod}}
- summary: Reloading Promehteus' configuration failed
-
- - alert: PrometheusNotificationQueueRunningFull
- expr: predict_linear(prometheus_notifications_queue_length[5m], 60 * 30) > prometheus_notifications_queue_capacity
- for: 10m
- labels:
- severity: warning
- annotations:
- description: Prometheus' alert notification queue is running full for {{$labels.namespace}}/{{
- $labels.pod}}
- summary: Prometheus' alert notification queue is running full
-
- - alert: PrometheusErrorSendingAlerts
- expr: rate(prometheus_notifications_errors_total[5m]) / rate(prometheus_notifications_sent_total[5m])
- > 0.01
- for: 10m
- labels:
- severity: warning
- annotations:
- description: Errors while sending alerts from Prometheus {{$labels.namespace}}/{{
- $labels.pod}} to Alertmanager {{$labels.Alertmanager}}
- summary: Errors while sending alert from Prometheus
-
- - alert: PrometheusErrorSendingAlerts
- expr: rate(prometheus_notifications_errors_total[5m]) / rate(prometheus_notifications_sent_total[5m])
- > 0.03
- for: 10m
- labels:
- severity: critical
- annotations:
- description: Errors while sending alerts from Prometheus {{$labels.namespace}}/{{
- $labels.pod}} to Alertmanager {{$labels.Alertmanager}}
- summary: Errors while sending alerts from Prometheus
-
- - alert: PrometheusNotConnectedToAlertmanagers
- expr: prometheus_notifications_alertmanagers_discovered < 1
- for: 10m
- labels:
- severity: warning
- annotations:
- description: Prometheus {{ $labels.namespace }}/{{ $labels.pod}} is not connected
- to any Alertmanagers
- summary: Prometheus is not connected to any Alertmanagers
-
- - alert: PrometheusTSDBReloadsFailing
- expr: increase(prometheus_tsdb_reloads_failures_total[2h]) > 0
- for: 12h
- labels:
- severity: warning
- annotations:
- description: '{{$labels.job}} at {{$labels.instance}} had {{$value | humanize}}
- reload failures over the last four hours.'
- summary: Prometheus has issues reloading data blocks from disk
-
- - alert: PrometheusTSDBCompactionsFailing
- expr: increase(prometheus_tsdb_compactions_failed_total[2h]) > 0
- for: 12h
- labels:
- severity: warning
- annotations:
- description: '{{$labels.job}} at {{$labels.instance}} had {{$value | humanize}}
- compaction failures over the last four hours.'
- summary: Prometheus has issues compacting sample blocks
-
- - alert: PrometheusTSDBWALCorruptions
- expr: tsdb_wal_corruptions_total > 0
- for: 4h
- labels:
- severity: warning
- annotations:
- description: '{{$labels.job}} at {{$labels.instance}} has a corrupted write-ahead
- log (WAL).'
- summary: Prometheus write-ahead log is corrupted
-
- - alert: PrometheusNotIngestingSamples
- expr: rate(prometheus_tsdb_head_samples_appended_total[5m]) <= 0
- for: 10m
- labels:
- severity: warning
- annotations:
- description: "Prometheus {{ $labels.namespace }}/{{ $labels.pod}} isn't ingesting samples."
