Automatic cleanup of old docker images would be nice

[bickelj]

In the test system over the weekend, the continuous integration script notified us of a couple failures. It turns out the disk was probably temporarily full. It is easy enough to run docker system prune --all when verifying that containers are up, but a more regular, targeted approach would be better because it is best to have the previous image present in case we need to roll back.

[bickelj]

A safer command that can leave at least one image assuming we built an image in the past 336 hours:
docker system prune -a --filter "until=336h"

Perhaps even safer:
[[ "$(docker ps | wc -l)" -eq "4" ]] && docker system prune -a --filter "until=336h"