Rollback
When a new version sends login error rate sharply up, roll back to the last known-good version; then confirm both running version and metrics recover.
When a new version sends login error rate sharply up, roll back to the last known-good version; then confirm both running version and metrics recover.
↳Rollback switches immediately to an existing older artifact; it does not wait for a forward fix.
Restore a known-good state to reduce impact: When a new version materially affects people, returning to a verified version or configuration can stop spread quickly. State the target clearly rather than saying only “restore it.”
A completed switch is not recovered health: First confirm the target version is actually running, then inspect login success, error rate, or key action recovery. If metrics do not recover, the problem may extend beyond this version.
If login error rate crosses its threshold, roll production back from v2 to the last known-good v1. Afterward, confirm both that v1 is actually running and that login error and success rates recovered; report both pieces of evidence.