Gracefully stop container before shutdown/reboot

[Krumbelfix]

Is your feature request related to a problem? Please describe.
While rebooting our Servers with rke2 we had several times the issue that the containers did not stop properly and therefore the filesystem could not be unmounted. The server hung for several minutes before rebooting.

Describe the solution you'd like
we implemented the attached systemd unit
systemd-unit.txt

Describe alternatives you've considered
A nice fix to speedup the server reboot provided by rke2 :)

[Martin-Weiss]

+1

[brandond]

Linking this in to graceful node shutdown as discussed at k3s-io/k3s#4319 would probably be the proper way to go about this.

[brandond]

I am not confident that this is something that RKE2 should provide for itself. Both K3s and RKE2 do not - by design - stop workload pods when shutting down. There is some discussion in the linked K3s issue about possible approaches that Harvester (or other users of K3s/RKE2) could take to ensuring that workloads are explicitly terminated as part of systemd shutdown. It could even just be something as simple as installing a service that runs the rke2-killall.sh script on shutdown:

# /etc/systemd/system/rke2-cleanup.service
[Unit]
Description=rke2-cleanup
DefaultDependencies=no
Before=shutdown.target

[Service]
Type=oneshot
ExecStart=/usr/local/bin/rke2-killall.sh
TimeoutStartSec=0

[Install]
WantedBy=shutdown.target

Edit: I see that this is the approach that has already been taken by the attached systemd unit. Lets have a discussion about whether or not this is something that should be bundled with RKE2, or if we want to just add it as an example in the documentation.

[Krumbelfix]

In my opinion it should be bundled, because it improves the rke experience. It does not have any disadvantages in the daily business and is helps keeping it easy.

[stale]

This repository uses a bot to automatically label issues which have not had any activity (commit/comment/label) for 180 days. This helps us manage the community issues better. If the issue is still relevant, please add a comment to the issue so the bot can remove the label and we know it is still valid. If it is no longer relevant (or possibly fixed in the latest release), the bot will automatically close the issue in 14 days. Thank you for your contributions.

[alancaldelas]

Was this actually resolved? I ran the kill-all and now my ceph pods are in unknown state