Hi everyone,
I just upgraded my lab from NKP 2.19.0 to 2.19.1, and this time the management cluster went through the UI as well (Administration → Upgrades), not only the workload cluster.
I wrote down the timings and a few details that may help anyone planning the same patch.
What I measured
- Management cluster (3 control planes + 5 workers): 43 min 28 s from the click until every HelmRelease was ready
- Workload cluster (3 control planes + 3 workers): 16 min 48 s
- Test app with 3 replicas: 4,089 of 4,089 requests answered during the whole window
Worth knowing before you start
- The dashboard flags new versions on its own, much like Prism Central LCM, from a release inventory that Kommander refreshes every hour.
- A patch still replaces every node. 2.19.1 keeps Kubernetes 1.36.2, but the new node image recreates every VM. Plan the window by the number of VMs, not by the version number.
- "Completed" comes before the last VM. In a patch, the tracker shows Completed while Cluster API is still replacing VMs. The ControlPlaneMachinesUpToDate and WorkerMachinesUpToDate conditions tell you when the window is really over.
- Keep headroom for the rolling update. It creates before it deletes: one extra VM per node group. In my lab, 12 vCPU, 48 GiB, 160 GiB and 2 IPs per cluster.
- Read the release notes on the portal. The known issues in the wizard are a release-day summary, and the portal listed two. Read them the day before.
Full write-up
Commands, the pre-upgrade checklist, the verification steps and two real-time videos of the node replacement:
👉 NKP 2.19.1 from the UI, on leimar.dev
Any comments or suggestions are very welcome.
Leimar