- summary: "Prometheus isn't ingesting samples"
-
- - alert: PrometheusTargetScapesDuplicate
- expr: increase(prometheus_target_scrapes_sample_duplicate_timestamp_total[5m]) > 0
- for: 10m
- labels:
- severity: warning
- annotations:
- description: "{{$labels.namespace}}/{{$labels.pod}} has many samples rejected due to duplicate timestamps but different values"
- summary: Prometheus has many samples rejected
+ prometheus.rules.yaml: "groups:\n- name: prometheus.rules\n rules:\n - alert:
+ PrometheusConfigReloadFailed\n expr: prometheus_config_last_reload_successful
+ == 0\n for: 10m\n labels:\n severity: warning\n annotations:\n description:
+ Reloading Prometheus' configuration has failed for {{$labels.namespace}}/{{$labels.pod}}\n
+ \ summary: Reloading Promehteus' configuration failed\n\n - alert: PrometheusNotificationQueueRunningFull\n
+ \ expr: predict_linear(prometheus_notifications_queue_length[5m], 60 * 30) >
+ prometheus_notifications_queue_capacity\n for: 10m\n labels:\n severity:
+ warning\n annotations:\n description: Prometheus' alert notification queue
+ is running full for {{$labels.namespace}}/{{\n $labels.pod}}\n summary:
+ Prometheus' alert notification queue is running full \n\n - alert: PrometheusErrorSendingAlerts\n
+ \ expr: rate(prometheus_notifications_errors_total[5m]) / rate(prometheus_notifications_sent_total[5m])\n
+ \ > 0.01\n for: 10m\n labels:\n severity: warning\n annotations:\n
+ \ description: Errors while sending alerts from Prometheus {{$labels.namespace}}/{{\n
+ \ $labels.pod}} to Alertmanager {{$labels.Alertmanager}}\n summary:
+ Errors while sending alert from Prometheus\n\n - alert: PrometheusErrorSendingAlerts\n
+ \ expr: rate(prometheus_notifications_errors_total[5m]) / rate(prometheus_notifications_sent_total[5m])\n
+ \ > 0.03\n for: 10m\n labels:\n severity: critical\n annotations:\n
+ \ description: Errors while sending alerts from Prometheus {{$labels.namespace}}/{{\n
+ \ $labels.pod}} to Alertmanager {{$labels.Alertmanager}}\n summary:
+ Errors while sending alerts from Prometheus\n\n - alert: PrometheusNotConnectedToAlertmanagers\n
+ \ expr: prometheus_notifications_alertmanagers_discovered < 1\n for: 10m\n
+ \ labels:\n severity: warning\n annotations:\n description: Prometheus
+ {{ $labels.namespace }}/{{ $labels.pod}} is not connected\n to any Alertmanagers\n
+ \ summary: Prometheus is not connected to any Alertmanagers\n\n - alert:
+ PrometheusTSDBReloadsFailing\n expr: increase(prometheus_tsdb_reloads_failures_total[2h])
+ > 0\n for: 12h\n labels:\n severity: warning\n annotations:\n description:
+ '{{$labels.job}} at {{$labels.instance}} had {{$value | humanize}}\n reload
+ failures over the last four hours.'\n summary: Prometheus has issues reloading
+ data blocks from disk\n\n - alert: PrometheusTSDBCompactionsFailing\n expr:
+ increase(prometheus_tsdb_compactions_failed_total[2h]) > 0\n for: 12h\n labels:\n
+ \ severity: warning\n annotations:\n description: '{{$labels.job}}
+ at {{$labels.instance}} had {{$value | humanize}}\n compaction failures
+ over the last four hours.'\n summary: Prometheus has issues compacting sample
+ blocks\n\n - alert: PrometheusTSDBWALCorruptions\n expr: tsdb_wal_corruptions_total
+ > 0\n for: 4h\n labels:\n severity: warning\n annotations:\n description:
+ '{{$labels.job}} at {{$labels.instance}} has a corrupted write-ahead\n log
+ (WAL).'\n summary: Prometheus write-ahead log is corrupted\n\n - alert:
+ PrometheusNotIngestingSamples\n expr: rate(prometheus_tsdb_head_samples_appended_total[5m])
+ <= 0\n for: 10m\n labels:\n severity: warning\n annotations:\n description:
+ \"Prometheus {{ $labels.namespace }}/{{ $labels.pod}} isn't ingesting samples.\"\n
+ \ summary: \"Prometheus isn't ingesting samples\"\n\n - alert: PrometheusTargetScapesDuplicate\n
