rename KubeNodeEviction, fix test case

Signed-off-by: TheRealNoob <[email protected]>

Author: TheRealNoob

alerts/kubelet.libsonnet

@@ -14,7 +14,7 @@ local utils = import '../lib/utils.libsonnet';
     kubeletCertExpirationCriticalSeconds: 1 * 24 * 3600,
 
     // Evictions per second that will trigger an alert. The default value will trigger on any evictions.
-    evictionRateThreshold: 0.0,
+    KubeNodeEvictionRateThreshold: 0.0,
   },
 
   prometheusAlerts+:: {
@@ -125,9 +125,13 @@ local utils = import '../lib/utils.libsonnet';
             },
           },
           {
-            alert: 'KubeNodeEvictions',
+            alert: 'KubeNodeEviction',
             expr: |||
-              sum(rate(kubelet_evictions{%(kubeletSelector)s}[15m])) by(%(clusterLabel)s, eviction_signal, node) > %(evictionRateThreshold)s
+              sum(rate(kubelet_evictions{%(kubeletSelector)s}[15m])) by(%(clusterLabel)s, eviction_signal, instance) > %(KubeNodeEvictionRateThreshold)s
+              * on (%(clusterLabel)s, instance) group_left(node)
+              max by (%(clusterLabel)s, instance, node) (
+                kubelet_node_name{%(kubeletSelector)s}
+              )
             ||| % $._config,
             labels: {
               severity: 'info',

runbook.md

@@ -191,8 +191,6 @@ This page collects this repositories alerts and begins the process of describing
 + *Severity*: info
 + *Runbook*: [Link](https://runbooks.prometheus-operator.dev/runbooks/kubernetes/kubenodepressure/)
 
-# If soft thresholds are crossed, pods will be evicted respecting TerminationGracePeriod. If Hard thresholds are crossed grace period will be ignored.
-
 ##### Alert Name: "KubeNodeUnreachable"
 + *Message*: `{{ $labels.node }} is unreachable and some workloads may be rescheduled.`
 + *Severity*: warning
@@ -208,10 +206,10 @@ This page collects this repositories alerts and begins the process of describing
 + *Severity*: warning
 + *Runbook*: [Link](https://runbooks.prometheus-operator.dev/runbooks/kubernetes/kubenodereadinessflapping/)
 
-##### Alert Name: "KubeNodeEvictions"
+##### Alert Name: "KubeNodeEviction"
 + *Message*: `Node {{ $labels.node }} is evicting Pods due to {{ $labels.eviction_signal }}. Eviction occurs when eviction thresholds are crossed, typically caused by Pods exceeding RAM/ephemeral-storage limits.`
 + *Severity*: info
-+ *Runbook*: [Link](https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubenodeevictions)
++ *Runbook*: [Link](https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubenodeeviction)
 
 ##### Alert Name: "KubeletPlegDurationHigh"
 + *Message*: `The Kubelet Pod Lifecycle Event Generator has a 99th percentile duration of {{ $value }} seconds on node {{ $labels.node }}.`

tests/tests.yaml

@@ -696,23 +696,17 @@ tests:
     values: '1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 3 3 3 3 3'
   alert_rule_test:
   - eval_time: 18m
-    alertname: KubeNodeEvictions
+    alertname: KubeNodeEviction
     exp_alerts:
     - exp_labels:
         eviction_signal: memory.available
         cluster: kubernetes
         node: minikube
         severity: info
-        endpoint: https-main
-        instance: 10.0.2.15:10250
-        job: kubelet
-        namespace: monitoring
-        pod: kube-state-metrics-b894d84cc-d6htw
-        service: kube-state-metrics
       exp_annotations:
         summary: "Node is evicting pods."
         description: 'Node minikube is evicting Pods due to memory.available.  Eviction occurs when eviction thresholds are crossed, typically caused by Pods exceeding RAM/ephemeral-storage limits.'
-        runbook_url: 'https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubenodeevictions'
+        runbook_url: 'https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubenodeeviction'
 
 # Verify that node:node_num_cpu:sum triggers no many-to-many errors.
 - interval: 1m