A livelinessProbe is important, especially in HA. We want to make sure the Job Manager was elected as a leader. Sometimes I observed it getting stuck if the Job manager start before the Zookeeper cluster is up and running. It hangs waiting for Zookeeper to elect a leader and never completes.
This livenessProbe has fixed that issue.
If the job manager cannot see a leader, it will return an error, which will get captured by the livenessProbe. Kubernetes will eventually restart the container
livenessProbe: httpGet: path: /overview port: 8081 initialDelaySeconds: 30 periodSeconds: 10
A livelinessProbe is important, especially in HA. We want to make sure the Job Manager was elected as a leader. Sometimes I observed it getting stuck if the Job manager start before the Zookeeper cluster is up and running. It hangs waiting for Zookeeper to elect a leader and never completes. This livenessProbe has fixed that issue.
If the job manager cannot see a leader, it will return an error, which will get captured by the livenessProbe. Kubernetes will eventually restart the container