kube-prometheus
Note that everything is experimental and may change significantly at any time.
This repository collects Kubernetes manifests, Grafana dashboards, and Prometheus rules combined with documentation and scripts to provide easy to operate end-to-end Kubernetes cluster monitoring with Prometheus using the Prometheus Operator.
The content of this project is written in jsonnet. This project could both be described as a package as well as a library.
Components included in this package:
- The Prometheus Operator
- Highly available Prometheus
- Highly available Alertmanager
- Prometheus node-exporter
- Prometheus Adapter for Kubernetes Metrics APIs
- kube-state-metrics
- Grafana
This stack is meant for cluster monitoring, so it is pre-configured to collect metrics from all Kubernetes components. In addition to that it delivers a default set of dashboards and alerting rules. Many of the useful dashboards and alerts come from the kubernetes-mixin project, similar to this project it provides composable jsonnet as a library for users to customize to their needs.
Warning
If you are migrating from release-0.7
branch or earlier please read what changed and how to migrate in our guide.
Table of contents
-
kube-prometheus
- Warning
- Table of contents
- Prerequisites
- Compatibility
- Quickstart
- Customizing Kube-Prometheus
- Update from upstream project
- Configuration
-
Customization Examples
- Cluster Creation Tools
- Internal Registry
- NodePorts
- Prometheus Object Name
- node-exporter DaemonSet namespace
- Alertmanager configuration
- Adding additional namespaces to monitor
- Monitoring all namespaces
- Static etcd configuration
- Pod Anti-Affinity
- Stripping container resource limits
- Customizing Prometheus alerting/recording rules and Grafana dashboards
- Exposing Prometheus/Alermanager/Grafana via Ingress
- Setting up a blackbox exporter
- Minikube Example
- Continuous Delivery
- Troubleshooting
- Contributing
- License
Prerequisites
You will need a Kubernetes cluster, that's it! By default it is assumed, that the kubelet uses token authentication and authorization, as otherwise Prometheus needs a client certificate, which gives it full access to the kubelet, rather than just the metrics. Token authentication and authorization allows more fine grained and easier access control.
This means the kubelet configuration must contain these flags:
-
--authentication-token-webhook=true
This flag enables, that aServiceAccount
token can be used to authenticate against the kubelet(s). This can also be enabled by setting the kubelet configuration valueauthentication.webhook.enabled
totrue
. -
--authorization-mode=Webhook
This flag enables, that the kubelet will perform an RBAC request with the API to determine, whether the requesting entity (Prometheus in this case) is allowed to access a resource, in specific for this project the/metrics
endpoint. This can also be enabled by setting the kubelet configuration valueauthorization.mode
toWebhook
.
This stack provides resource metrics by deploying the Prometheus Adapter. This adapter is an Extension API Server and Kubernetes needs to be have this feature enabled, otherwise the adapter has no effect, but is still deployed.
minikube
To try out this stack, start minikube with the following command:
$ minikube delete && minikube start --kubernetes-version=v1.20.0 --memory=6g --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
The kube-prometheus stack includes a resource metrics API server, so the metrics-server addon is not necessary. Ensure the metrics-server addon is disabled on minikube:
$ minikube addons disable metrics-server
Compatibility
Kubernetes compatibility matrix
The following versions are supported and work as we test against these versions in their respective branches. But note that other versions might work!
kube-prometheus stack | Kubernetes 1.18 | Kubernetes 1.19 | Kubernetes 1.20 | Kubernetes 1.21 | Kubernetes 1.22 |
---|---|---|---|---|---|
release-0.6 |
✗ | ✗ | ✗ | ✗ | |
release-0.7 |
✗ | ✗ | ✗ | ||
release-0.8 |
✗ | ✗ | ✗ | ||
release-0.9 |
✗ | ✗ | ✗ | ||
HEAD |
✗ | ✗ | ✗ |
Quickstart
Note: For versions before Kubernetes v1.21.z refer to the Kubernetes compatibility matrix in order to choose a compatible branch.
This project is intended to be used as a library (i.e. the intent is not for you to create your own modified copy of this repository).
Though for a quickstart a compiled version of the Kubernetes manifests generated with this library (specifically with example.jsonnet
) is checked into this repository in order to try the content out quickly. To try out the stack un-customized run:
- Create the monitoring stack using the config in the
manifests
directory:
# Create the namespace and CRDs, and then wait for them to be available before creating the remaining resources
kubectl create -f manifests/setup
until kubectl get servicemonitors --all-namespaces ; do date; sleep 1; echo ""; done
kubectl create -f manifests/
We create the namespace and CustomResourceDefinitions first to avoid race conditions when deploying the monitoring components.
Alternatively, the resources in both folders can be applied with a single command
kubectl create -f manifests/setup -f manifests
, but it may be necessary to run the command multiple times for all components to
be created successfullly.
