If a Bamboo remote agent fails in the midst of a Job that had been using a docker environment then we will see the docker process spawned by Bamboo remaining across restarts of the agent.

If docker is used by Bamboo in a Job then Bamboo will spawn a docker container running
"tail -f /dev/null". Thus holding the container active to be used by the Job.
This docker instance would normally be cleaned up and removed during the Docker cleanup task in the job.

If the remote agent aborts whilst before the Docker cleanup task is executed these processes persist.

$ docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
5c1ab003564f xxxxx:1.0 "tail -f /dev/null" 4 days ago Up 4 days c8c1332b-2a06-4cc1-9a70-68d79c8202b881887236
1a144aded739 xxxxx:1.0 "tail -f /dev/null" 4 days ago Up 4 days 586e5563-ffa8-421c-b323-6bec9eccfaeb81887236

This enhancement request is for the Bamboo agent to maintain some persistent knowledge about the processes, docker instances, that it has spawned and clean these up when it restarts, as necessary.