Skip to content

Deployment Convergence

During a deployment, BOSH tries to converge to the intended state, i.e. the state described in the deployment manifest, from the current state.

When a update is triggered using bosh cck or bosh<start|stop|restart|recreate>, they can lead to different intended states.

bosh cck

bosh cck will always recreate the instance with the current deployed instance state. This means that the desired instance plan is based off of the information that is encoded in the instances' current deployment spec that is present in the database.

bosh <start|stop|restart|recreate>

bosh <start|stop|restart|recreate> will always converge the instance to the last successfully deployed desired state. This means that the desired instance plan is based off of the information that is persisted in the last successfully deployed manifest. These commands will also detect any changes on other instances that conflict with the last deployed successful state, and attempt to converge the deployment to the desired state.

--no-converge flag

As of BOSH v270.4.0 and BOSH CLI v6.0.0, the start, stop, restart, and recreate commands all support a --no-converge flag. When this flag is specified, the corresponding command will act ONLY on the specified instance. The instance will based off of the current deployed instance state that is recorded in the database.