- And to teardown the stack:
kubectl delete --ignore-not-found=true -f manifests/ -f manifests/setup
Access the dashboards
Prometheus, Grafana, and Alertmanager dashboards can be accessed quickly using kubectl port-forward
after running the quickstart via the commands below. Kubernetes 1.10 or later is required.
Note: There are instructions on how to route to these pods behind an ingress controller in the Exposing Prometheus/Alermanager/Grafana via Ingress section.
Prometheus
$ kubectl --namespace monitoring port-forward svc/prometheus-k8s 9090
Then access via http://localhost:9090
Grafana
$ kubectl --namespace monitoring port-forward svc/grafana 3000
Then access via http://localhost:3000 and use the default grafana user:password of admin:admin
.
Alert Manager
$ kubectl --namespace monitoring port-forward svc/alertmanager-main 9093
Then access via http://localhost:9093
Customizing Kube-Prometheus
This section:
- describes how to customize the kube-prometheus library via compiling the kube-prometheus manifests yourself (as an alternative to the Quickstart section).
- still doesn't require you to make a copy of this entire repository, but rather only a copy of a few select files.
Installing
The content of this project consists of a set of jsonnet files making up a library to be consumed.
Install this library in your own project with jsonnet-bundler (the jsonnet package manager):
$ mkdir my-kube-prometheus; cd my-kube-prometheus
$ jb init # Creates the initial/empty `jsonnetfile.json`
# Install the kube-prometheus dependency
$ jb install github.com/prometheus-operator/kube-prometheus/jsonnet/kube-prometheus@release-0.9 # Creates `vendor/` & `jsonnetfile.lock.json`, and fills in `jsonnetfile.json`
$ wget https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/release-0.9/example.jsonnet -O example.jsonnet
$ wget https://raw.githubusercontent.com/prometheus-operator/kube-prometheus/release-0.9/build.sh -O build.sh
jb
can be installed withgo get github.com/jsonnet-bundler/jsonnet-bundler/cmd/jb
An e.g. of how to install a given version of this library:
jb install github.com/prometheus-operator/kube-prometheus/jsonnet/kube-prometheus@release-0.9
In order to update the kube-prometheus dependency, simply use the jsonnet-bundler update functionality:
$ jb update
Compiling
e.g. of how to compile the manifests: ./build.sh example.jsonnet
before compiling, install
gojsontoyaml
tool withgo get github.com/brancz/gojsontoyaml
andjsonnet
withgo get github.com/google/go-jsonnet/cmd/jsonnet
Here's example.jsonnet:
Note: some of the following components must be configured beforehand. See configuration and customization-examples.
local kp =
(import 'kube-prometheus/main.libsonnet') +
// Uncomment the following imports to enable its patches
// (import 'kube-prometheus/addons/anti-affinity.libsonnet') +
// (import 'kube-prometheus/addons/managed-cluster.libsonnet') +
// (import 'kube-prometheus/addons/node-ports.libsonnet') +
// (import 'kube-prometheus/addons/static-etcd.libsonnet') +
// (import 'kube-prometheus/addons/custom-metrics.libsonnet') +
// (import 'kube-prometheus/addons/external-metrics.libsonnet') +
{
values+:: {
common+: {
namespace: 'monitoring',
},
},
};
{ 'setup/0namespace-namespace': kp.kubePrometheus.namespace } +
{
['setup/prometheus-operator-' + name]: kp.prometheusOperator[name]
for name in std.filter((function(name) name != 'serviceMonitor' && name != 'prometheusRule'), std.objectFields(kp.prometheusOperator))
} +
// serviceMonitor and prometheusRule are separated so that they can be created after the CRDs are ready
{ 'prometheus-operator-serviceMonitor': kp.prometheusOperator.serviceMonitor } +
{ 'prometheus-operator-prometheusRule': kp.prometheusOperator.prometheusRule } +
{ 'kube-prometheus-prometheusRule': kp.kubePrometheus.prometheusRule } +
{ ['alertmanager-' + name]: kp.alertmanager[name] for name in std.objectFields(kp.alertmanager) } +
{ ['blackbox-exporter-' + name]: kp.blackboxExporter[name] for name in std.objectFields(kp.blackboxExporter) } +
{ ['grafana-' + name]: kp.grafana[name] for name in std.objectFields(kp.grafana) } +
{ ['kube-state-metrics-' + name]: kp.kubeStateMetrics[name] for name in std.objectFields(kp.kubeStateMetrics) } +
{ ['kubernetes-' + name]: kp.kubernetesControlPlane[name] for name in std.objectFields(kp.kubernetesControlPlane) }
{ ['node-exporter-' + name]: kp.nodeExporter[name] for name in std.objectFields(kp.nodeExporter) } +
{ ['prometheus-' + name]: kp.prometheus[name] for name in std.objectFields(kp.prometheus) } +
{ ['prometheus-adapter-' + name]: kp.prometheusAdapter[name] for name in std.objectFields(kp.prometheusAdapter) }
And here's the build.sh script (which uses vendor/
to render all manifests in a json structure of {filename: manifest-content}
):
