From 3b8e685082c403c9844895e972e4a4dd1ee3febd Mon Sep 17 00:00:00 2001
From: karancode <karan.thanvi@paypay-corp.co.jp>
Date: Tue, 22 Oct 2019 02:24:07 +0900
Subject: [PATCH] add aws-eks-cni service
---
examples/kustomize.jsonnet | 2 +-
.../kube-prometheus-aws-eks-cni.libsonnet | 13 +++
jsonnetfile.lock.json | 14 +--
kustomization.yaml | 1 +
manifests/grafana-dashboardDefinitions.yaml | 98 ++++++++++++-------
manifests/prometheus-rules.yaml | 16 +++
.../prometheus-serviceMonitorAwsEksCNI.yaml | 19 ++++
7 files changed, 119 insertions(+), 44 deletions(-)
create mode 100644 jsonnet/kube-prometheus/kube-prometheus-aws-eks-cni.libsonnet
create mode 100644 manifests/prometheus-serviceMonitorAwsEksCNI.yaml
diff --git a/examples/kustomize.jsonnet b/examples/kustomize.jsonnet
index db3ba344..438e69b1 100644
--- a/examples/kustomize.jsonnet
+++ b/examples/kustomize.jsonnet
@@ -1,5 +1,5 @@
local kp =
- (import 'kube-prometheus/kube-prometheus.libsonnet') + {
+ (import 'kube-prometheus/kube-prometheus-aws-eks-cni.libsonnet') + {
_config+:: {
namespace: 'monitoring',
},
diff --git a/jsonnet/kube-prometheus/kube-prometheus-aws-eks-cni.libsonnet b/jsonnet/kube-prometheus/kube-prometheus-aws-eks-cni.libsonnet
new file mode 100644
index 00000000..e6dce38b
--- /dev/null
+++ b/jsonnet/kube-prometheus/kube-prometheus-aws-eks-cni.libsonnet
@@ -0,0 +1,13 @@
+local k = import 'ksonnet/ksonnet.beta.4/k.libsonnet';
+local service = k.core.v1.service;
+local servicePort = k.core.v1.service.mixin.spec.portsType;
+
+{
+ prometheus+: {
+ kubePrometheusAwsEksCniMetricService:
+ service.new('aws-eks-cni', { 'k8s-app' : 'eks-cni' } , servicePort.newNamed('cni-metrics-port', 61678, 61678)) +
+ service.mixin.metadata.withNamespace('kube-system') +
+ service.mixin.metadata.withLabels({ 'k8s-app': 'eks-cni' }) +
+ service.mixin.spec.withClusterIp('None'),
+ },
+}
diff --git a/jsonnetfile.lock.json b/jsonnetfile.lock.json
index 5d054e2b..66bfafd6 100644
--- a/jsonnetfile.lock.json
+++ b/jsonnetfile.lock.json
@@ -7,7 +7,7 @@
"directory": "jsonnet/kube-prometheus"
}
},
- "version": ""
+ "version": "aws_eks_cni"
},
{
"name": "ksonnet",
@@ -27,7 +27,7 @@
"subdir": ""
}
},
- "version": "15ddfa20a6921ffbd43172eb54f6bdc1bcf8d3d3"
+ "version": "3ad401ea3ef7fb0879298fa411772984ffa7f31f"
},
{
"name": "grafonnet",
@@ -37,7 +37,7 @@
"subdir": "grafonnet"
}
},
- "version": "69bc267211790a1c3f4ea6e6211f3e8ffe22f987"
+ "version": "47db72da03fc4a7a0658a87791e13c3315a3a252"
},
{
"name": "grafana-builder",
@@ -47,7 +47,7 @@
"subdir": "grafana-builder"
}
},
- "version": "e59d64a96a73e65ba53ba7fe05c9598974cc4a52"
+ "version": "3fe9a46d5fe0b70cbcabec1d2054f8ac3b3faae7"
},
{
"name": "grafana",
@@ -77,7 +77,7 @@
"subdir": "Documentation/etcd-mixin"
}
},
- "version": "3ef2ad8e115449a7004b628a873e2629855ed468"
+ "version": "5dc12f27251ad6f5f0744ad33ea7d731480f4b87"
},
{
"name": "prometheus",
@@ -87,7 +87,7 @@
"subdir": "documentation/prometheus-mixin"
}
},
- "version": "b05b5f9a300b0209689c06d70f676291f23774c4"
+ "version": "b5a16a8f861c29799f9a903f1e0859f513e862ed"
},
{
"name": "node-mixin",
@@ -107,7 +107,7 @@
"subdir": "lib/promgrafonnet"
}
},
- "version": "15ddfa20a6921ffbd43172eb54f6bdc1bcf8d3d3"
+ "version": "3ad401ea3ef7fb0879298fa411772984ffa7f31f"
}
]
}
diff --git a/kustomization.yaml b/kustomization.yaml
index a580ed8e..79f9624f 100644
--- a/kustomization.yaml
+++ b/kustomization.yaml
@@ -62,6 +62,7 @@ resources:
- ./manifests/prometheus-serviceAccount.yaml
- ./manifests/prometheus-serviceMonitor.yaml
- ./manifests/prometheus-serviceMonitorApiserver.yaml
+- ./manifests/prometheus-serviceMonitorAwsEksCNI.yaml
- ./manifests/prometheus-serviceMonitorCoreDNS.yaml
- ./manifests/prometheus-serviceMonitorKubeControllerManager.yaml
- ./manifests/prometheus-serviceMonitorKubeScheduler.yaml
diff --git a/manifests/grafana-dashboardDefinitions.yaml b/manifests/grafana-dashboardDefinitions.yaml
index 1065be04..9ab00dad 100644
--- a/manifests/grafana-dashboardDefinitions.yaml
+++ b/manifests/grafana-dashboardDefinitions.yaml
@@ -5522,7 +5522,7 @@ items:
],
"targets": [
{
- "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5540,7 +5540,7 @@ items:
"step": 10
},
{