+ \ expr: increase(prometheus_target_scrapes_sample_duplicate_timestamp_total[5m])
+ > 0\n for: 10m\n labels:\n severity: warning\n annotations:\n description:
+ \"{{$labels.namespace}}/{{$labels.pod}} has many samples rejected due to duplicate
+ timestamps but different values\"\n summary: Prometheus has many samples
+ rejected\n"
+kind: ConfigMap
+metadata:
+ name: prometheus-k8s-rules
+ namespace: monitoring
diff --git a/manifests/prometheus/prometheus-k8s-service-account.yaml b/manifests/prometheus-k8s/prometheus-k8s-service-account.yaml
similarity index 74%
rename from manifests/prometheus/prometheus-k8s-service-account.yaml
rename to manifests/prometheus-k8s/prometheus-k8s-service-account.yaml
index 58d5342dfcebb4f667b0f2a916d8e52396651a03..3e55fad675b8691f3aab29f152fdaa5416605745 100644
--- a/manifests/prometheus/prometheus-k8s-service-account.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-service-account.yaml
@@ -2,3 +2,4 @@ apiVersion: v1
kind: ServiceAccount
metadata:
name: prometheus-k8s
+ namespace: monitoring
diff --git a/manifests/prometheus/prometheus-k8s-service-monitor-alertmanager.yaml b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-alertmanager.yaml
similarity index 81%
rename from manifests/prometheus/prometheus-k8s-service-monitor-alertmanager.yaml
rename to manifests/prometheus-k8s/prometheus-k8s-service-monitor-alertmanager.yaml
index 19669e3e8f6a8fc9bd8343dcce34b06ea34fc680..e4e75ccc736027f662b95a4ff190afa3f61f50f5 100644
--- a/manifests/prometheus/prometheus-k8s-service-monitor-alertmanager.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-alertmanager.yaml
@@ -1,16 +1,17 @@
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
- name: alertmanager
labels:
k8s-app: alertmanager
+ name: alertmanager
+ namespace: monitoring
spec:
- selector:
- matchLabels:
- alertmanager: main
+ endpoints:
+ - interval: 30s
+ port: web
namespaceSelector:
matchNames:
- monitoring
- endpoints:
- - port: web
- interval: 30s
+ selector:
+ matchLabels:
+ alertmanager: main
diff --git a/manifests/prometheus/prometheus-k8s-service-monitor-apiserver.yaml b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-apiserver.yaml
similarity index 81%
rename from manifests/prometheus/prometheus-k8s-service-monitor-apiserver.yaml
rename to manifests/prometheus-k8s/prometheus-k8s-service-monitor-apiserver.yaml
index 40361f04f8c24eb2639cda02b310d93281d0d5f5..0cffe5418bf65194c8cb7d2b2e71e87394ab5310 100644
--- a/manifests/prometheus/prometheus-k8s-service-monitor-apiserver.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-apiserver.yaml
@@ -1,23 +1,24 @@
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
- name: kube-apiserver
labels:
k8s-app: apiserver
+ name: kube-apiserver
+ namespace: monitoring
spec:
- jobLabel: component
- selector:
- matchLabels:
- component: apiserver
- provider: kubernetes
- namespaceSelector:
- matchNames:
- - default
endpoints:
- - port: https
+ - bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
interval: 30s
+ port: https
scheme: https
tlsConfig:
caFile: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt
serverName: kubernetes
- bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
+ jobLabel: component
+ namespaceSelector:
+ matchNames:
+ - default
+ selector:
+ matchLabels:
+ component: apiserver
+ provider: kubernetes
diff --git a/manifests/prometheus/prometheus-k8s-service-monitor-coredns.yaml b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-coredns.yaml
similarity index 69%
rename from manifests/prometheus/prometheus-k8s-service-monitor-coredns.yaml
rename to manifests/prometheus-k8s/prometheus-k8s-service-monitor-coredns.yaml
index 362ac89959cf8e2be1835e6b08ef20feeca01b7e..12a4c5bf9fb3858dfb520d13c8b44b9a78bca341 100644
