Show more information in the UI about why Rollout gets aborted

[andrii-korotkov-verkada]

Summary

Extract information from Argo Rollout or the controller to give users more context why rollout was aborted.

Motivation

Right now if the failure happened on the step before analysis run is configured to happen, there might be no indication what went wrong, e.g. no bad events and generic message in status like

    - lastTransitionTime: '2024-11-19T21:48:19Z'
      lastUpdateTime: '2024-11-19T21:48:19Z'
      message: Rollout aborted update to revision 386
      reason: RolloutAborted
      status: 'False'
      type: Progressing

The info+ logs weren't helpful either.

Proposal

Add more information to the UI about why rollout was aborted, showing it in the Events and/or status.

[crenshaw-dev]

This might be a rollouts issue, if there's truly nothing else helpful in the status field.

[andrii-korotkov-verkada]

Let me open a Rollout issue as well then.

[andrii-korotkov-verkada]

argoproj/argo-rollouts#3957

[andrii-korotkov-verkada]

Looks like the healthcheck already has

for _, condition in ipairs(obj.status.conditions) do
  if condition.type == "InvalidSpec" then
    hs.status = "Degraded"
    hs.message = condition.message
    return hs
  end
  if condition.type == "Progressing" and condition.reason == "RolloutAborted" then
    hs.status = "Degraded"
    hs.message = condition.message
    return hs
  end
  if condition.type == "Progressing" and condition.reason == "ProgressDeadlineExceeded" then
    hs.status = "Degraded"
    hs.message = condition.message
    return hs
  end
end

So it seems to be more of a rollout not providing enough info issue.

[andrii-korotkov-verkada]

Tho I'm not sure why we check for type Progressing there.

[andrii-korotkov-verkada]

Looks like it's how it's done

    - lastTransitionTime: '2024-12-19T17:30:48Z'
      lastUpdateTime: '2024-12-19T17:30:48Z'
      message: >-
        Rollout aborted update to revision 18: Metric "apdex-rate-fill" assessed
        Failed due to failed (4) > failureLimit (3)
      reason: RolloutAborted
      status: 'False'
      type: Progressing

[andrii-korotkov-verkada]

Here's the relevant code in Rollouts:

	if isAborted {
		revision, _ := replicasetutil.Revision(c.rollout)
		message := fmt.Sprintf(conditions.RolloutAbortedMessage, revision)
		if c.pauseContext.abortMessage != "" {
			message = fmt.Sprintf("%s: %s", message, c.pauseContext.abortMessage)
		}
		condition := conditions.NewRolloutCondition(v1alpha1.RolloutProgressing, corev1.ConditionFalse, conditions.RolloutAbortedReason, message)
		if conditions.SetRolloutCondition(&newStatus, *condition) {
			c.recorder.Warnf(c.rollout, record.EventOptions{EventReason: conditions.RolloutAbortedReason}, message)
		}
	}

[andrii-korotkov-verkada]

So looks like abortMessage can be missing sometimes.