Skip to main content
BumbleTap
Add to ChromeInstall

Kubernetes 1.35 End of Life (EOL): 28 February 2027, and the EKS, GKE and AKS Dates

Upstream Kubernetes stops patching 1.35 on 28 February 2027. EKS, GKE and AKS set later deadlines, then charge for extended support or upgrade you.

Written Shahzeb Umer

Subscribe

Status

Supported

Support ends .

Time left
101 days
Latest release
1.35.10
Move to
Kubernetes 1.37fixed until 28 Oct 2027

1.2 years of support

  • Bug and security fixes17 Dec 2025 – 28 Dec 2026
  • Security fixes only28 Dec 2026 – 28 Feb 2027

Supported for 1.2 years, from 17 Dec 2025 to 28 Feb 2027.

Who still patches it

Five places Kubernetes 1.35 runs, five deadlines.

  • Kubernetes (upstream): Patches end 28 Feb 2027
  • Amazon EKS: Extended support to 27 Mar 2028
  • Google GKE: Extended channel to 11 Feb 2028
  • Azure AKS (months only): LTS to Mar 2028
  • DigitalOcean DOKS: Upgraded from 28 Mar 2027
Sources: kubernetes.io, Amazon EKS version calendar, GKE release schedule, AKS supported versions, DigitalOcean supported releases. Checked 11 Nov 2026.

Where to move

Be on Kubernetes 1.37.

The short version

  • The Kubernetes project stops patching 1.35 on 28 February 2027. It enters maintenance mode on 28 December 2026, after which a 1.35 release ships only for CVE fixes, dependency updates and critical bugs. The latest patch is 1.35.10.
  • The managed providers set most teams' deadlines. EKS standard support ends on 27 March 2027, and extended support at $0.60 per cluster per hour instead of $0.10 runs to 27 March 2028. GKE ends standard support on 11 April 2027 and AKS in March 2027, and both upgrade clusters left behind.
  • Move to 1.37, going through 1.36, because control planes upgrade one minor version at a time. No beta or GA API version is removed in 1.36 or 1.37. The work is in node runtimes (containerd 2.x), SELinux volume sharing, stricter IP validation and renamed metrics.
On this page
  1. What ends upstream
  2. The dates that actually decide it: EKS, GKE and AKS
  3. What breaks first
  4. How to check what you run
  5. Which version to move to: 1.37
  6. How to upgrade
  7. Changes most likely to bite from 1.35 onward
  8. Sources

Kubernetes 1.35 reaches end of life upstream on 28 February 2027. After that date the Kubernetes project publishes no more 1.35 patch releases, security fixes included. The latest is 1.35.10, tagged on 27 October 2026. A running 1.35 cluster keeps working afterwards, because nothing in it checks a date.

Most teams run Kubernetes on a managed service, and the managed services set the dates that matter. Amazon EKS ends standard support on 27 March 2027 and then bills six times the hourly cluster fee. Google GKE ends standard support on 11 April 2027 and Azure AKS in March 2027, and both upgrade clusters still on 1.35 afterwards. On GKE's Regular channel, automatic upgrades from 1.35 to 1.36 are scheduled to begin on or after about 11 November 2026.

What ends upstream

The Kubernetes project maintains release branches for the three most recent minor versions, currently 1.37, 1.36 and 1.35. Each gets roughly 14 months of patches: twelve months of monthly patch releases, then two months of maintenance mode. In maintenance mode, release managers may still cut a release, but only for vulnerabilities with an assigned CVE, dependency issues including base image updates, and critical core component issues. At the end of maintenance mode the branch is end of life, and cherry picks to it close soon afterwards.

For 1.35 that means:

  • 17 December 2025: 1.35.0 released.
  • Until 28 December 2026: monthly patch releases.
  • 28 December 2026 to 28 February 2027: maintenance mode.
  • 28 February 2027: end of life. No further releases.

1.34 reached end of life on 27 October 2026, so 1.35 is now the oldest branch the project maintains. AWS adds that the community discourages CVE submissions against unsupported versions, so a flaw specific to 1.35 may never be reported at all.

