{"id":11941,"date":"2026-11-16T17:30:00","date_gmt":"2026-11-16T17:30:00","guid":{"rendered":"https:\/\/www.cotocus.com\/blog\/?p=11941"},"modified":"2026-11-19T07:30:00","modified_gmt":"2026-11-19T07:30:00","slug":"nine-numbers-kubernetes-cluster-health","status":"publish","type":"post","link":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/","title":{"rendered":"Nine Numbers That Show How Healthy a Kubernetes Cluster Is"},"content":{"rendered":"<p>Cluster upkeep tends to be argued about in opinions and settled, eventually, by outages. A better habit is to agree on a short list of numbers and review them on a schedule. The list below follows the metrics catalogue in <a href=\"https:\/\/www.xopsschool.com\/kubernetesops\/\">the Ops Atlas page on KubernetesOps<\/a>, with my own commentary on how to read each one. Treat them as nine gauges on one dashboard, not a scorecard to be gamed, and expect to retire or replace some as your own practice matures and different questions become more pressing.<\/p>\n<h2>1. Version lag<\/h2>\n<p>How many minor releases behind the newest supported version is production? The Atlas suggests a target of one or fewer. Because upgrades have to be taken sequentially, lag is the number that compounds: a team three minors behind faces three separate hops, each with its own removals to check, and the longer the gap the more tempting it is to delay again.<\/p>\n<h2>2. Days to end of support<\/h2>\n<p>Lag counts versions; this counts time. Each minor receives around fourteen months of patches, twelve standard and two in maintenance for critical fixes. On 16 October 2026, the Atlas notes, one branch had eleven days left. Putting a day count on a dashboard turns an abstract worry into a deadline a manager can read.<\/p>\n<h2>3. Upgrade lead time<\/h2>\n<p>Measure the elapsed time from a new minor appearing to your production clusters running it. If the figure keeps growing, your process is accumulating friction: untested runbooks, unclear owners, too many manual steps. A shrinking lead time is the clearest sign that a team has stopped dreading upgrades.<\/p>\n<h2>4. Request allocation<\/h2>\n<p>What share of a node pool&#8217;s allocatable capacity is already promised to pods through requests? In the Atlas worked case, ten machines with sixteen CPUs each expose 145 allocatable CPUs after system reservations, and 120 requested gives an allocation of 82.8 percent. High allocation with idle machines is a sign of inflated requests, not of a busy cluster.<\/p>\n<h2>5. Usage-to-request ratio<\/h2>\n<p>The companion to number four. If the 95th-percentile usage in the same example is 42 CPUs against 120 requested, the ratio is 35 percent. Sizing requests at the observed peak with a 30 percent cushion, then planning for 75 percent utilisation, lands near 55 CPUs, which means six machines where ten were running. A ratio far below one is money; a ratio near or above one is risk, since pods that exceed their requests are first in line for eviction.<\/p>\n<h2>6. Share of pods with requests<\/h2>\n<p>An easy one to measure and a good gate for a maturity step. The Atlas ties a managed level of practice to at least 98 percent of pods declaring requests. Anything below that leaves the scheduler and autoscalers guessing, and an admission rule can enforce the number automatically.<\/p>\n<h2>7. Unschedulable pods<\/h2>\n<p>Count pods stuck pending because nothing satisfies their requirements. A brief spike during scale-up is normal; a persistent plateau means a capacity gap, a mis-sized request or an over-tight placement rule. Alert on the duration, not on the instant count.<\/p>\n<h2>8. Restart rate and API latency<\/h2>\n<p>I group these because both are early warnings. Restarts climbing across replicas often point to probes that check dependencies, a pattern the Atlas flags as a cause of cascading failure. For the API server, the community&#8217;s scalability objectives are a 99th-percentile latency of one second for single-object writes and five seconds for stateless pod start. If you drift well beyond those, look at the number of objects, webhooks in the request path and the health of the datastore.