Argo Rollouts

A dashboard of your Argo Rollouts, with phase, step, canary weight, revisions and AnalysisRuns, and Promote, Abort, Retry and Restart.

Rollouts in the sidebar’s Tools is a dashboard of the cluster’s Argo Rollouts. It appears only on a cluster that serves the rollouts.argoproj.io CRD. Kubebay reads the Rollout objects, their ReplicaSets and their AnalysisRun objects through the Kubernetes API with your kubeconfig. If you open the page by address on a cluster without Argo Rollouts, it says so.

What each Rollout shows

  • The name, which opens the Rollout’s detail, the namespace, a phase pill (Healthy, Progressing, Paused, Degraded or Aborted) and the strategy, canary or blueGreen.
  • For a canary, step N of M from status.currentStepIndex and the strategy’s steps, and the traffic split as N% canary and N% stable. The split is the controller’s own status.canary.weights when it reports one, otherwise the last setWeight step already completed.
  • desired / updated / ready / available replicas.
  • The stable revision and the canary revision: the ReplicaSet’s name, its rollout.argoproj.io/revision and its container images. There is no canary line when the new version is already the stable one.
  • When it is paused, Paused: and the reason or message, such as CanaryPauseStep.
  • Each AnalysisRun the Rollout owns, with its phase.

Actions

Each Rollout offers the actions that make sense for its state. They go through the same path as an edit in the YAML tab: a JSON merge patch with your own permissions, written to the audit log. On a cluster whose name looks like production, Kubebay asks first.

Action Shown Patch Confidence
Promote not aborted, not Healthy spec.paused: false and status.pauseConditions: null Follows what kubectl argo rollouts promote sends. Not tested against a live cluster.
Promote fully not aborted, not Healthy the same, plus status.promoteFull: true As above.
Abort not aborted, not Healthy status.abort: true As above.
Retry aborted status.abort: false As above.
Restart always spec.restartAt set to now Documented Rollout field. Not tested against a live cluster.

Not offered: skip the current step (status.currentStepIndex) and Pause, because the patch Kubebay would send for them wasn’t verified. Use kubectl argo rollouts for those.

The status fields are written on the Rollout itself, as the kubectl plugin does. If your Rollout CRD has the status subresource turned on, the API server ignores them and the action does nothing.

Not here yet

Argo Workflows and Argo Events have no pages of their own. Their objects are listed under Custom Resources in the sidebar.