Update

Flux¶

This is automated using Github Action. Sometimes the Flux components may fail to start due to outdated CRDs. In that case, make sure you are on the latest Flux CLI (brew upgrade fluxcd/tap/flux) and run buvisctl update flux.

Cilium¶

Cilium is deployed as a Talos inline manifest (not managed by Helm or Flux). Values are in infrastructure/talos/cilium-values.yaml.

To upgrade, run buvisctl update cilium <version> in the cluster directory. This renders manifests, updates the Talos patch for future bootstraps, and applies to the running cluster.

Talos¶

When siderolabs release a new Talos version (<VERSION_TAG>):

1. Check for new issues to see if the new version is safe to use
2. Update nodes: buvisctl update talos
3. Check nodes versions: talosctl -n $NODE_IPS version | awk '/^[[:space:]]*Client:/{ctx="Client"} /^[[:space:]]*NODE:/{ctx=$2} /^[[:space:]]*Tag:/{printf "%s %s %s\n", (ctx=="client"?"0":"1"), ctx, $2}' | sort | awk 'BEGIN {printf "\n%-15s %-8s\n", "Node", "Version"; print "--------------- --------"} {printf "%-15s %-8s\n", $2, $3}'
4. Push Talos version changes to the repository

Proxmox¶

1. Connect to Proxmox node: ssh <NODE_NAME>
2. Update packages: apt update && apt dist-upgrade && apt autoremove
3. Reboot: reboot

Post-update¶

Longhorn managers may need to be restarted after any update: kubectl -n storage rollout restart ds/longhorn-manager