docker cleanup issues about buildfarm_deployment HOT 4 CLOSED

Non graceful termination:

2015-02-21 02:15:01,616 Disk space satified ending
2015-02-21 02:15:01,617 Uncaught exception in cleanup_docker_image.py: Traceback (most recent call last):
  File "/home/jenkins-slave/cleanup_docker_images.py", line 173, in <module>
    main()
  File "/home/jenkins-slave/cleanup_docker_images.py", line 153, in main
    sys.exit(0)
SystemExit: 0

from buildfarm_deployment.

Clean termination: 13d1e9d
Better usage of disk space on larger disks: 04e69e5

I believe the old images had retained cron jobs. I've rolled over all the old executors.

from buildfarm_deployment.

The base images will be cleaned up only if there are no containers using them. And if we go the full timeout and nothing is using them It's probably ok to clean them. They are only ~ 100 Mb to redownload and explicitly listing anything is just another thing to maintiain adding complexity and overhead. I think whitelisting the base images or something like that would be an overoptimization. In practice we will likely never be that idle and they'll be protected by the active containers.

from buildfarm_deployment.

Filed slow rmi upstream: moby/moby#11053

from buildfarm_deployment.