The dates that actually decide it: EKS, GKE and AKS

Platform Standard support ends Extended support ends Extra cost during extended support What happens at the end Source
Kubernetes upstream 28 February 2027 (maintenance mode from 28 December 2026) Not offered None No more patch releases kubernetes.io releases
Amazon EKS 27 March 2027 27 March 2028 $0.60 per cluster per hour, up from $0.10 Clusters enter extended support by default. After 27 March 2028, EKS upgrades the control plane; nodes stay on 1.35 EKS version calendar
Google GKE 11 April 2027 11 February 2028 (Extended channel only) $0.50 per cluster per hour on top of the $0.10 management fee GKE upgrades control planes and nodes to the next minor version GKE release schedule
Azure AKS March 2027 (month only) March 2028 with LTS (month only) LTS needs the Premium tier: $0.60 per cluster per hour, against $0.10 for Standard 60-day grace period, then a platform-initiated upgrade, planned for June 2027 (June 2028 on LTS) AKS supported versions
DigitalOcean DOKS 28 February 2027 Not published Not published Automatic upgrade to the next minor version from 28 March 2027 DOKS supported releases

Amazon EKS counts 14 months of standard support from its own 1.35 release on 27 January 2026, and AWS backports applicable security patches for that whole period, past the upstream date. Extended support keeps patching the control plane, the VPC CNI, kube-proxy and CoreDNS add-ons and the EKS optimized AMIs, and it is on by default, so a 1.35 cluster nobody touches moves into it on 27 March 2027. Setting the upgrade policy to STANDARD avoids the charge, and EKS upgrades the cluster at the end of standard support instead. Decide before the date: a cluster in extended support cannot change its policy until it runs a version in standard support again. At the end of extended support EKS upgrades the control plane without notice, leaves node groups on 1.35, and does not allow a rollback.

Google GKE counts 14 months from 1.35's arrival in the Regular channel on 11 February 2026. Clusters on the Rapid, Regular and Stable channels, and those with no channel, are upgraded on their channel's schedule before 11 April 2027, and GKE upgrades any left after that date, nodes typically within a month. Extended channel clusters can stay until 11 February 2028 with security patches only, and are then upgraded even if they use deprecated APIs. In an emergency, a maintenance exclusion with the No upgrades scope delays the end of support upgrade by up to 90 days.

Azure AKS lists 1.35's end of life as March 2027 and gives only the month. Its documentation says a month-only date means the last day of that month, so 31 March 2027. After that the cluster is in platform support: Microsoft supports the Azure platform but not Kubernetes bugs or CVEs, and you can no longer create new 1.35 clusters. A 60-day grace period follows, and then AKS runs a platform-initiated upgrade, which its calendar schedules for June 2027. Long Term Support backports security fixes to 1.35 until March 2028, with an upgrade to the next LTS version planned for June 2028. AKS never enrolls a cluster in LTS for you.

DigitalOcean follows the upstream date and starts upgrading remaining 1.35 clusters to the next minor version on 28 March 2027, in their maintenance windows, whether or not auto-upgrade is enabled. It emails 30 days, 7 days and one day before.

What breaks first

Nothing in a 1.35 cluster fails on a fixed day. The trouble arrives as a series of upgrades and invoices, in this order.

GKE upgrades start now. Google's schedule estimates Regular channel upgrades to 1.36 on or after 11 November 2026 and Stable channel upgrades in January 2027. GKE pauses the upgrade if it detects use of a feature removed in the next minor version, but says it cannot guarantee detection. A maintenance exclusion is the only reliable hold.

Patches thin out from 28 December 2026. In maintenance mode a 1.35 release ships only for CVEs, dependency updates and critical bugs. From 28 February 2027 nothing ships.

The bills change in spring 2027. On EKS, from 27 March 2027, each cluster gains the usage type AmazonEKS-Hours:extendedSupport at $0.50 an hour beside AmazonEKS-Hours:perCluster at $0.10, each with a region prefix such as USE1-. That is about $365 a month extra per cluster at 730 hours. On GKE, Extended channel clusters pay the GKE extended period cluster management fee from 11 April 2027. On AKS, the Premium tier that LTS requires costs $0.60 an hour from the day you switch, and Azure's retail price list carries that rate as a meter named Standard Long Term Support.

