Troubleshooting

• ActionDirty

  A stage with an incomplete once-per-lifetime action failed repeatedly, so the controller halted auto-retry; clear the halt with a manual reconcile once the cause is fixed.

• ArtifactNotFound

  The referenced ExternalArtifact could not be found (transient; the controller requeues).

• AwaitingPromotion

  A stage is applied and healthy and the rollout is holding for a manual promotion.

• BudgetExhausted

  A rollout is frozen because the service is out of its SLO error budget.

• BudgetSourceUnavailable

  A metric source for an error-budget freeze or a promotion analysis could not be read.

• Controller pod down

  A stageset-controller pod has been NotReady for the alert window.

• DependencyNotReady

  A StageSet named in spec.dependsOn is not yet Ready.

• DowngradeNotAllowed

  The desired version is below the deployed version and downgrades are not enabled.

• DowngradeRequiresMigration

  A downgrade is enabled, but a version boundary it crosses declares no down actions and is irreversible.

• ImageUnverified

  A stage referenced an image that fails an ImageVerificationPolicy; the stage is held before apply.

• InvalidSpec

  The StageSet spec is invalid; the Message names the offending field or action.

• InvalidVersion

  A version source or value could not be parsed as semver.

• MigrationArtifactInvalid

  A migration ladder sourced via spec.migrationsSourceRef could not be parsed or failed content validation; the version is held back until the artifact is fixed.

• MigrationCoverageMissing

  A major-version transition has no migration covering it, and requireMigrationCoverage is set; the version is held back.

• MigrationDirty

  A migration failed repeatedly, so the controller halted auto-retry; clear the halt with a manual reconcile once the cause is fixed.

• MigrationFailed

  A migration's action failed; the run halts at that stage and retries with backoff, skipping already-completed actions.

• MigrationSourceNotPinned

  A sourced migration ladder's source is pinned to a mutable tag/branch instead of an immutable digest/commit, and pinning is required.

• MigrationSourceNotVerified

  A sourced migration ladder's source is not signature-verified; the destructive ladder is refused until verification passes.

• MigrationStageNotFound

  A version-selected migration anchors to a stage that doesn't exist in this StageSet; the version is held back rather than skipping the migration.

• PreviousRevisionUnavailable

  rollbackOnFailure is set but the last-good revisions could not be restored.

• Producer watch engagement failing

  A dynamic watch on a producer source kind failed to engage, so StageSets referencing that kind no longer re-trigger on the producer's upstream changes.

• PromotionBlocked

  A stage's promotion analysis breached its thresholds, so the rollout will not advance past it.

• RBACDenied

  An apiserver call the controller made — resolving a source, an impersonated apply, or a tenant get — failed with Forbidden or referenced a kind the apiserver does not know.

• Reconcile latency high

  Reconcile p99 latency for the StageSet controller is above threshold.

• ResolveFailed

  A source reference could not be resolved to a ready ExternalArtifact.

• Runbooks

  Symptom-to-remediation pages for every StageSet status reason and operational alert, grouped by what each means for the rollout.

• Soaking

  A stage is applied and healthy and the rollout is holding through its promotion soak window before advancing.

• SourceNotReady

  The source exists but has not published a ready artifact yet (transient).

• StageFailed

  A stage failed to fetch, build, apply, verify, or run an action; the run halts there.

• StageInventory

  The controller-managed inventory of objects each stage applied.

• StageProgressing

  A stage's verify wait ran out of time while its objects were still coming up, so the run is holding rather than failing — it converges on its own.

• Stalled

  The run cannot make progress and will not retry until the spec changes.

• Suspended

  Reconciliation is paused via spec.suspend.

• TeardownForced

  A deleting StageSet's finalizer was force-dropped because teardown could not complete, possibly orphaning objects on the target cluster.

• UpdateDeferred

  A new revision is held by a closed update window.

• Webhook cert renewal failing

  The self-signed admission webhook certificate is not being rotated.

• Workqueue saturation

  The controller cannot drain its reconcile queue fast enough.