- "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_requests_memory_bytes{node=\"$node\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_requests_memory_bytes{node=\"$node\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5558,7 +5558,7 @@ items:
"step": 10
},
{
- "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_limits_memory_bytes{node=\"$node\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_limits_memory_bytes{node=\"$node\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5567,7 +5567,7 @@ items:
"step": 10
},
{
- "expr": "sum(container_memory_rss{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_memory_rss{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5576,7 +5576,7 @@ items:
"step": 10
},
{
- "expr": "sum(container_memory_cache{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_memory_cache{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5585,7 +5585,7 @@ items:
"step": 10
},
{
- "expr": "sum(container_memory_swap{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
+ "expr": "sum(node_namespace_pod_container:container_memory_swap{cluster=\"$cluster\", node=\"$node\",container!=\"\"}) by (pod)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -8717,7 +8717,7 @@ items:
"tableColumn": "",
"targets": [
{
- "expr": "sum(up{job=\"kubelet\"})",
+ "expr": "sum(up{cluster=\"$cluster\", job=\"kubelet\"})",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "",
@@ -8801,7 +8801,7 @@ items:
"tableColumn": "",
"targets": [
{
- "expr": "sum(kubelet_running_pod_count{job=\"kubelet\", instance=~\"$instance\"})",
+ "expr": "sum(kubelet_running_pod_count{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"})",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -8885,7 +8885,7 @@ items:
"tableColumn": "",
"targets": [
{
- "expr": "sum(kubelet_running_container_count{job=\"kubelet\", instance=~\"$instance\"})",
+ "expr": "sum(kubelet_running_container_count{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"})",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -8969,7 +8969,7 @@ items:
"tableColumn": "",
"targets": [
{
- "expr": "sum(volume_manager_total_volumes{job=\"kubelet\", instance=~\"$instance\", state=\"actual_state_of_world\"})",
+ "expr": "sum(volume_manager_total_volumes{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\", state=\"actual_state_of_world\"})",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -9053,7 +9053,7 @@ items:
"tableColumn": "",
"targets": [
{
- "expr": "sum(volume_manager_total_volumes{job=\"kubelet\", instance=~\"$instance\",state=\"desired_state_of_world\"})",
+ "expr": "sum(volume_manager_total_volumes{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\",state=\"desired_state_of_world\"})",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -9137,7 +9137,7 @@ items:
"tableColumn": "",
"targets": [
{
- "expr": "sum(rate(kubelet_node_config_error{job=\"kubelet\", instance=~\"$instance\"}[5m]))",
+ "expr": "sum(rate(kubelet_node_config_error{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -9217,7 +9217,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(kubelet_runtime_operations_total{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (operation_type, instance)",
+ "expr": "sum(rate(kubelet_runtime_operations_total{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (operation_type, instance)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} {{operation_type}}",
@@ -9308,7 +9308,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(kubelet_runtime_operations_errors_total{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_type)",
+ "expr": "sum(rate(kubelet_runtime_operations_errors_total{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_type)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} {{operation_type}}",
@@ -9412,7 +9412,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(kubelet_runtime_operations_duration_seconds_bucket{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_type, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(kubelet_runtime_operations_duration_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_type, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} {{operation_type}}",
@@ -9516,14 +9516,14 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(kubelet_pod_start_duration_seconds_count{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance)",
