Releases and rollback

Health checkCopy Anchor Link

After the service starts, Clusterz asks health_check on its port until it answers, within a time limit. A worker must keep running instead. If one service never answers, the new version of every service is taken down, the versions that were running stay, and the deploy is marked failed with the reason.

ReleaseCopy Anchor Link

A deploy builds, then runs release_command, once for each service that has one, then starts the new version, then moves traffic. If a release command fails, nothing moves.

RollbackCopy Anchor Link

Every deploy is kept with its image. Go back to the one before, or to a chosen one:

cluz deployments list
cluz rollback            # the deploy before the active one
cluz rollback 4 --wait   # deploy number 4

A rollback redeploys the earlier image. It does not undo a migration, so write migrations that the earlier version still runs on.