#!/usr/bin/env bash
# This script uses arg $1 (name of *.jsonnet file to use) to generate the manifests/*.yaml files.
set -e
set -x
# only exit with zero if all commands of the pipeline exit successfully
set -o pipefail
# Make sure to use project tooling
PATH="$(pwd)/tmp/bin:${PATH}"
# Make sure to start with a clean 'manifests' dir
rm -rf manifests
mkdir -p manifests/setup
# Calling gojsontoyaml is optional, but we would like to generate yaml, not json
jsonnet -J vendor -m manifests "${1-example.jsonnet}" | xargs -I{} sh -c 'cat {} | gojsontoyaml > {}.yaml' -- {}
# Make sure to remove json files
find manifests -type f ! -name '*.yaml' -delete
rm -f kustomization
Note you need
jsonnet
(go get github.com/google/go-jsonnet/cmd/jsonnet
) andgojsontoyaml
(go get github.com/brancz/gojsontoyaml
) installed to runbuild.sh
. If you just want json output, not yaml, then you can skip the pipe and everything afterwards.
This script runs the jsonnet code, then reads each key of the generated json and uses that as the file name, and writes the value of that key to that file, and converts each json manifest to yaml.
Apply the kube-prometheus stack
The previous steps (compilation) has created a bunch of manifest files in the manifest/ folder.
Now simply use kubectl
to install Prometheus and Grafana as per your configuration:
# Update the namespace and CRDs, and then wait for them to be available before creating the remaining resources
$ kubectl apply -f manifests/setup
$ kubectl apply -f manifests/
Alternatively, the resources in both folders can be applied with a single command
kubectl apply -Rf manifests
, but it may be necessary to run the command multiple times for all components to
be created successfullly.
Check the monitoring namespace (or the namespace you have specific in namespace:
) and make sure the pods are running. Prometheus and Grafana should be up and running soon.
Containerized Installing and Compiling
If you don't care to have jb
nor jsonnet
nor gojsontoyaml
installed, then use quay.io/coreos/jsonnet-ci
container image. Do the following from this kube-prometheus
directory:
$ docker run --rm -v $(pwd):$(pwd) --workdir $(pwd) quay.io/coreos/jsonnet-ci jb update
$ docker run --rm -v $(pwd):$(pwd) --workdir $(pwd) quay.io/coreos/jsonnet-ci ./build.sh example.jsonnet
Update from upstream project
You may wish to fetch changes made on this project so they are available to you.
Update jb
jb
may have been updated so it's a good idea to get the latest version of this binary:
$ go get -u github.com/jsonnet-bundler/jsonnet-bundler/cmd/jb
Update kube-prometheus
The command below will sync with upstream project:
$ jb update
Compile the manifests and apply
Once updated, just follow the instructions under "Compiling" and "Apply the kube-prometheus stack" to apply the changes to your cluster.
Configuration
Jsonnet has the concept of hidden fields. These are fields, that are not going to be rendered in a result. This is used to configure the kube-prometheus components in jsonnet. In the example jsonnet code of the above Customizing Kube-Prometheus section, you can see an example of this, where the namespace
is being configured to be monitoring
. In order to not override the whole object, use the +::
construct of jsonnet, to merge objects, this way you can override individual settings, but retain all other settings and defaults.
The available fields and their default values can be seen in main.libsonnet. Note that many of the fields get their default values from variables, and for example the version numbers are imported from versions.json.
Configuration is mainly done in the values
map. You can see this being used in the example.jsonnet
to set the namespace to monitoring
. This is done in the common
field, which all other components take their default value from. See for example how Alertmanager is configured in main.libsonnet
:
alertmanager: {
name: 'main',
// Use the namespace specified under values.common by default.
namespace: $.values.common.namespace,
version: $.values.common.versions.alertmanager,
image: $.values.common.images.alertmanager,
mixin+: { ruleLabels: $.values.common.ruleLabels },
},
The grafana definition is located in a different project (https://github.com/brancz/kubernetes-grafana), but needed configuration can be customized from the same top level values
field. For example to allow anonymous access to grafana, add the following values
section:
grafana+:: {
config: { // http://docs.grafana.org/installation/configuration/
sections: {
"auth.anonymous": {enabled: true},
},
},
},
Customization Examples
Jsonnet is a turing complete language, any logic can be reflected in it. It also has powerful merge functionalities, allowing sophisticated customizations of any kind simply by merging it into the object the library provides.
Cluster Creation Tools
A common example is that not all Kubernetes clusters are created exactly the same way, meaning the configuration to monitor them may be slightly different. For the following clusters there are mixins available to easily configure them:
- aws
- bootkube
- eks
- gke
- kops-coredns
- kubeadm
- kubespray
These mixins are selectable via the platform
field of kubePrometheus:
(import 'kube-prometheus/main.libsonnet') +
{
values+:: {
common+: {
platform: 'example-platform',
},
},
}