nextcloud
cron.sh doesn't stop without SIGKILL
The cron.sh script doesn't stop with the default SIGWINCH signal (or SIGTERM), which causes docker stop to wait for 10 seconds and fail over to SIGKILL. To reproduce:
docker run --name nextcloud-cron --rm --entrypoint /cron.sh docker.io/library/nextcloud
docker stop nextcloud-cron # Takes 10 seconds to complete
Docker hides this by silently falling back to SIGKILL, but it makes stopping take an unnecessary 10 seconds. If using podman, podman stop nextcloud-cron returns
WARN[0010] StopSignal SIGWINCH failed to stop container nextcloud-cron in 10 seconds, resorting to SIGKILL
This may be the result of upstream base image handling (the php image) for the Dockerfile STOPSIGNAL. I believe the default was changed there to SIGWINCH a long time ago because Apache likes that for gracefully shutting down.
On the other hand, it looks like busybox crond prefers SIGTERM to do a graceful shutdown.
Not at my desk at the moment, but should be easy to test with either...
For an already running cron container:
docker kill -s TERM (possible "SIGTERM" not sure offhand)
Or by overriding the default STOPSIGNAL for when running the cron container:
docker run -stop-signal [...]
Since keeping it for Apache is desirable and we don't use separate Dockerfiles for Cron, not sure how to best handle this on a permanent basis at the moment... 🤔