--- a/manifests/prometheus/prometheus-k8s-service-monitor-coredns.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-coredns.yaml
@@ -4,16 +4,17 @@ metadata:
labels:
k8s-app: coredns
name: coredns
+ namespace: monitoring
spec:
+ endpoints:
+ - bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
+ interval: 15s
+ port: http-metrics
jobLabel: k8s-app
+ namespaceSelector:
+ matchNames:
+ - kube-system
selector:
matchLabels:
- k8s-app: coredns
component: metrics
- namespaceSelector:
- matchNames:
- - kube-system
- endpoints:
- - port: http-metrics
- interval: 15s
- bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
+ k8s-app: coredns
diff --git a/manifests/prometheus/prometheus-k8s-service-monitor-kube-controller-manager.yaml b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-kube-controller-manager.yaml
similarity index 82%
rename from manifests/prometheus/prometheus-k8s-service-monitor-kube-controller-manager.yaml
rename to manifests/prometheus-k8s/prometheus-k8s-service-monitor-kube-controller-manager.yaml
index 681c320d87c13c05f267e4d9104538eb3ecfe748..dfb2a25d7508d24d71d1a2f47769c1c1427d4bd7 100644
--- a/manifests/prometheus/prometheus-k8s-service-monitor-kube-controller-manager.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-kube-controller-manager.yaml
@@ -1,17 +1,18 @@
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
- name: kube-controller-manager
labels:
k8s-app: kube-controller-manager
+ name: kube-controller-manager
+ namespace: monitoring
spec:
- jobLabel: k8s-app
endpoints:
- - port: http-metrics
- interval: 30s
- selector:
- matchLabels:
- k8s-app: kube-controller-manager
+ - interval: 30s
+ port: http-metrics
+ jobLabel: k8s-app
namespaceSelector:
matchNames:
- kube-system
+ selector:
+ matchLabels:
+ k8s-app: kube-controller-manager
diff --git a/manifests/prometheus/prometheus-k8s-service-monitor-kube-scheduler.yaml b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-kube-scheduler.yaml
similarity index 81%
rename from manifests/prometheus/prometheus-k8s-service-monitor-kube-scheduler.yaml
rename to manifests/prometheus-k8s/prometheus-k8s-service-monitor-kube-scheduler.yaml
index 6927f58e2995498101f5463378303bb909bdc7ff..f00db0e474023495ed4a825d61674367b1dfd7b8 100644
--- a/manifests/prometheus/prometheus-k8s-service-monitor-kube-scheduler.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-kube-scheduler.yaml
@@ -1,17 +1,18 @@
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
- name: kube-scheduler
labels:
k8s-app: kube-scheduler
+ name: kube-scheduler
+ namespace: monitoring
spec:
- jobLabel: k8s-app
endpoints:
- - port: http-metrics
- interval: 30s
- selector:
- matchLabels:
- k8s-app: kube-scheduler
+ - interval: 30s
+ port: http-metrics
+ jobLabel: k8s-app
namespaceSelector:
matchNames:
- kube-system
+ selector:
+ matchLabels:
+ k8s-app: kube-scheduler
diff --git a/manifests/prometheus/prometheus-k8s-service-monitor-kube-state-metrics.yaml b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-kube-state-metrics.yaml
similarity index 71%
rename from manifests/prometheus/prometheus-k8s-service-monitor-kube-state-metrics.yaml
rename to manifests/prometheus-k8s/prometheus-k8s-service-monitor-kube-state-metrics.yaml
index 1433a5feb168e30e8ffcb012970ccd1870ca05fb..cca52f696071938a307dc6386421fe0443a69cca 100644
--- a/manifests/prometheus/prometheus-k8s-service-monitor-kube-state-metrics.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-kube-state-metrics.yaml
@@ -1,28 +1,29 @@
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
- name: kube-state-metrics
labels:
k8s-app: kube-state-metrics
+ name: kube-state-metrics
+ namespace: monitoring
spec:
- jobLabel: k8s-app
- selector:
- matchLabels:
- k8s-app: kube-state-metrics
- namespaceSelector:
