Bluegreen rollout does not resume if autoPromotionEnabled: false is set with autoPromotionSeconds and previewReplicaCount

[thatsmydoing]

Summary

In a bluegreen rollout with autoPromotionEnabled: false and autoPromotionSeconds and previewReplicaCount also set, the rollout will stay paused indefinitely until it is manually promoted or any of previewReplicaCount or autoPromotionSeconds, autoPromotionEnabled is removed.

This might be a documentation issue as it states:

• autoPromotionEnabled defaults to true
• autoPromotionSeconds is ignored if autoPromotionEnabled is set to true

So it gives the impression that to use autoPromotionSeconds, one has to set autoPromotionEnabled: false, otherwise it would just promote immediately without waiting. In practice, autoPromotionEnabled must be true or unset for this to work. If not, it just never promotes.

Historically, I think pre-1.0.0, it did work with autoPromotionEnabled: false. I'm not sure if that was ever necessary though.

Diagnostics

argo-rollouts 1.0.7 and 1.2.0

These logs just keep on repeating

time="2022-04-26T16:29:48Z" level=info msg="Reconciling stable ReplicaSet 'realtime-7bb5bcd68f'" namespace=branch-master-1612593156 rollout=realtime
time="2022-04-26T16:29:48Z" level=info msg="reconciling pause (autoPromotionSeconds: 60)" namespace=branch-master-1612593156 rollout=realtime
time="2022-04-26T16:29:48Z" level=info msg="pause incomplete" namespace=branch-master-1612593156 rollout=realtime
time="2022-04-26T16:29:48Z" level=info msg="skipping active service switch: New RS 'realtime-856b7557fd' is not fully saturated" namespace=branch-master-1612593156 rollout=realtime
time="2022-04-26T16:29:48Z" level=info msg="Timed out (false) [last progress check: 2022-04-26 16:28:48 +0000 UTC - now: 2022-04-26 16:29:48.002012168 +0000 UTC m=+3407.751665669]" namespace=branch-master-1612593156 rollout=realtime
time="2022-04-26T16:29:48Z" level=info msg="No status changes. Skipping patch" generation=24 namespace=branch-master-1612593156 resourceVersion=360523416 rollout=realtime
time="2022-04-26T16:29:48Z" level=info msg="Reconciliation completed" generation=24 namespace=branch-master-1612593156 resourceVersion=360523416 rollout=realtime time_ms=1.3044790000000002

---

Message from the maintainers:

Impacted by this bug? Give it a 👍. We prioritize the issues with the most 👍.

[KipsasJaujoj]

Imho, it is a documentation issue. It behaves same whether previewReplicaCount is present or not.
With previewReplicaCount:

() kubectl get rollouts go-hello-world-app-sandbox1-common-rollout -o yaml | grep autoPromotion
      autoPromotionEnabled: false
      autoPromotionSeconds: 60

It’s paused indefinitely:

  - lastTransitionTime: "2023-01-19T16:12:10Z"
    lastUpdateTime: "2023-01-19T16:12:10Z"
    message: Rollout is paused
    reason: RolloutPaused
    status: "True"
    type: Paused
() TZ=UTC date
Thu Jan 19 16:17:50 UTC 2023

Without previewReplicaCount:

()   kubectl get rollouts go-hello-world-app-sandbox1-common-rollout -o yaml | grep autoPromotion
      autoPromotionEnabled: false
      autoPromotionSeconds: 60

Paused:

() kubectl get rollouts go-hello-world-app-sandbox1-common-rollout -o yaml
...
  - lastTransitionTime: "2023-01-19T16:44:32Z"
    lastUpdateTime: "2023-01-19T16:44:32Z"
    message: Rollout is paused
    reason: RolloutPaused
    status: "True"
    type: Paused
  controllerPause: true
() TZ=UTC date
Thu Jan 19 16:46:36 UTC 2023

and with autoPromotionEnabled: true:

() kubectl get rollouts go-hello-world-app-sandbox1-common-rollout -o yaml | grep autoPromotion
      autoPromotionEnabled: true
      autoPromotionSeconds: 60

And it’s already promoted after a minute:

() kubectl get rollouts go-hello-world-app-sandbox1-common-rollout -o yaml
...
  - lastTransitionTime: "2023-01-19T16:50:47Z"
    lastUpdateTime: "2023-01-19T16:50:47Z"
    message: Rollout is paused
    reason: RolloutPaused
    status: "False"
    type: Paused
  currentPodHash: cb87897b7
  observedGeneration: "80"
  phase: Healthy

[KipsasJaujoj]

It was a mistake in documentation. Sorted with #2525

[github-actions]

This issue is stale because it has been open 60 days with no activity.