AKS flags the cluster after 31 March 2027. A cluster on an unsupported version reports Degraded in Azure Resource Health, and the portal shows a banner naming when AKS will upgrade it. The June 2027 upgrade ignores your maintenance window, skips the deprecated API check, and can override a Pod Disruption Budget that blocks a node drain.

Forced upgrades follow: DigitalOcean from 28 March 2027, GKE from 11 April 2027, AKS from June 2027, and EKS after 27 March 2028 (or from 27 March 2027 for clusters with the STANDARD policy). On EKS only the control plane moves, and the node groups are left for you.

Removed APIs are a smaller risk this time. The deprecated API migration guide lists its most recent removal of beta or GA API versions in 1.32, and none since. The 1.36 and 1.37 changelogs drop only alpha versions (scheduling.k8s.io/v1alpha1 and v1alpha2), which are off by default. An automatic upgrade is unlikely to fail on a missing apiVersion. The risks are behaviour changes, listed at the end of this page.

How to check what you run

The control plane version, the kubelet on every node, and the container runtime:

# Server Version is the control plane; Client Version is your kubectl
kubectl version

# VERSION is each node's kubelet; CONTAINER-RUNTIME shows containerd 1.x or 2.x
kubectl get nodes -o wide

Amazon EKS, per region:

# One cluster
aws eks describe-cluster --name my-cluster --query cluster.version --output text

# Every cluster in a region, with its version and upgrade policy (STANDARD or EXTENDED)
for c in $(aws eks list-clusters --region us-east-1 --query 'clusters[]' --output text); do
  echo "$c $(aws eks describe-cluster --region us-east-1 --name "$c" \
    --query 'cluster.[version,upgradePolicy.supportType]' --output text)"
done

# The published support dates for 1.35
aws eks describe-cluster-versions --cluster-versions 1.35

Google GKE:

# Every cluster in the project, with control plane version and release channel
gcloud container clusters list \
  --format="table(name,location,currentMasterVersion,releaseChannel.channel)"

# Node pool versions, then upgrade targets and end of support dates
gcloud container node-pools list --cluster=CLUSTER_NAME --location=LOCATION \
  --format="table(name,version)"
gcloud container clusters get-upgrade-info CLUSTER_NAME --location=LOCATION

Azure AKS:

# Every cluster in the subscription
az aks list -o table

# Version, support plan (KubernetesOfficial or AKSLongTermSupport) and tier
az aks list --query "[].{name:name, group:resourceGroup, version:currentKubernetesVersion, plan:supportPlan, tier:sku.tier}" -o table

# The versions this cluster can upgrade to
az aks get-upgrades --resource-group RESOURCE_GROUP --name CLUSTER_NAME -o table

Which version to move to: 1.37

Upstream maintains 1.37, 1.36 and 1.35 (latest patches 1.37.2, 1.36.6 and 1.35.10). EKS has 1.37 to 1.34 in standard support. GKE offers 1.37 in the Rapid, Regular and Extended channels and expects it in Stable in January 2027. AKS made 1.37 generally available in October 2026.

Version Upstream end of life EKS standard support ends GKE standard support ends AKS end of life
1.36 28 June 2027 2 August 2027 9 August 2027 June 2027 (month only)
1.37 28 October 2027 1 December 2027 22 December 2027 October 2027 (month only)

Target 1.37. 1.36 buys about four months over 1.35 and puts you back here in mid 2027. 1.37 lasts until late 2027 on every platform. On GKE's Stable channel, take 1.36 now and 1.37 when it arrives.

You cannot skip 1.36 on the control plane. The Kubernetes version skew policy says kube-apiserver must not skip minor versions, even in a single-instance cluster, and EKS, GKE and AKS all enforce it. The one exception is AKS, where clusters enrolled in Long Term Support can jump between LTS versions.

Nodes can lag. Upstream allows a kubelet up to three minor versions older than kube-apiserver and never newer, so 1.35 nodes are within policy under a 1.37 control plane. GKE documents a two-version limit for its nodes, and EKS recommends matching node and control plane versions before each control plane upgrade. Drain a node before a minor kubelet upgrade: the skew policy does not support upgrading it in place.