- matchNames:
- - monitoring
endpoints:
- - port: https-main
- scheme: https
- interval: 30s
+ - bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
honorLabels: true
- bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
+ interval: 30s
+ port: https-main
+ scheme: https
tlsConfig:
insecureSkipVerify: true
- - port: https-self
- scheme: https
+ - bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
interval: 30s
- bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
+ port: https-self
+ scheme: https
tlsConfig:
insecureSkipVerify: true
+ jobLabel: k8s-app
+ namespaceSelector:
+ matchNames:
+ - monitoring
+ selector:
+ matchLabels:
+ k8s-app: kube-state-metrics
diff --git a/manifests/prometheus/prometheus-k8s-service-monitor-kubelet.yaml b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-kubelet.yaml
similarity index 71%
rename from manifests/prometheus/prometheus-k8s-service-monitor-kubelet.yaml
rename to manifests/prometheus-k8s/prometheus-k8s-service-monitor-kubelet.yaml
index 16c9752df755bb69d7c04ee8f14cccd8548c7322..06ec7fc84b3efda7befa08d1d6256136bbac4ef4 100644
--- a/manifests/prometheus/prometheus-k8s-service-monitor-kubelet.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-kubelet.yaml
@@ -1,29 +1,30 @@
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
- name: kubelet
labels:
k8s-app: kubelet
+ name: kubelet
+ namespace: monitoring
spec:
- jobLabel: k8s-app
endpoints:
- - port: https-metrics
- scheme: https
+ - bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
interval: 30s
+ port: https-metrics
+ scheme: https
tlsConfig:
insecureSkipVerify: true
- bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
- - port: https-metrics
- scheme: https
- path: /metrics/cadvisor
- interval: 30s
+ - bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
honorLabels: true
+ interval: 30s
+ path: /metrics/cadvisor
+ port: https-metrics
+ scheme: https
tlsConfig:
insecureSkipVerify: true
- bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
- selector:
- matchLabels:
- k8s-app: kubelet
+ jobLabel: k8s-app
namespaceSelector:
matchNames:
- kube-system
+ selector:
+ matchLabels:
+ k8s-app: kubelet
diff --git a/manifests/prometheus/prometheus-k8s-service-monitor-node-exporter.yaml b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-node-exporter.yaml
similarity index 78%
rename from manifests/prometheus/prometheus-k8s-service-monitor-node-exporter.yaml
rename to manifests/prometheus-k8s/prometheus-k8s-service-monitor-node-exporter.yaml
index 0dd72e759a7234996a16b2e724ca703baea6b442..529f2944f61f18a616a1a99121214e660ae3effd 100644
--- a/manifests/prometheus/prometheus-k8s-service-monitor-node-exporter.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-node-exporter.yaml
@@ -1,21 +1,22 @@
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
- name: node-exporter
labels:
k8s-app: node-exporter
+ name: node-exporter
+ namespace: monitoring
spec:
- jobLabel: k8s-app
- selector:
- matchLabels:
- k8s-app: node-exporter
- namespaceSelector:
- matchNames:
- - monitoring
endpoints:
- - port: https
- scheme: https
+ - bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
interval: 30s
- bearerTokenFile: /var/run/secrets/kubernetes.io/serviceaccount/token
+ port: https
+ scheme: https
tlsConfig:
insecureSkipVerify: true
+ jobLabel: k8s-app
+ namespaceSelector:
+ matchNames:
+ - monitoring
+ selector:
+ matchLabels:
+ k8s-app: node-exporter
diff --git a/manifests/prometheus/prometheus-k8s-service-monitor-prometheus-operator.yaml b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-prometheus-operator.yaml
similarity index 90%
rename from manifests/prometheus/prometheus-k8s-service-monitor-prometheus-operator.yaml
rename to manifests/prometheus-k8s/prometheus-k8s-service-monitor-prometheus-operator.yaml