<\/p>\n<h2>9. Restore readiness<\/h2>\n<p>The most neglected gauge. Record the date of your last successful restore from a backup and how long it took. The Atlas retells a spring 2023 incident at a big social network: about five hours spent restoring from a datastore backup, with the company admitting its written procedure had gone stale. A restore-readiness figure such as &#8220;drilled 41 days ago in 38 minutes&#8221; is far more reassuring than a statement that backups exist.<\/p>\n<h2>Reading the nine together<\/h2>\n<p>No single gauge is decisive. Versions that are current but backups that are untested give you a smooth ride into a hard stop. Excellent requests with a lag of four minors give you efficient machines that will fall out of support. The Atlas ties the numbers to its maturity ladder: a measured practice reviews them monthly and has tracked both lead time and usage-to-request ratio for two quarters.<\/p>\n<p>If you already run a measurement culture around data pipelines, the approach will feel familiar. Our pieces on <a href=\"https:\/\/www.cotocus.com\/blog\/the-essential-role-of-mlops-in-machine-learning-success\/\">why operational discipline decides machine learning outcomes<\/a> and on <a href=\"https:\/\/www.cotocus.com\/blog\/mlops-roadmap-a-complete-guide-for-data-scientists-and-engineering-teams\/\">a staged roadmap for machine learning teams<\/a> make the same argument in another domain: pick a few numbers, review them regularly and let trends, not anecdotes, direct the work.<\/p>\n<h2>Setting up the dashboard<\/h2>\n<p>Resist the urge to build a perfect system before you start. Three gauges in a spreadsheet, updated by hand each month, will teach you more than nine gauges that nobody trusts.<\/p>\n<p>You do not need exotic tooling. Cluster state exporters and the API server&#8217;s own metrics supply most of the inputs, and a cost-allocation tool can fill in the capacity gauges. One caution from the Atlas: the standard resource metrics component is meant only for autoscaling, so do not treat it as a general monitoring source. Begin with a monthly review and add trend lines after two quarters. Publish the dashboard somewhere the whole team already looks. What you are after is a shared, unexciting picture of reality that survives staff changes and busy weeks.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Nine measurable signals for judging a cluster&#8217;s upkeep, from version lag to restore readiness, with the thresholds and worked figures that make each one useful.<\/p>\n","protected":false},"author":37,"featured_media":11942,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[147],"tags":[],"class_list":["post-11941","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-devops"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.6 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Nine Numbers That Show How Healthy a Kubernetes Cluster Is - Cotocus<\/title>\n<meta name=\"description\" content=\"Nine KubernetesOps metrics worth tracking, from version lag and days to end of support to restore readiness, each with a threshold or worked example.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Nine Numbers That Show How Healthy a Kubernetes Cluster Is - Cotocus\" \/>\n<meta property=\"og:description\" content=\"Nine KubernetesOps metrics worth tracking, from version lag and days to end of support to restore readiness, each with a threshold or worked example.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/\" \/>\n<meta property=\"og:site_name\" content=\"Cotocus\" \/>\n<meta property=\"article:published_time\" content=\"2026-11-16T17:30:00+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-11-19T07:30:00+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.cotocus.com\/blog\/wp-content\/uploads\/2026\/11\/b2a-banner-286054-banner.