Stopping containers with docker.keepContainer false fails with docker error

[fabric8io]

Stopping containers with docker.keepContainer false fails with docker error

[rhuss]

That's interesting. In case of the error, if you retry to delete the container sometimes later, does it the work ?

I ask because I could imagine to simply retry the removal if the first try fails (instead of inserting a pause from the very beginning since that problem never happened to me until now).

Another question: Did you really slept 20 minutes (1000 seconds) or this a typo in your command above (and you meant 1000 ms) ?

[chris-snider]

It should have been a sleep 1, not the 1000 for command line. I used 1000 (ms) for an update to the AbstractDockerMojo.java.

Attempting to remove the container after getting the error results in another error indicating the container was not found. Restarting the docker engine showed the same container ID, but a different name associated. Executing docker rm ImgID succeeds and removes the container(s).

I added a Thread.sleep(1000) just before the access.removeContainer in the AbstractDockerMojo as

            if (!keepContainer) {
                try {
                    Thread.sleep(1000); //need to give the docker engine the split second to finish
                } catch (InterruptedException e) {
                    //just eat it, eat it...
                }
                // Remove the container
                access.removeContainer(container);
            }

And this successfully completes my integration tests from starting all my containers to stopping and removing them all.

[rhuss]

Ok, this seems that docker has indeed multiple issues when removing images (search for docker run rm busy). So its probably a good idea to add this pause as a workaround (wonder if it could be 500ms as well ;-)

'will add this for the next release, maybe even configurable (in the <wait> section).

[chris-snider]

I believe 500 ms will work. It seems that it just has to have a pause of some kind to complete the previous command.

[rhuss]

Just pust a new snapshot 0.10.6-SNAPSHOT which allow the specificaton on a grace period after stopping a container. See #70 for details how to configure this.

Does this fix your issue ?

[chris-snider]

Is the change adding an additional tag to the already existing wait? Such as

<wait>
    <log>Server started in</log>
    <!--... but a max of 20 seconds-->
    <time>20000</time>
    <!-- Wait 500 ms between stopping and removing the container -->
    <shutdown>500</shutdown>
</wait>

[rhuss]

Yes, that's it.

[chris-snider]

I can confirm that the new shutdown works. I was able to launch, test, shutdown and remove my containers.