I'm running the kubelet 1.5.4 and finding that even when I run low on disk space it will not delete old containers.
I have may exited containers, over an hour hold:
$ sudo docker ps -a | grep 'Exited (0) About an hour' | wc -l 78
But the kubelet does not seem to notice them (using a log level of 10 I see no references in kubelet logs to the container IDs).
Is this something that has changed recently? My previous 1.2 setup seemed to GC old containers.
Kubelet has a feature to gc images but not sure what you mean by gc'ing containers?
See the proposal spec here (it does seem as if it's due to be deprecated as well)