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1200\" \/>\n\t<meta property=\"og:image:height\" content=\"630\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Nikhil Farooqui\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Nikhil Farooqui\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/nine-numbers-kubernetes-cluster-health\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/nine-numbers-kubernetes-cluster-health\\\/\"},\"author\":{\"name\":\"Nikhil Farooqui\",\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/#\\\/schema\\\/person\\\/419007bef27e8d85890550d783b96c50\"},\"headline\":\"Nine Numbers That Show How Healthy a Kubernetes Cluster Is\",\"datePublished\":\"2026-11-16T17:30:00+00:00\",\"dateModified\":\"2026-11-19T07:30:00+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/nine-numbers-kubernetes-cluster-health\\\/\"},\"wordCount\":912,\"commentCount\":0,\"image\":{\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/nine-numbers-kubernetes-cluster-health\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/b2a-banner-286054-banner.png\",\"articleSection\":[\"DevOps\"],\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/nine-numbers-kubernetes-cluster-health\\\/#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/nine-numbers-kubernetes-cluster-health\\\/\",\"url\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/nine-numbers-kubernetes-cluster-health\\\/\",\"name\":\"Nine Numbers That Show How Healthy a Kubernetes Cluster Is - Cotocus\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/nine-numbers-kubernetes-cluster-health\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/nine-numbers-kubernetes-cluster-health\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/b2a-banner-286054-banner.png\",\"datePublished\":\"2026-11-16T17:30:00+00:00\",\"dateModified\":\"2026-11-19T07:30:00+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/#\\\/schema\\\/person\\\/419007bef27e8d85890550d783b96c50\"},\"description\":\"Nine KubernetesOps metrics worth tracking, from version lag and days to end of support to restore readiness, each with a threshold or worked example.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/nine-numbers-kubernetes-cluster-health\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/nine-numbers-kubernetes-cluster-health\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/nine-numbers-kubernetes-cluster-health\\\/#primaryimage\",\"url\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/b2a-banner-286054-banner.png\",\"contentUrl\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/b2a-banner-286054-banner.png\",\"width\":1200,\"height\":630},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/nine-numbers-kubernetes-cluster-health\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Nine Numbers That Show How Healthy a Kubernetes Cluster Is\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/#website\",\"url\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/\",\"name\":\"Cotocus\",\"description\":\"Shaping Tomorrow\u2019s Tech Today\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/#\\\/schema\\\/person\\\/419007bef27e8d85890550d783b96c50\",\"name\":\"Nikhil Farooqui\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/d855702dac00a5af0405bb329101be495718d611233e52c86fe360c8fca6f6dc?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/d855702dac00a5af0405bb329101be495718d611233e52c86fe360c8fca6f6dc?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/d855702dac00a5af0405bb329101be495718d611233e52c86fe360c8fca6f6dc?s=96&d=mm&r=g\",\"caption\":\"Nikhil Farooqui\"},\"url\":\"https:\\\/\\\/www.cotocus.com\\\/blog\\\/author\\\/nikhil-farooqui\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Nine Numbers That Show How Healthy a Kubernetes Cluster Is - Cotocus","description":"Nine KubernetesOps metrics worth tracking, from version lag and days to end of support to restore readiness, each with a threshold or worked example.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/","og_locale":"en_US","og_type":"article","og_title":"Nine Numbers That Show How Healthy a Kubernetes Cluster Is - Cotocus","og_description":"Nine KubernetesOps metrics worth tracking, from version lag and days to end of support to restore readiness, each with a threshold or worked example.","og_url":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/","og_site_name":"Cotocus","article_published_time":"2026-11-16T17:30:00+00:00","article_modified_time":"2026-11-19T07:30:00+00:00","og_image":[{"width":1200,"height":630,"url":"https:\/\/www.cotocus.com\/blog\/wp-content\/uploads\/2026\/11\/b2a-banner-286054-banner.png","type":"image\/png"}],"author":"Nikhil Farooqui","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Nikhil Farooqui","Est. reading time":"4 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/#article","isPartOf":{"@id":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/"},"author":{"name":"Nikhil