Duplicate container is left over when recreate fails

[lieryan]

This is related to #1045. After hitting the issue mentioned in #1045, subsequent docker-compose up fails with:

$ sudo docker-compose up -d
Recreating foo_bar_1...
Traceback (most recent call last):
...
requests.exceptions.ReadTimeout: UnixHTTPConnectionPool(host='localhost', port=None): Read timed out. (read timeout=60)
$ sudo DOCKER_CLIENT_TIMEOUT=300 docker-compose up
Recreating foo_bar_1...
Recreating 9c1b31a841_foo_bar_1...
Conflict. The name "foo_bar_1" is already in use by container 4af37fe03b76. You have to delete (or rename) that container to be able to reuse that name.

[szuliq]

I also see a lot of these.

I'm not sure if those are caused by #1045

Ubuntu 15.04; Docker 1.7

[aanand]

Looks related to #1349 - Compose has somehow gotten into a situation where there are two containers with the suffix _1, one of which has been renamed but then failed to recreate (perhaps because there was an HTTP timeout).

[dnephin]

I think after the failure, 9c1b31a841_foo_bar_1 still has the correct labels, so it's still considered an instance of the service. So there are two instances with the id of 1.

compose would need to only recreate once per instance id. In many ways this is related to #1516.

[aanand]

OK yeah, looks like this happens if we successfully create the new container but fail to start it. I can reproduce by sticking a sys.exit() in recreate_container:

new_container = self.create_container(...)
sys.exit(1)
self.start_container(new_container)

This results in a renamed old container, plus an unstarted new container, both with container-number=1.

In many ways this is related to #1516.

Random container numbers would solve the immediate error, yeah, but we'd still be left with two containers where we only wanted one.

[DavidBruant]

I also see a lot of these.

Ubuntu 15.04; Docker 1.7

Same combination, same problem. I also doubt it's #1045

I could get back to work by coming back to Docker 1.4.1 and docker-compose 1.2.0.

[lieryan]

This bug seems to show up if there is an exception (e.g. with #1045) or if you Ctrl+C in the middle of a docker-compose up. Essentially, docker-compose fails to clean up the temporary containers.

[iMerica]

Which version can I downgrade to so that I can get around this issue?

[aanand]

@iMerica Run docker-compose rm -f to clear everything out, then downgrade to 1.2.0.