Docker run improvements

Background

docker run is meant to run docker containers as if they were UNIX processes. docker run doesn't behave just like UNIX processes because it doesn't return the exit code of the process running in the container and it doesn't have support for temporary containers.

A proof of concept implementation which wraps docker run can be found at Dockrun at github.com.

Purpose

The goal of these changes is to make docker run easy to monitor with systemd, upstart, supervisor and other process monitoring solutions and systems. These changes would make docker run standalone redundant. Standalone mode wouldn't be needed any more because monitoring the docker daemon is enough and monitoring is going to be used for docker run anyway, one more process to monitor won't cause any problems.

Proposed changes

The proposed changes are separated to maximize their usefulness to make them useful in more scenarios. docker run needs the following features in order to have a more UNIX process like behavior:

"temporary" flag

This flag would make docker run delete the container when the process exits. This flag could be -r because it's not currently used by docker run.

"exit code" flag

This flag would make docker run return the exit code of the process running in the container, instead of returning 0 when the container process exits. The flag used for this could be -s. Note: docker run will have to return the exit code of the process running in the container even if a signal to terminate is received by docker run.