index 0b8028e778b4fa8aef77abaed10860c8e4b09ba7..10e0059aa4c6f3785dc0beda7e2b36fbf8b91a44 100644
--- a/manifests/prometheus/prometheus-k8s-service-monitor-prometheus-operator.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-prometheus-operator.yaml
@@ -1,9 +1,10 @@
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
- name: prometheus-operator
labels:
k8s-app: prometheus-operator
+ name: prometheus-operator
+ namespace: monitoring
spec:
endpoints:
- port: http
diff --git a/manifests/prometheus/prometheus-k8s-service-monitor-prometheus.yaml b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-prometheus.yaml
similarity index 81%
rename from manifests/prometheus/prometheus-k8s-service-monitor-prometheus.yaml
rename to manifests/prometheus-k8s/prometheus-k8s-service-monitor-prometheus.yaml
index c3d11e57b38fb77a1d26b19a0c7dd0a8d8b044ec..90b25476e057ae87e33a0708b8b3f01865bcabe2 100644
--- a/manifests/prometheus/prometheus-k8s-service-monitor-prometheus.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-service-monitor-prometheus.yaml
@@ -1,16 +1,17 @@
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
- name: prometheus
labels:
k8s-app: prometheus
+ name: prometheus
+ namespace: monitoring
spec:
- selector:
- matchLabels:
- prometheus: k8s
+ endpoints:
+ - interval: 30s
+ port: web
namespaceSelector:
matchNames:
- monitoring
- endpoints:
- - port: web
- interval: 30s
+ selector:
+ matchLabels:
+ prometheus: k8s
diff --git a/manifests/prometheus/prometheus-k8s-service.yaml b/manifests/prometheus-k8s/prometheus-k8s-service.yaml
similarity index 77%
rename from manifests/prometheus/prometheus-k8s-service.yaml
rename to manifests/prometheus-k8s/prometheus-k8s-service.yaml
index 5cd3b65b7b50a0e21da2416a65d771151f375536..85b007f8b787324932918166acdc470aaef2b21c 100644
--- a/manifests/prometheus/prometheus-k8s-service.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s-service.yaml
@@ -4,13 +4,12 @@ metadata:
labels:
prometheus: k8s
name: prometheus-k8s
+ namespace: monitoring
spec:
- type: NodePort
ports:
- name: web
- nodePort: 30900
port: 9090
- protocol: TCP
targetPort: web
selector:
+ app: prometheus
prometheus: k8s
diff --git a/manifests/prometheus/prometheus-k8s.yaml b/manifests/prometheus-k8s/prometheus-k8s.yaml
similarity index 54%
rename from manifests/prometheus/prometheus-k8s.yaml
rename to manifests/prometheus-k8s/prometheus-k8s.yaml
index 8f243eb08eb6612ef5e47d2bbc19313378189994..324d96c71f3d74c6ea37bd8a0dfbf9eb5ead1388 100644
--- a/manifests/prometheus/prometheus-k8s.yaml
+++ b/manifests/prometheus-k8s/prometheus-k8s.yaml
@@ -1,29 +1,27 @@
apiVersion: monitoring.coreos.com/v1
kind: Prometheus
metadata:
- name: k8s
labels:
prometheus: k8s
+ name: k8s
+ namespace: monitoring
spec:
+ alerting:
+ alertmanagers:
+ - name: alertmanager-main
+ namespace: monitoring
+ port: web
replicas: 2
- version: v2.2.1
- serviceAccountName: prometheus-k8s
- serviceMonitorSelector:
- matchExpressions:
- - {key: k8s-app, operator: Exists}
- ruleSelector:
- matchLabels:
- role: alert-rules
- prometheus: k8s
resources:
requests:
- # 2Gi is default, but won't schedule if you don't have a node with >2Gi
- # memory. Modify based on your target and time-series count for
- # production use. This value is mainly meant for demonstration/testing
- # purposes.
memory: 400Mi
- alerting:
- alertmanagers:
- - namespace: monitoring
- name: alertmanager-main
- port: web
+ ruleSelector:
+ matchLabels:
+ prometheus: k8s
+ role: alert-rules
+ serviceAccountName: prometheus-k8s
+ serviceMonitorSelector:
+ matchExpressions:
+ - key: k8s-app
+ operator: Exists
+ version: v2.2.1
diff --git a/manifests/prometheus-operator/prometheus-operator-cluster-role-binding.yaml b/manifests/prometheus-operator/prometheus-operator-cluster-role-binding.yaml