How to upgrade

  1. Read what the provider already found. EKS runs upgrade insights on every cluster every 24 hours. GKE publishes deprecation insights. AKS shows Kubernetes API deprecations under Diagnose and solve problems in the portal and alerts through Azure Advisor.

    # EKS upgrade insights
    aws eks list-insights --region us-east-1 --cluster-name my-cluster
    
    # GKE deprecation insights
    gcloud recommender insights list \
      --insight-type=google.container.DiagnosisInsight \
      --location=LOCATION --project=PROJECT_ID
  2. Find deprecated API calls yourself. The API server records every deprecated API version that was requested, with the release that removes it. For manifests and Helm releases, Pluto checks apiVersion fields, and the kubectl convert plugin rewrites old manifests.

    kubectl get --raw /metrics | grep apiserver_requested_deprecated_apis
    pluto detect-helm -owide
    kubectl convert -f ./my-deployment.yaml --output-version apps/v1
  3. Upgrade the control plane to 1.36.

    # EKS
    aws eks update-cluster-version --name my-cluster --kubernetes-version 1.36 --region us-east-1
    
    # GKE (a minor version valid for the cluster's release channel)
    gcloud container clusters upgrade CLUSTER_NAME --master \
      --location=LOCATION --cluster-version=1.36
    
    # AKS (control plane first, then each node pool; VERSION from az aks get-upgrades)
    az aks upgrade --resource-group RESOURCE_GROUP --name CLUSTER_NAME --kubernetes-version VERSION
  4. Upgrade nodes and add-ons. EKS does not upgrade managed node groups, self-managed nodes or Fargate Pods with the control plane, and recommends updating the VPC CNI, CoreDNS and kube-proxy add-ons next. GKE upgrades node pools automatically by default. An in-place AKS upgrade does the node pools after the control plane.

  5. Repeat for 1.37. EKS can roll an in-place upgrade back to the previous minor version within 7 days, so leave a few days between steps.

Changes most likely to bite from 1.35 onward

1. Nodes still on containerd 1.x (High likelihood on custom node images). The 1.35 release announcement calls 1.35 the last release to support containerd 1.x and says to switch to containerd 2.0 or later before upgrading. containerd's own compatibility table lists only 2.2 and 2.3 for Kubernetes 1.36, and containerd 1.7 reached end of life on 30 September 2026. The 1.36 and 1.37 changelogs deferred removal of the related kubelet flags and fallback behaviour to 1.38, so an old node may keep running for now. Look for containerd://1. in the CONTAINER-RUNTIME column, or watch the kubelet's kubelet_cri_losing_support metric.

2. SELinux volume labelling in 1.37 (Medium; High on SELinux-enforcing nodes). SELinuxMount is on by default in 1.37. A volume mount carries one SELinux label, so Pods with different SELinux labels sharing a volume on the same node can stay in ContainerCreating. It applies only where the CSI driver sets seLinuxMount: true. Setting seLinuxChangePolicy: Recursive on a Pod keeps the old behaviour, and 1.37 is the last release where the feature can be disabled cluster-wide.

kubectl get csidriver -o custom-columns=NAME:.metadata.name,SELINUXMOUNT:.spec.seLinuxMount

3. Stricter IP and CIDR validation in 1.36 (Medium). StrictIPCIDRValidation is on by default, so the API server rejects IP addresses with leading zeros, such as 010.000.000.005, and CIDRs with host bits set, such as 192.168.0.5/24. Objects already stored are kept, but creates and updates are rejected, so the failure shows up the next time a Helm chart or pipeline applies the value.

4. Renamed and removed metrics (Medium). 1.36 renamed volume_operation_total_errors to volume_operation_errors_total and etcd_bookmark_counts to etcd_bookmark_total. In 1.37 the kubelet stopped exporting container_cpu_load_average_10s, container_cpu_load_d_average_10s and container_tasks_state. Dashboards and alerts on these go quiet without any error.

5. Ingress NGINX is retired (Medium). Kubernetes retired Ingress NGINX on 24 March 2026, with no further releases, bug fixes or security fixes. Existing installs keep working, and its last release predates Kubernetes 1.36.

