One last thing that must be mentioned in this section as well is the fact that most of plugins that handle the moving of volumes can only handle being attached to a single node at any one time due to the volume being writable by multiple sources is going to generally cause serious issues so most drivers disallow it.

This however is in conflict with the main feature of most orchestration engines which, on changes to Docker services, will leave the original service running until the new one is started and passes health checks, causing the need to mount the same volume on both the old and new service task in effect, creating a chicken-egg paradox.

In most cases, this can be worked around by making sure that Docker completely kills the old service before starting the new one, but even then, you can expect that occasionally the old volume will not be unmounted quickly enough from the old node, so the new service will fail to start.