index e7e03a29bcb9a50a9f4e80853d25dc9782568f39..8f1b4d36ca2d94b20ab75794e37849845c0b71d4 100644
--- a/manifests/prometheus-operator/prometheus-operator-cluster-role-binding.yaml
+++ b/manifests/prometheus-operator/prometheus-operator-cluster-role-binding.yaml
@@ -1,4 +1,4 @@
-apiVersion: rbac.authorization.k8s.io/v1beta1
+apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
name: prometheus-operator
diff --git a/manifests/prometheus-operator/prometheus-operator-cluster-role.yaml b/manifests/prometheus-operator/prometheus-operator-cluster-role.yaml
index 1b13c89951c4a0c88ac94d88a97840f9adb558a5..9ecd917e9d2258457a22a80ee032f3afc6d04abf 100644
--- a/manifests/prometheus-operator/prometheus-operator-cluster-role.yaml
+++ b/manifests/prometheus-operator/prometheus-operator-cluster-role.yaml
@@ -1,4 +1,4 @@
-apiVersion: rbac.authorization.k8s.io/v1beta1
+apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
name: prometheus-operator
@@ -8,13 +8,13 @@ rules:
resources:
- thirdpartyresources
verbs:
- - "*"
+ - '*'
- apiGroups:
- apiextensions.k8s.io
resources:
- customresourcedefinitions
verbs:
- - "*"
+ - '*'
- apiGroups:
- monitoring.coreos.com
resources:
@@ -24,31 +24,45 @@ rules:
- alertmanagers/finalizers
- servicemonitors
verbs:
- - "*"
+ - '*'
- apiGroups:
- apps
resources:
- statefulsets
- verbs: ["*"]
-- apiGroups: [""]
+ verbs:
+ - '*'
+- apiGroups:
+ - ""
resources:
- configmaps
- secrets
- verbs: ["*"]
-- apiGroups: [""]
+ verbs:
+ - '*'
+- apiGroups:
+ - ""
resources:
- pods
- verbs: ["list", "delete"]
-- apiGroups: [""]
+ verbs:
+ - list
+ - delete
+- apiGroups:
+ - ""
resources:
- services
- - endpoints
- verbs: ["get", "create", "update"]
-- apiGroups: [""]
+ verbs:
+ - get
+ - create
+ - update
+- apiGroups:
+ - ""
resources:
- nodes
- verbs: ["list", "watch"]
-- apiGroups: [""]
+ verbs:
+ - list
+ - watch
+- apiGroups:
+ - ""
resources:
- namespaces
- verbs: ["list"]
+ verbs:
+ - list
diff --git a/manifests/prometheus-operator/prometheus-operator.yaml b/manifests/prometheus-operator/prometheus-operator-deployment.yaml
similarity index 97%
rename from manifests/prometheus-operator/prometheus-operator.yaml
rename to manifests/prometheus-operator/prometheus-operator-deployment.yaml
index d00301111466c972e8f97e5175ade17ded9e5ac9..1c70ef0a51a2b25756191298de971d18ea77ec16 100644
--- a/manifests/prometheus-operator/prometheus-operator.yaml
+++ b/manifests/prometheus-operator/prometheus-operator-deployment.yaml
@@ -4,6 +4,7 @@ metadata:
labels:
k8s-app: prometheus-operator
name: prometheus-operator
+ namespace: monitoring
spec:
replicas: 1
selector:
diff --git a/manifests/prometheus-operator/prometheus-operator-service-account.yaml b/manifests/prometheus-operator/prometheus-operator-service-account.yaml
index 38d18cce71105613f954591c54f1eaa4ddd15b06..11898a6fbef1c3e2e47f4f554967e5d5651c808c 100644
--- a/manifests/prometheus-operator/prometheus-operator-service-account.yaml
+++ b/manifests/prometheus-operator/prometheus-operator-service-account.yaml
@@ -2,3 +2,4 @@ apiVersion: v1
kind: ServiceAccount
metadata:
name: prometheus-operator
+ namespace: monitoring
diff --git a/manifests/prometheus-operator/prometheus-operator-service.yaml b/manifests/prometheus-operator/prometheus-operator-service.yaml
index 8882d4a793ed26bc1e4fb3c85134ca16061a25f4..8a82538740f2cb4f3563918b4e27669e24ea89d0 100644
--- a/manifests/prometheus-operator/prometheus-operator-service.yaml
+++ b/manifests/prometheus-operator/prometheus-operator-service.yaml
@@ -2,14 +2,11 @@ apiVersion: v1
kind: Service
metadata:
name: prometheus-operator
- labels:
- k8s-app: prometheus-operator
+ namespace: monitoring
spec:
- type: ClusterIP
ports:
- name: http
port: 8080
targetPort: http
- protocol: TCP
selector:
