Symptom

READY=False, REASON=MigrationFailed. The Message names the migration and the failed action. The run halts at the anchoring stage; status.version is not advanced. The controller retries with backoff.

Cause

A migration anchored before a stage ran an action that failed (a job that errored, an http call that returned an error, a wait that timed out, an apply/patch/delete the apiserver rejected). This reason is distinct from StageFailed so you can tell a migration failure from the stage’s own apply.

Retries are safe for the actions that already completed: the per-action ledger (status.executedMigrationActions, keyed by migration content digest) records each finished action, so a retry resumes at the failed action rather than re-running destructive work from the top.

Remediation

• Read the Message and the migration’s Events (MigrationStarted / MigrationFailed) to find the failed action and its error.
• Fix the underlying cause (the target object, RBAC for the action, the endpoint an http action calls, the image a job runs).
• The migration re-attempts automatically on the next retry. After repeated failures it escalates to MigrationDirty , which halts auto-retry — see that page for the manual-recovery step.