{"id":1188,"date":"2026-11-16T20:10:00","date_gmt":"2026-11-16T20:10:00","guid":{"rendered":"https:\/\/charterkings.com\/blog\/?p=1188"},"modified":"2026-11-19T15:10:00","modified_gmt":"2026-11-19T15:10:00","slug":"kubernetesops-upgrade-steps-trip-plan","status":"publish","type":"post","link":"https:\/\/charterkings.com\/blog\/kubernetesops-upgrade-steps-trip-plan\/","title":{"rendered":"Eight Steps to Upgrade a Cluster, Planned Like a Charter Trip"},"content":{"rendered":"<p>A charter flight succeeds on sequence: crew briefed, aircraft inspected, passengers boarded, doors closed, then wheels up. Move one step and the whole trip wobbles. Upgrading a Kubernetes cluster is much the same, and the people who do it for a living call this kind of work KubernetesOps. What follows is a plain, numbered procedure for moving a cluster up by one minor version, written for readers who like the logic of a trip plan. It condenses the guidance in <a href=\"https:\/\/www.xopsschool.com\/kubernetesops\/\">the Ops Atlas page on KubernetesOps<\/a>, and your own platform&#8217;s documentation should still be the final word.<\/p>\n<h2>Step 1: Confirm where you stand<\/h2>\n<p>Before any move, write down three facts: the version you run, the newest release available, and the date your version stops getting patches. The upstream project ships about three minors a year, and the Atlas notes that as of mid-October 2026 one older branch was due to leave maintenance on 27 October. If your cluster sits on a branch like that, the upgrade is no longer optional. Think of checking the weather and the aircraft&#8217;s inspection due date before quoting a client.<\/p>\n<h2>Step 2: Hunt for removed features<\/h2>\n<p>After a beta interface is deprecated it stays available for nine months or three minor releases, which gives you a window but not a long one. Your goal is to find workloads and tools that still call interfaces about to disappear. The Atlas names two sources of evidence: a metric on the API server that counts requests to deprecated interfaces, and audit annotations that mark them. One warning matters here: the metric only covers requests since the server last started, so a freshly restarted instance can look cleaner than it really is.<\/p>\n<h2>Step 3: Check the skew rules<\/h2>\n<p>Do not upgrade anything until the version rules are on paper. The API server must go up one minor at a time. With several API servers for resilience, the gap between them is capped at one minor. Node agents are allowed to trail the API server by as many as three minors, yet they may never run ahead of it. If you use a hosted service, confirm whether its nodes must match the control plane, as the Atlas says is the case for managed and serverless nodes on one major provider.<\/p>\n<h2>Step 4: Take a backup and test the way back<\/h2>\n<p>Self-managed clusters need an etcd snapshot, and the Atlas&#8217;s warning is blunt: whoever can read that store holds root-level power over the cluster, so lock the snapshot away. Also confirm what rollback really means on your platform. A kubeadm rerun can be repeated safely and keeps backups under a temporary configuration folder; a hosted service may only permit going back one minor within a limited window. Know this before you need it, the way a crew knows the diversion options before departure. It is much easier to choose between an undo and a rebuild in a calm meeting than in the middle of an incident.<\/p>\n<h2>Step 5: Upgrade the control plane<\/h2>\n<ol>\n<li>Upgrade the API server first.<\/li>\n<li>Then upgrade the controller manager, the scheduler and, where present, the cloud controller manager.<\/li>\n<li>Confirm health checks pass before moving on.<\/li>\n<\/ol>\n<p>Resist the temptation to do everything in parallel. The order exists so that newer components never talk to older ones in unsupported ways.<\/p>\n<h2>Step 6: Move the nodes one at a time<\/h2>\n<p>For each node, drain it so its workloads move elsewhere, upgrade the kubelet, and return the node to service. Drains use the Eviction API. If the API replies with a 429 status, an eviction was refused for now, perhaps because a disruption budget would be broken, and the right response is to retry. A 500 can signal that two budgets select the same pod. The Atlas also recommends setting the policy that lets unhealthy pods be evicted, since the default can leave a broken pod blocking the whole drain. Force-deleting pods is a last resort to agree with the workload&#8217;s owner, not a shortcut.<\/p>\n<p>A practical rule for timing: schedule node moves for quiet hours, and give each hop its own change window instead of stacking several. A trip planner would never stack three tight connections on one itinerary, and the same instinct applies here. If one node refuses to drain, stop, find the budget or the workload responsible and fix that cause before touching the next machine.<\/p>\n<h2>Step 7: Finish with the network proxy and the add-ons<\/h2>\n<p>The proxy follows the kubelets. After that, review the add-ons that sit alongside the cluster, such as the metrics server, which the Atlas notes is meant only for autoscaling use. Check each add-on&#8217;s compatibility with the new version instead of assuming it moved along with the rest.<\/p>\n<h2>Step 8: Debrief and update the calendar<\/h2>\n<p>Record how long each stage took, what surprised you, and what the next hop should change. The Atlas suggests tracking upgrade lead time as a measure, and teams that have done two consecutive hops from a written runbook are, in its terms, repeatable. Then book the next date while the memory is fresh.<\/p>\n<h2>Two reading suggestions from this site<\/h2>\n<p>Planning habits translate across industries. For the passenger-facing side of preparation, see <a href=\"https:\/\/charterkings.com\/blog\/luxury-travel-preferences-every-charter-provider-should-consider\/\">how luxury travellers judge a charter operator<\/a>. For the shorter-hop mindset, <a href=\"https:\/\/charterkings.com\/blog\/helicopter-travel-etiquette-for-first-time-passengers\/\">etiquette for first-time helicopter passengers<\/a> shows how clear briefings prevent confusion.<\/p>\n<h2>A closing checklist<\/h2>\n<p>Keep it short enough to print: version recorded, removed features hunted, skew rules checked, backup taken and rollback understood, control plane done, nodes one by one, add-ons reviewed, debrief written. If every box is ticked, you have done what the discipline asks, and the next hop becomes easier because the steps are already on paper.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>A numbered, no-nonsense procedure for a one-minor Kubernetes upgrade, from checking support dates to draining nodes and writing the debrief.<\/p>\n","protected":false},"author":6,"featured_media":1189,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[798],"tags":[],"class_list":["post-1188","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-aviation-careers"],"_links":{"self":[{"href":"https:\/\/charterkings.com\/blog\/wp-json\/wp\/v2\/posts\/1188","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/charterkings.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/charterkings.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/charterkings.com\/blog\/wp-json\/wp\/v2\/users\/6"}],"replies":[{"embeddable":true,"href":"https:\/\/charterkings.com\/blog\/wp-json\/wp\/v2\/comments?post=1188"}],"version-history":[{"count":1,"href":"https:\/\/charterkings.com\/blog\/wp-json\/wp\/v2\/posts\/1188\/revisions"}],"predecessor-version":[{"id":1313,"href":"https:\/\/charterkings.com\/blog\/wp-json\/wp\/v2\/posts\/1188\/revisions\/1313"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/charterkings.com\/blog\/wp-json\/wp\/v2\/media\/1189"}],"wp:attachment":[{"href":"https:\/\/charterkings.com\/blog\/wp-json\/wp\/v2\/media?parent=1188"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/charterkings.com\/blog\/wp-json\/wp\/v2\/categories?post=1188"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/charterkings.com\/blog\/wp-json\/wp\/v2\/tags?post=1188"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}