k8s-app: prometheus-operator
diff --git a/manifests/prometheus/prometheus-k8s-role-bindings.yaml b/manifests/prometheus/prometheus-k8s-role-bindings.yaml
deleted file mode 100644
index 5f190e7ab1a21e09673c6a9bef9e683b22e98304..0000000000000000000000000000000000000000
--- a/manifests/prometheus/prometheus-k8s-role-bindings.yaml
+++ /dev/null
@@ -1,54 +0,0 @@
-apiVersion: rbac.authorization.k8s.io/v1beta1
-kind: RoleBinding
-metadata:
- name: prometheus-k8s
- namespace: monitoring
-roleRef:
- apiGroup: rbac.authorization.k8s.io
- kind: Role
- name: prometheus-k8s
-subjects:
-- kind: ServiceAccount
- name: prometheus-k8s
- namespace: monitoring
----
-apiVersion: rbac.authorization.k8s.io/v1beta1
-kind: RoleBinding
-metadata:
- name: prometheus-k8s
- namespace: kube-system
-roleRef:
- apiGroup: rbac.authorization.k8s.io
- kind: Role
- name: prometheus-k8s
-subjects:
-- kind: ServiceAccount
- name: prometheus-k8s
- namespace: monitoring
----
-apiVersion: rbac.authorization.k8s.io/v1beta1
-kind: RoleBinding
-metadata:
- name: prometheus-k8s
- namespace: default
-roleRef:
- apiGroup: rbac.authorization.k8s.io
- kind: Role
- name: prometheus-k8s
-subjects:
-- kind: ServiceAccount
- name: prometheus-k8s
- namespace: monitoring
----
-apiVersion: rbac.authorization.k8s.io/v1beta1
-kind: ClusterRoleBinding
-metadata:
- name: prometheus-k8s
-roleRef:
- apiGroup: rbac.authorization.k8s.io
- kind: ClusterRole
- name: prometheus-k8s
-subjects:
-- kind: ServiceAccount
- name: prometheus-k8s
- namespace: monitoring
diff --git a/manifests/prometheus/prometheus-k8s-roles.yaml b/manifests/prometheus/prometheus-k8s-roles.yaml
deleted file mode 100644
index 4f738e776e7b5de3a5d6fe5a543392bf36ba0213..0000000000000000000000000000000000000000
--- a/manifests/prometheus/prometheus-k8s-roles.yaml
+++ /dev/null
@@ -1,55 +0,0 @@
-apiVersion: rbac.authorization.k8s.io/v1beta1
-kind: Role
-metadata:
- name: prometheus-k8s
- namespace: monitoring
-rules:
-- apiGroups: [""]
- resources:
- - nodes
- - services
- - endpoints
- - pods
- verbs: ["get", "list", "watch"]
-- apiGroups: [""]
- resources:
- - configmaps
- verbs: ["get"]
----
-apiVersion: rbac.authorization.k8s.io/v1beta1
-kind: Role
-metadata:
- name: prometheus-k8s
- namespace: kube-system
-rules:
-- apiGroups: [""]
- resources:
- - services
- - endpoints
- - pods
- verbs: ["get", "list", "watch"]
----
-apiVersion: rbac.authorization.k8s.io/v1beta1
-kind: Role
-metadata:
- name: prometheus-k8s
- namespace: default
-rules:
-- apiGroups: [""]
- resources:
- - services
- - endpoints
- - pods
- verbs: ["get", "list", "watch"]
----
-apiVersion: rbac.authorization.k8s.io/v1beta1
-kind: ClusterRole
-metadata:
- name: prometheus-k8s
-rules:
-- apiGroups: [""]
- resources:
- - nodes/metrics
- verbs: ["get"]
-- nonResourceURLs: ["/metrics"]
- verbs: ["get"]
diff --git a/manifests/prometheus/prometheus-k8s-service-coredns-metrics.yaml b/manifests/prometheus/prometheus-k8s-service-coredns-metrics.yaml
deleted file mode 100644
index cd90a55e614a0c39fa5a59a96f91adc93ef121c6..0000000000000000000000000000000000000000
--- a/manifests/prometheus/prometheus-k8s-service-coredns-metrics.yaml
+++ /dev/null
@@ -1,18 +0,0 @@
-apiVersion: v1
-kind: Service
-metadata:
- name: coredns-prometheus-discovery
- namespace: kube-system
- labels:
- k8s-app: coredns
- component: metrics
-spec:
- ports:
- - name: http-metrics
- port: 9153
- protocol: TCP
- targetPort: 9153
- selector:
- k8s-app: coredns
- type: ClusterIP
- clusterIP: None
diff --git a/requirements.txt b/requirements.txt
deleted file mode 100644
index fdc2b200388fc3a7d1fc9f67c8e846aebff90d8b..0000000000000000000000000000000000000000
--- a/requirements.txt
+++ /dev/null
@@ -1 +0,0 @@
-git+https://github.com/aknuds1/grafanalib.git@v0.4.0
\ No newline at end of file