Hi there !

We test monitoring our kubernetes cluster via zabbix with the "Kubernetes nodes by HTTP" template version 6.2.1

Description:

When a container is failing in a pod, kubernetes tries to restart it until it reaches the "CrashLoopBackoff" state. But while a container inside a pod is in this state, no alert is shown in zabbix.

Expected behaviour:

When any container is in "CrashLoopBackoff" state, a warning should be triggered by default.

Steps to reproduce:

1. Create a faulty container, let's call it Gorbatchev

apiVersion: v1
kind: Pod
metadata:
  name: gorbatchev
  namespace: test-zabbix
spec: 
  containers: 
    - image: "busybox"
      name: gorbatchev
      #This command will cause the container to fail
      args: ["perestroika"]
 

2. wait until it reaches the CrashLoopBackoff state

NAME             READY   STATUS             RESTARTS   AGE
pod/gorbatchev   0/1     CrashLoopBackOff   7          13m 

Result:
No alert is shown.

Expected:
A "trigger prototype" should be added to the template to alert when any pod has Conditions: Containers ready to false.

We were going to create the trigger on our zabbix instance but maybe this should be the default on the template ? Any pod in this state is an issue for cluster admins to investigate IMO.

Thanks in advance

–

Julien