Farooqui","@id":"https:\/\/www.cotocus.com\/blog\/#\/schema\/person\/419007bef27e8d85890550d783b96c50"},"headline":"Nine Numbers That Show How Healthy a Kubernetes Cluster Is","datePublished":"2026-11-16T17:30:00+00:00","dateModified":"2026-11-19T07:30:00+00:00","mainEntityOfPage":{"@id":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/"},"wordCount":912,"commentCount":0,"image":{"@id":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/#primaryimage"},"thumbnailUrl":"https:\/\/www.cotocus.com\/blog\/wp-content\/uploads\/2026\/11\/b2a-banner-286054-banner.png","articleSection":["DevOps"],"inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/#respond"]}]},{"@type":"WebPage","@id":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/","url":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/","name":"Nine Numbers That Show How Healthy a Kubernetes Cluster Is - Cotocus","isPartOf":{"@id":"https:\/\/www.cotocus.com\/blog\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/#primaryimage"},"image":{"@id":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/#primaryimage"},"thumbnailUrl":"https:\/\/www.cotocus.com\/blog\/wp-content\/uploads\/2026\/11\/b2a-banner-286054-banner.png","datePublished":"2026-11-16T17:30:00+00:00","dateModified":"2026-11-19T07:30:00+00:00","author":{"@id":"https:\/\/www.cotocus.com\/blog\/#\/schema\/person\/419007bef27e8d85890550d783b96c50"},"description":"Nine KubernetesOps metrics worth tracking, from version lag and days to end of support to restore readiness, each with a threshold or worked example.","breadcrumb":{"@id":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/#primaryimage","url":"https:\/\/www.cotocus.com\/blog\/wp-content\/uploads\/2026\/11\/b2a-banner-286054-banner.png","contentUrl":"https:\/\/www.cotocus.com\/blog\/wp-content\/uploads\/2026\/11\/b2a-banner-286054-banner.png","width":1200,"height":630},{"@type":"BreadcrumbList","@id":"https:\/\/www.cotocus.com\/blog\/nine-numbers-kubernetes-cluster-health\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.cotocus.com\/blog\/"},{"@type":"ListItem","position":2,"name":"Nine Numbers That Show How Healthy a Kubernetes Cluster Is"}]},{"@type":"WebSite","@id":"https:\/\/www.cotocus.com\/blog\/#website","url":"https:\/\/www.cotocus.com\/blog\/","name":"Cotocus","description":"Shaping Tomorrow\u2019s Tech Today","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.cotocus.com\/blog\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Person","@id":"https:\/\/www.cotocus.com\/blog\/#\/schema\/person\/419007bef27e8d85890550d783b96c50","name":"Nikhil Farooqui","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/secure.gravatar.com\/avatar\/d855702dac00a5af0405bb329101be495718d611233e52c86fe360c8fca6f6dc?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/d855702dac00a5af0405bb329101be495718d611233e52c86fe360c8fca6f6dc?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/d855702dac00a5af0405bb329101be495718d611233e52c86fe360c8fca6f6dc?s=96&d=mm&r=g","caption":"Nikhil Farooqui"},"url":"https:\/\/www.cotocus.com\/blog\/author\/nikhil-farooqui\/"}]}},"_links":{"self":[{"href":"https:\/\/www.cotocus.com\/blog\/wp-json\/wp\/v2\/posts\/11941","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.cotocus.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.cotocus.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.cotocus.com\/blog\/wp-json\/wp\/v2\/users\/37"}],"replies":[{"embeddable":true,"href":"https:\/\/www.cotocus.com\/blog\/wp-json\/wp\/v2\/comments?post=11941"}],"version-history":[{"count":1,"href":"https:\/\/www.cotocus.com\/blog\/wp-json\/wp\/v2\/posts\/11941\/revisions"}],"predecessor-version":[{"id":12069,"href":"https:\/\/www.cotocus.com\/blog\/wp-json\/wp\/v2\/posts\/11941\/revisions\/12069"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.cotocus.com\/blog\/wp-json\/wp\/v2\/media\/11942"}],"wp:attachment":[{"href":"https:\/\/www.cotocus.com\/blog\/wp-json\/wp\/v2\/media?parent=11941"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.cotocus.com\/blog\/wp-json\/wp\/v2\/categories?post=11941"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.cotocus.com\/blog\/wp-json\/wp\/v2\/tags?post=11941"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}