Migrate archive-production-rke2 to the zfs snapshotter
After #5269 (closed), it looks like the zfs snapshotter works fine on our common workloads, so we can migrate the production cluster to it.
Prerequisite operations:
-
disable puppet on all nodes -
push change of snapshotter to puppet on all relevant nodes
For each node:
- cordon / drain the node
- apt-get update / dist-upgrade the node
- systemctl disable rke2-agent rancher-system-agent
- reboot the node
- move /var/lib/rancher/rke2/agent/containerd directory out of the way
- run puppet on the node
- enable rke2-agent rancher-system-agent.service
- start rancher-system-agent, wait for it to finish starting, start rke2-agent
- check system containers are being pulled and start
- uncordon the node
- clean up the old /var/lib/rancher/rke2/agent/containerd directory
Nodes:
-
rancher-node-production-rke2-mgmt1 -
rancher-node-metal01 -
rancher-node-metal02 -
rancher-node-metal03 -
rancher-node-metal04 -
saam -
banco