Contiv v2 plugin fail to come up after docker service restart

[g1rana]

Issue : Contiv v2 plugin is getting failed to come up after Docker service restart

Root-cause: In current swarm installation, Contiv v2plugin and ETCD start as container Contiv-v2plugin(Start in context of docker daemon) has dependency on ETCD and ETCD container has dependency at docker daemon. Due to this cycle dependency , docker service restart couldn't enable the v2plugin.

Suggested Solution:
Approach 1 : Add dumy Systemd service which run after ETCD service and enable v2plugin which failed at first time during docker daemon spawn context. i.e " small bash script which run following command : <docker plugin enable <contiv-container ID> "

With this approach, changes require in swarm installer to add new service which run after ETCD service. Approach 2: Instead of running ETCD as container , just run it as host process and make docker service as depended on ETCD. With this approach , V2 plugin always find ETCD cluster and won't failed. This approach require changes in existing docker.service file and current ETCD installer script. Let me know which approach folks like to proceed further .

[unclejack]

@g1rana The second approach would be better. The first approach would make things difficult and we'd have to maintain that in the future.