Other swarm members can't mount a volume that was attached to a terminated instance

[zbyte64]

• Setup an AWS docker swarm with convoy installed on each node
• Create an EBS volume through docker py:

swarm_client.create_volume(
        name='someuuid',
        driver='convoy',
        driver_opts={
            'size': '20G',
        }
    )

• Create a container that mounts the EBS volume
• Terminate container
• Build new container with the same instructions and join the docker swarm
• Create the same container and it creates a new volume but it says internally it can't find the volume (from the logs):

{"level":"debug","msg":"Handle plugin get volume: POST /VolumeDriver.Get","pkg":"daemon","time":"2016-07-03T06:57:47Z"}
{"level":"debug","msg":"Request from docker: \u0026{b-rethinkdb-montage-c18f2db3_c6d2_4f2f_a841_6d6bbcba8183 map[]}","pkg":"daemon","time":"2016-07-03T06:57:47Z"}
{"level":"debug","msg":"Response:  {\n\t\"Err\": \"Could not find volume b-rethinkdb-montage-c18f2db3_c6d2_4f2f_a841_6d6bbcba8183.\"\n}","pkg":"daemon","time":"2016-07-03T06:57:47Z"}

• docker inspect <the_container_id> and notice that volumes changed to local:

"Mounts": [
            {
                "Name": "b-rethinkdb-montage-5d1d7480_67cb_4644_a3be_380cfa8fc3bc",
                "Source": "/var/lib/docker/volumes/b-rethinkdb-montage-5d1d7480_67cb_4644_a3be_380cfa8fc3bc/_data",
                "Destination": "/data",
                "Driver": "local",
                "Mode": "rw",
                "RW": true,
                "Propagation": "rprivate"
            }
        ],

• EBS volume exists and is detached
• create another volume and container and see the docker inspect lists the driver as convoy

This might be a duplicate of #125

Edit:

• Convoy 5.0-rc1
• Docker 1.11.2
• Swarm 1.2.3

[yasker]

@zbyte64 looks like it's in a different host and EBS volume is not recognized. Convoy is not a distributed system so the Convoy on the other host doesn't know about the volume at all.

You can create the volume on the host beforehand using --id parameter for convoy create, but I think it's may not be the optimal solution to you. I need to look into this and see if we can find a better way to do this.

[zbyte64]

Having to create the volume again would not work well with swarm rescheduler: https://docs.docker.com/swarm/scheduler/rescheduling/