6. Self-managed node and kubeadm configuration (Low). In 1.37 the kubelet refuses to start if a deprecated cAdvisor flag is set (--housekeeping-interval is kept), eventRecordQPS: 0 now means no limit, and static Pods can no longer reference Secrets or ConfigMaps. kubeadm 1.37 removed its v1beta3 configuration API, and kubeadm config migrate from the 1.35 binary converts it. EKS says its optimized AMIs set neither the deprecated cAdvisor flags nor eventRecordQPS.

7. gitRepo volumes stop running in 1.36 (Low). The gitRepo volume plugin is permanently disabled. The API still accepts such Pods, and the kubelet refuses to run them. Use an init container or a git-sync sidecar.

8. Warnings now, removals later (Low). Service .spec.externalIPs warns from 1.36, with removal planned for 1.43. kube-proxy's ipvs mode logs a deprecation warning from 1.37, and the release announcement expects it to be off by default by 1.40.

Sources

Common questions

When does Kubernetes 1.35 reach end of life?
Upstream, on 28 February 2027. The Kubernetes project puts 1.35 into maintenance mode on 28 December 2026 and publishes no 1.35 patch releases after 28 February 2027. Managed services run later: Amazon EKS ends standard support on 27 March 2027, Google GKE on 11 April 2027, and Azure AKS in March 2027 (Microsoft gives only the month).
When does EKS end support for 1.35?
Standard support ends on 27 March 2027 and extended support on 27 March 2028, per the Amazon EKS release calendar. Clusters move into extended support automatically unless their upgrade policy is set to STANDARD. After 27 March 2028 EKS upgrades the control plane to the next version, without advance notice, and leaves managed node groups and self-managed nodes on 1.35.
How much does EKS extended support cost?
$0.60 per cluster per hour, against $0.10 in standard support. That is $0.50 an hour more, or about $365 a month per cluster at 730 hours. Billing starts at the beginning of the day standard support ends, in UTC, which for 1.35 is 27 March 2027. In AWS billing data the extra charge is the usage type AmazonEKS-Hours:extendedSupport with a region prefix, such as USE1-AmazonEKS-Hours:extendedSupport.
Will GKE upgrade my 1.35 cluster automatically?
Yes. Google's release schedule estimates that Regular channel clusters start moving to 1.36 on or after 11 November 2026, and Stable channel clusters in January 2027 (an estimate given to the month). GKE pauses those upgrades if it detects use of a feature removed in the next minor version, but only until 1.35's end of standard support on 11 April 2027. Extended channel clusters can stay on 1.35 until 11 February 2028 for an extra $0.50 per cluster per hour from 11 April 2027.
Which Kubernetes version should I upgrade to?
1.37. Upstream patches it until 28 October 2027, and all three big providers offer it: EKS standard support runs to 1 December 2027, GKE standard support to 22 December 2027 and AKS support to October 2027 (month only). Getting there from 1.35 takes two control plane upgrades, through 1.36. GKE's Stable channel does not offer 1.37 until about January 2027, so Stable clusters stop at 1.36 for now.
Can I skip Kubernetes versions when upgrading?
Not on the control plane. The Kubernetes version skew policy says kube-apiserver must not skip minor versions, and EKS, GKE and AKS all upgrade a control plane one minor version at a time, so 1.35 to 1.37 is two upgrades. Nodes have more room: upstream allows a kubelet up to three minor versions older than the API server, and GKE supports two. AKS lets clusters enrolled in Long Term Support jump between LTS versions.
Does AKS offer extended support for 1.35?
Yes, through Long Term Support, which keeps 1.35 patched until March 2028 (Microsoft gives the month only). LTS requires the Premium tier, which Azure prices at $0.60 per cluster per hour against $0.10 for the Standard tier, and you pay it from the day you switch. Without LTS, 1.35 reaches end of life in March 2027 and AKS plans to upgrade remaining clusters from June 2027.

Also on the calendar

End-of-life dates for the rest of your stack.

Dates checked against endoflife.date (MIT). How we check datesReport a wrong dateEvery Kubernetes version