The grid manager checks its persistent storage to ensure that the service is processing the running objects.

The grid manager then starts undeploying the object from the service. The grid manager sends a stop message to all the controller instances where the objects are running. The objects stops consuming events from the queue, clears the backlog, and releases all the resources that it holds.

After all the objects respond successfully to the stop message, the grid manager undeploys the object from the service instances.

The node removes the transport queue if the object is the last one that is producing events to it.

The grid manager then repeats the same steps to undeploy the processes in other nodes.

Successful undeployment removes the objects from the persistent repository and reports a success.

If there are any errors while undeploying an object, the grid manager retries the operation for the configured number of times before starting the rollback process.

As part of the rollback, the grid manager redeploys the undeployed objects with the objects stored in its persistent storage.

After redeployment, the grid manager returns an error back to the user interface to notify about the failure.