+ "expr": "sum(rate(kubelet_pod_start_duration_seconds_count{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} pod",
"refId": "A"
},
{
- "expr": "sum(rate(kubelet_pod_worker_duration_seconds_count{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance)",
+ "expr": "sum(rate(kubelet_pod_worker_duration_seconds_count{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} worker",
@@ -9614,14 +9614,14 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(kubelet_pod_start_duration_seconds_count{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(kubelet_pod_start_duration_seconds_count{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} pod",
"refId": "A"
},
{
- "expr": "histogram_quantile(0.99, sum(rate(kubelet_pod_worker_duration_seconds_bucket{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(kubelet_pod_worker_duration_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} worker",
@@ -9727,7 +9727,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(storage_operation_duration_seconds_count{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_name, volume_plugin)",
+ "expr": "sum(rate(storage_operation_duration_seconds_count{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_name, volume_plugin)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} {{operation_name}} {{volume_plugin}}",
@@ -9820,7 +9820,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(storage_operation_errors_total{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_name, volume_plugin)",
+ "expr": "sum(rate(storage_operation_errors_total{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_name, volume_plugin)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} {{operation_name}} {{volume_plugin}}",
@@ -9926,7 +9926,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(storage_operation_duration_seconds_bucket{job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, operation_name, volume_plugin, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(storage_operation_duration_seconds_bucket{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, operation_name, volume_plugin, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} {{operation_name}} {{volume_plugin}}",
@@ -10030,7 +10030,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(kubelet_cgroup_manager_duration_seconds_count{job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, operation_type)",
+ "expr": "sum(rate(kubelet_cgroup_manager_duration_seconds_count{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, operation_type)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{operation_type}}",
@@ -10121,7 +10121,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(kubelet_cgroup_manager_duration_seconds_bucket{job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, operation_type, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(kubelet_cgroup_manager_duration_seconds_bucket{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, operation_type, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} {{operation_type}}",
@@ -10226,7 +10226,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(kubelet_pleg_relist_duration_seconds_count{job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance)",
+ "expr": "sum(rate(kubelet_pleg_relist_duration_seconds_count{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -10317,7 +10317,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(kubelet_pleg_relist_interval_seconds_bucket{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(kubelet_pleg_relist_interval_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -10421,7 +10421,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(kubelet_pleg_relist_duration_seconds_bucket{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(kubelet_pleg_relist_duration_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -10525,28 +10525,28 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(rest_client_requests_total{job=\"kubelet\", instance=~\"$instance\",code=~\"2..\"}[5m]))",
+ "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\",code=~\"2..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "2xx",
"refId": "A"
},
{
