What's the proper way to gracefully and safely shut down an entire k3s cluster in preparation for a physical move?

[matthewadams]

Hi,

We've seen https://docs.k3s.io/upgrades/killall#k3s-service but aren't clear that this is the safest, most graceful way to shutdown the whole cluster. We are prepping to physically move the three physical nodes to another location. By safe & graceful, I mean that we won't lose any data on volumes, we won't corrupt dbs, etc.

Also, for good measure, what would be the proper way to bring the cluster back up?

Like, we'd love a k3s server and/or agent command like k3s shutdown or similar that gracefully brings down the whole kit & caboodle, preserving all data, etc. This seems like a problem unique to on-prem clusters, so it seems warranted for k3s to provide a convenient means to achieve shutdown & restart.

[brandond]

Just drain the nodes in whatever order your application prefers to handle terminating the workload, then run the killall script to clean up the rest.