[SPARK-54328] Add configurable startupProbe and enhance liveness/readiness probes in Helm chart

[jiangzho]

What changes were proposed in this pull request?

This PR adds a configurable startupProbe and enhances the existing livenessProbe and readinessProbe with additional configurable parameters in the Helm chart for the spark-kubernetes-operator.

Why are the changes needed?

Previously, these values were either using Kubernetes defaults or not configured at all. This change makes them explicitly configurable via Helm values, giving operators more control over pod lifecycle management in different cluster environments (small clusters vs large production clusters).

Does this PR introduce any user-facing change?

Yes. Users can now configure these probe settings in their values.yaml.

How was this patch tested?

E2E coverage for default value, and local dry-run for value overrides.

Was this patch authored or co-authored using generative AI tooling?

No.

[jiangzho]

cc @peter-toth for review - thanks a lot in advance!

[peter-toth]

Thank you @jiangzho , it is always better to have more control over these values.

Could you please check the PR description though? The link seems to be broken.

[jiangzho]

Thanks for the catch! Seems GitHub interprates that as a JIRA automatically. Updated the description.

[peter-toth]

Thank you @jiangzho.

Merged to main (0.7.0).

[dongjoon-hyun]

Hi, @jiangzho and @peter-toth .

Unfortunately, this seems to introduce a severe race condition which causes the Apache Spark restart frequently. Technically, WebSocket is supposed to restart regularly and the reconnect time is scheduled after 1000ms. HealthProbe seems to check this time and kills the operator because this PR makes failureThreshold=1 which means no failure is allowed.. Here is the full log.

26/02/08 01:59:42 DEBUG   o.a.s.k.o.p.HealthProbe Checking informer health
26/02/08 01:59:42 DEBUG   i.j.o.p.e.s.i.InformerWrapper Informer status: HEALTHY for type: SparkApplication, namespace: default, details [is running: true, has synced: true, is watching: true]
26/02/08 01:59:42 DEBUG   i.j.o.p.e.s.i.InformerWrapper Informer status: HEALTHY for type: Pod, namespace: default, details [is running: true, has synced: true, is watching: true]
26/02/08 01:59:42 DEBUG   i.j.o.p.e.s.i.InformerWrapper Informer status: HEALTHY for type: SparkCluster, namespace: default, details [is running: true, has synced: true, is watching: true]
26/02/08 01:59:42 DEBUG   i.j.o.p.e.s.i.InformerWrapper Informer status: HEALTHY for type: Pod, namespace: default, details [is running: true, has synced: true, is watching: true]
26/02/08 01:59:51 DEBUG   i.f.k.c.d.i.AbstractWatchManager Closing the current watch
26/02/08 01:59:51 DEBUG   i.f.k.c.d.i.WatcherWebSocketListener WebSocket close received. code: 1000, reason: null
26/02/08 01:59:51 DEBUG   i.f.k.c.d.i.AbstractWatchManager Scheduling reconnect task in 1000 ms
26/02/08 01:59:52 DEBUG   o.a.s.k.o.p.HealthProbe Checking informer health
26/02/08 01:59:52 DEBUG   i.j.o.p.e.s.i.InformerWrapper Informer status: HEALTHY for type: SparkApplication, namespace: default, details [is running: true, has synced: true, is watching: true]
26/02/08 01:59:52 DEBUG   i.j.o.p.e.s.i.InformerWrapper Informer status: HEALTHY for type: Pod, namespace: default, details [is running: true, has synced: true, is watching: true]
26/02/08 01:59:52 DEBUG   i.j.o.p.e.s.i.InformerWrapper Informer status: UNHEALTHY for type: SparkCluster, namespace: default, details [is running: true, has synced: true, is watching: false]
26/02/08 01:59:52 DEBUG   i.j.o.p.e.s.i.InformerWrapper Informer status: HEALTHY for type: Pod, namespace: default, details [is running: true, has synced: true, is watching: true]
26/02/08 01:59:52 DEBUG   i.j.o.p.e.s.i.InformerWrapper Informer status: UNHEALTHY for type: SparkCluster, namespace: default, details [is running: true, has synced: true, is watching: false]
26/02/08 01:59:52 DEBUG   i.j.o.p.e.s.i.InformerWrapper Informer status: UNHEALTHY for type: SparkCluster, namespace: default, details [is running: true, has synced: true, is watching: false]
26/02/08 01:59:52 ERROR   o.a.s.k.o.p.HealthProbe Controller: sparkclusterreconciler, Event Source: ControllerResourceEventSource, Informer: UNHEALTHY is in default, not a healthy state

[dongjoon-hyun]

Also, this PR introduces an inconsistency for readinessProbe.failureThreahold. Here, this PR uses 1 while the default is introduced by 30.

{{- default 30 .Values.operatorDeployment.operatorPod.operatorContainer.probes.readinessProbe.failureThreshold }}

[dongjoon-hyun]

I made a PR to fix this stability issue.

• [SPARK-55421] Increase livenessProbe.failureThreshold to 3 #491

[dongjoon-hyun]

I also made a PR for #417 (comment)

• [SPARK-55422] Fix the default value of readinessProbe.failureThreshold to 1 #492