- "expr": "sum(rate(rest_client_requests_total{job=\"kubelet\", instance=~\"$instance\",code=~\"3..\"}[5m]))",
+ "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\",code=~\"3..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "3xx",
"refId": "B"
},
{
- "expr": "sum(rate(rest_client_requests_total{job=\"kubelet\", instance=~\"$instance\",code=~\"4..\"}[5m]))",
+ "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\",code=~\"4..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "4xx",
"refId": "C"
},
{
- "expr": "sum(rate(rest_client_requests_total{job=\"kubelet\", instance=~\"$instance\",code=~\"5..\"}[5m]))",
+ "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\",code=~\"5..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "5xx",
@@ -10650,7 +10650,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, verb, url, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, verb, url, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} {{verb}} {{url}}",
@@ -10754,7 +10754,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "process_resident_memory_bytes{job=\"kubelet\",instance=~\"$instance\"}",
+ "expr": "process_resident_memory_bytes{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -10845,7 +10845,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "rate(process_cpu_seconds_total{job=\"kubelet\",instance=~\"$instance\"}[5m])",
+ "expr": "rate(process_cpu_seconds_total{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -10936,7 +10936,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "go_goroutines{job=\"kubelet\",instance=~\"$instance\"}",
+ "expr": "go_goroutines{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -11020,6 +11020,32 @@ items:
"allValue": null,
"current": {
+ },
+ "datasource": "$datasource",
+ "hide": 2,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info, cluster)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+
},
"datasource": "$datasource",
"hide": 0,
@@ -11030,7 +11056,7 @@ items:
"options": [
],
- "query": "label_values(kubelet_runtime_operations{job=\"kubelet\"}, instance)",
+ "query": "label_values(kubelet_runtime_operations{cluster=\"$cluster\", job=\"kubelet\"}, instance)",
"refresh": 2,
"regex": "",
"sort": 0,
diff --git a/manifests/prometheus-rules.yaml b/manifests/prometheus-rules.yaml
index 007b3548..63a606c4 100644
--- a/manifests/prometheus-rules.yaml
+++ b/manifests/prometheus-rules.yaml
@@ -79,6 +79,22 @@ spec:
rate(container_cpu_usage_seconds_total{job="kubelet", image!="", container!="POD"}[5m])
) * on (namespace, pod) group_left(node) max by(namespace, pod, node) (kube_pod_info)
record: node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate
+ - expr: |
+ container_memory_working_set_bytes{job="kubelet", image!=""}
+ * on (namespace, pod) group_left(node) max by(namespace, pod, node) (kube_pod_info)
+ record: node_namespace_pod_container:container_memory_working_set_bytes
+ - expr: |
+ container_memory_rss{job="kubelet", image!=""}
+ * on (namespace, pod) group_left(node) max by(namespace, pod, node) (kube_pod_info)
+ record: node_namespace_pod_container:container_memory_rss
+ - expr: |
+ container_memory_cache{job="kubelet", image!=""}
+ * on (namespace, pod) group_left(node) max by(namespace, pod, node) (kube_pod_info)
+ record: node_namespace_pod_container:container_memory_cache
+ - expr: |
+ container_memory_swap{job="kubelet", image!=""}
+ * on (namespace, pod) group_left(node) max by(namespace, pod, node) (kube_pod_info)
+ record: node_namespace_pod_container:container_memory_swap
- expr: |
sum(container_memory_usage_bytes{job="kubelet", image!="", container!="POD"}) by (namespace)
record: namespace:container_memory_usage_bytes:sum
diff --git a/manifests/prometheus-serviceMonitorAwsEksCNI.yaml b/manifests/prometheus-serviceMonitorAwsEksCNI.yaml
new file mode 100644
index 00000000..07d14f1f
--- /dev/null
+++ b/manifests/prometheus-serviceMonitorAwsEksCNI.yaml
@@ -0,0 +1,19 @@
+apiVersion: monitoring.coreos.com/v1
+kind: ServiceMonitor
+metadata:
+ labels:
+ k8s-app: eks-cni
+ name: awsekscni
+ namespace: monitoring
+spec:
+ endpoints:
+ - interval: 30s
+ path: /metrics
+ port: cni-metrics-port
+ jobLabel: k8s-app
+ namespaceSelector:
+ matchNames:
+ - kube-system
+ selector:
+ matchLabels:
+ k8s-app: eks-cni
--
GitLab