Some job metrics are missing if job has no conditions

[DerRockWolf]

What happened:
Some job metrics (kube_job_status_failed, kube_job_complete, kube_job_failed) are missing for Jobs without conditions.

What you expected to happen:
Metrics are present regardless if conditions exist.

How to reproduce it (as minimally and precisely as possible):

1. Create this job

job.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: test
spec:
  template:
    spec:
      containers:
      - image: alpine
        name: test
        command:
          - "sh"
          - "-c"
        args:
          - sleep 5 && exit 1
      restartPolicy: Never

2. Observe that the .status is missing a conditions objects (before the backoffLimit is reached)
3. Observe that e.g., the kube_job_status_failed metric is missing (curl localhost:8080/metrics | grep kube_job_status_failed)

Anything else we need to know?:

The reason for this bug is that the labels and values are only set within a for loop (if a condition of type Failed exists):

kube-state-metrics/internal/store/job.go

Lines 222 to 224 in 85d1423

	for _, c := range j.Status.Conditions {
	condition := c
	if condition.Type == v1batch.JobFailed {

I can provide a fix that sets the value regardless if there are any condition with type Failed.

Environment:

• kube-state-metrics version: v2.12.0 (master 85d1423)
• Kubernetes version (use kubectl version): v1.29.6
• Cloud provider or hardware configuration: homelab

[dgrisonnet]

/assign @richabanker
/triage accepted

[richabanker]

I just tried to reproduce this, so far I am actually able to see the kube_job_status_failed metric being reported

# HELP kube_job_status_failed The number of pods which reached Phase Failed.
# TYPE kube_job_status_failed gauge
kube_job_status_failed{namespace="default",job_name="test"} 0

but yes, I dont see the kube_job_complete,kube_job_failed metrics, which I believe is WAI since these metrics should only be reported when the job's Status.Condition.Type changes to Complete or Failed

[DerRockWolf]

kube_job_status_failed was there for roughly 30 seconds and then disappeared.

From my perspective the kube_job_status_succeeded, kube_job_status_failed & kube_job_status_active metrics should all be emitted regardless of the overall job conditions. kube_job_status_failed is currently the only one depending on the condition being present.

I listed kube_job_complete & kube_job_failed because they also depend on the existence of an condition, but I agree that these probably work as designed.

I also found that kube_job_status_failed isn't correctly implemented.
The description states: "The number of pods which reached Phase Failed and the reason for failure", but currently the number of failed pods is only present if reasonKnown is false...

[richabanker]

Discussed briefly with @dgrisonnet, the suggestion to emit kube_job_status_failed even when there are no job conditions seems worthwhile. @DerRockWolf would you be open to create a PR for that? If not, then I can get started on one, please let us know what you prefer. Thanks!