Nomad should wait with cleaning docker images for faster startup times

[bsphere]

Nomad version

0.4

Operating system and Environment details

CoreOS stable

Issue

Nomad removes docker images when tasks are stopped, but in case the task is restarted (e.g. consul-template update), startup times are wait too long because it pulls the images again.

Maybe Nomad can wait for some time before cleaning images to improve startup times.

[camerondavison]

this could also help with some of the docker race conditions that make cleaning up the images break docker due to docker bugs.

[diptanu]

We will most likely solve this when we introduce artifact cache in the Nomad client. Docker containers or any other binaries would be downloaded and removed if available disk space is low and not in use.

[dbresson]

This enhancement will break the ability to update to a new revision of the same image tag.

[camerondavison]

@dbresson see #639 I think you are mentioning that you would like to be able to restart a service and get a image even though the tag is the same, this is currently not supported, and you should not rely on this behavior. With #639 done then whatever flag is used to notify the client that it should always pull the latest image from the registry would make sure that this issue does not break the behavior you are talking about.

[dadgar]

Hey this was fixed in 0.5.5 and made more robust in 0.5.6

[github-actions]

I'm going to lock this issue because it has been closed for 120 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.