{"id":1379,"date":"2026-11-16T23:30:00","date_gmt":"2026-11-16T23:30:00","guid":{"rendered":"https:\/\/cotocus.org\/blog\/?p=1379"},"modified":"2026-11-19T23:30:00","modified_gmt":"2026-11-19T23:30:00","slug":"six-cluster-myths-kubernetesops","status":"publish","type":"post","link":"https:\/\/cotocus.org\/blog\/six-cluster-myths-kubernetesops\/","title":{"rendered":"Six Cluster Myths I Keep Hearing in KubernetesOps Conversations"},"content":{"rendered":"<p>Every few months I sit in a room where somebody says something about Kubernetes with total confidence and half the heads nod. Sometimes it is true. More often it is a half-truth that has survived because it is convenient. Here are six that keep resurfacing in conversations about running clusters, checked against the entry for KubernetesOps in <a href=\"https:\/\/www.xopsschool.com\/kubernetesops\/\">the Ops Atlas<\/a> and against ordinary operating experience.<\/p>\n<h2>Myth 1: &#8220;KubernetesOps is a job title with its own certificate&#8221;<\/h2>\n<p>It sounds official, which is why it spreads. In reality the phrase is informal. The Atlas is upfront that no foundation, book or exam adopts the label, and that it clashes with several unrelated products and a consultancy&#8217;s methodology. The work itself is cluster administration, and the credential most people mean is the CKA, a hands-on exam from the Linux Foundation where troubleshooting carries the heaviest weighting at 30 percent.<\/p>\n<p><strong>Reality check:<\/strong> use the term as shorthand if you like, but put the actual skill, cluster administration, on the job advert.<\/p>\n<h2>Myth 2: &#8220;We can jump straight to the latest version&#8221;<\/h2>\n<p>I have watched a team plan exactly this over a long weekend. The API server cannot skip minor versions; you move one at a time, with control-plane parts first, node agents after, and nodes drained before each hop. The Atlas suggests that if the gap is too wide to cross sequentially, the honest alternative is a new cluster and a migration, not a heroic leap.<\/p>\n<p><strong>Reality check:<\/strong> the cheapest upgrade is the one you take while the gap is still one step.<\/p>\n<h2>Myth 3: &#8220;A disruption budget makes our service safe&#8221;<\/h2>\n<p>Budgets are valuable, and I would never advise removing them. But they only guard against voluntary disruptions such as drains. When a node runs short of memory, the kubelet evicts pods regardless of any budget, picking victims by how far their usage exceeds their requests. Preemption by higher-priority workloads honours budgets only on a best-effort basis. And a budget that permits zero disruption does not protect a service so much as freeze maintenance.<\/p>\n<p><strong>Reality check:<\/strong> pair budgets with honest resource requests and a plan for node loss.<\/p>\n<h2>Myth 4: &#8220;Managed means we can stop thinking about the control plane&#8221;<\/h2>\n<p>A managed service really does remove the burden of running the API server and the datastore. It does not remove everything. The Atlas notes that one major provider regularly backs up its managed clusters, yet customers still need their configuration in Git, backups of their own data and a rebuild they have rehearsed. The treadmill of versions, the add-on inventory and request hygiene stay with you.<\/p>\n<p><strong>Reality check:<\/strong> managed shifts the line of responsibility; it does not erase it.<\/p>\n<h2>Myth 5: &#8220;Ingress has been removed, so we must migrate this quarter&#8221;<\/h2>\n<p>Two separate facts get blended here, and that is where teams get hurt. The older Ingress interface has stopped gaining features, yet the project says it has no intention of deleting it. A newer routing model, the Gateway API, is what the docs steer new work toward. Then there is the controller: one very common implementation had its repository archived in late March 2026 and ships no further security fixes, which is a perfectly good reason to schedule a move even though the interface itself is staying put. The Atlas calls the conversion utility an assistant for migrations, not a replacement you run once and forget, and notes that certain annotations, for instance the one limiting body size, lack a counterpart.<\/p>\n<p><strong>Reality check:<\/strong> the API is staying, the controller is the risk, and the migration needs a human to read the output.<\/p>\n<h2>Myth 6: &#8220;Our platform is self-driving&#8221;<\/h2>\n<p>I like automation as much as anyone, but I distrust the phrase. As the Atlas frames it, claims of a self-driving cluster deserve scepticism until the vendor can show how the system handles worker machines, add-ons and retired APIs. Anyone who has watched an autoscaler follow bad requests straight into waste knows that machines amplify the quality of the inputs they are given.<\/p>\n<p><strong>Reality check:<\/strong> automate the routine steps, and keep a person accountable for the decisions around them. This applies equally to the operational analytics approach covered in <a href=\"https:\/\/cotocus.org\/blog\/aiops-roadmap-a-practical-guide-for-devops-sre-and-operations-teams\/\">our practical roadmap for operations teams adopting machine-assisted analysis<\/a>, where tools sharpen judgement rather than replace it.<\/p>\n<h2>Why these stories stick around<\/h2>\n<p>Each of the six offers a way to avoid a boring task: reading removal notes, drilling a restore, fixing requests, keeping a calendar. That convenience is exactly why they survive. A myth that asks nothing of you will always beat a fact that asks for an afternoon, and repeating it in meetings gives it a respectability it has not earned. Believing them feels cheaper than doing the work, right up to the day the bill arrives. The Atlas lists &#8220;we&#8217;ll upgrade next quarter&#8221; as an anti-pattern for much the same reason, since delays compound and hops must be sequential.<\/p>\n<p>The Atlas also suggests a gentle way to turn scepticism into process: write each corrected belief into the team runbook, with the date you tested it, so the next newcomer inherits evidence instead of folklore.<\/p>\n<p>Should you act on only one thing here, keep it modest. Choose whichever myth you suspect is alive in your own team and put it to a test within seven days. Check the version lag, list the budgets that allow no disruption, or time a restore into a scratch cluster. Teams building reliable data products will see the same instinct at work in <a href=\"https:\/\/cotocus.org\/blog\/mastering-mlops-for-scalable-and-reliable-ai-products\/\">our guide to dependable machine learning products<\/a>: assumptions are cheap to state and costly to leave untested.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Six persistent beliefs about running Kubernetes clusters, from skipping versions to trusting managed services blindly, and what the evidence actually says.<\/p>\n","protected":false},"author":4,"featured_media":1380,"comment_status":"","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[130],"tags":[],"class_list":["post-1379","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-it-operations"],"_links":{"self":[{"href":"https:\/\/cotocus.org\/blog\/wp-json\/wp\/v2\/posts\/1379","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cotocus.org\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/cotocus.org\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/cotocus.org\/blog\/wp-json\/wp\/v2\/users\/4"}],"replies":[{"embeddable":true,"href":"https:\/\/cotocus.org\/blog\/wp-json\/wp\/v2\/comments?post=1379"}],"version-history":[{"count":1,"href":"https:\/\/cotocus.org\/blog\/wp-json\/wp\/v2\/posts\/1379\/revisions"}],"predecessor-version":[{"id":1515,"href":"https:\/\/cotocus.org\/blog\/wp-json\/wp\/v2\/posts\/1379\/revisions\/1515"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/cotocus.org\/blog\/wp-json\/wp\/v2\/media\/1380"}],"wp:attachment":[{"href":"https:\/\/cotocus.org\/blog\/wp-json\/wp\/v2\/media?parent=1379"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/cotocus.org\/blog\/wp-json\/wp\/v2\/categories?post=1379"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/cotocus.org\/blog\/wp-json\/wp\/v2\/tags?post=1379"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}