[EKS]: One-click Full Cluster Upgrade

[mohitanchlia]

Automation in EKS as a service has come in only bits and pieces. EKS with managed nodes is not really that useful without having a one click full upgrade where EKS version, aws-node, dns etc. along with the worker nodes get upgraded without running or orchestrating commands manually. This can be broken down into 2 pieces 1) Full cluster upgrade with nodes 2) Only worker node upgrades for ongoing AMI rotation. This is a critical functionality

[tabern]

Renaming this to 'One-click Full Cluster Upgrade'. Today EKS managed nodes supports worker node upgrades for AMI rotation, but this is not yet in the console (#605). See API documentation here: https://docs.aws.amazon.com/eks/latest/APIReference/API_UpdateNodegroupVersion.html

[mtparet]

I just read the official documentation of AWS to upgrade an eks cluster : we have to manually execute kubectl commands to upgrade critical components of Kubernetes.
Really, for a "managed" service, is it a joke ?
Even updating my on-premise cluster is easier than updating the called "managed" AWS Kubernetes service.

https://docs.aws.amazon.com/eks/latest/userguide/update-cluster.html

[irperez]

In Azure AKS, this is a single click operation with a version drop down.

[sandrom]

The upgrade process cant get enough love, please work on this! It is incredibly important for production workloads to have a managed process for this :)

[raravena80]

+1 for this

[ghost]

+1
EKS need to close the gaps with AKS and GKE

[nydalal]

+1

[damscott]

This feature could impact people working under the assumption that AWS will not modify resources running inside the cluster, notably kube-proxy (#657)

[kylecompassion]

+1

[alexey-pankratyev]

+1

[smrutiranjantripathy]

Hi Team,

There is a sample package eks-one-click-cluster-upgrade in aws-samples which provides similar functionality. This is a cli utility which can be used to carry out upgrade. Please check this package and share your feedback.

[kareem-elsayed]

We still can say that EKS it's not a fully managed k8s engine if we are going for every release to spend time and effort to check the release note for every addon.

[MMartyn]

One other aspect of upgrades that would be great to be a part of this effort would be the ability to configure the upgrade timeout for node groups. Currently if it takes longer than 15 minutes to replace a node it fails the upgrade and rolls back.

Related doc: https://docs.aws.amazon.com/eks/latest/userguide/managed-node-update-behavior.html
Relevant part:

Drains the pods from the node. If the pods don't leave the node within 15 minutes and there's no force flag, the upgrade phase fails with a PodEvictionFailure error. For this scenario, you can apply the force flag with the update-nodegroup-version request to delete the pods.

[neerajprem]

+1, we must have it.

[eperdeme]

I'm not really understanding the value here. In a cluster where you've got many addons to make the cluster function such as istio, Argo, external-dns, Prometheus, cert-manager...etc.etc. VPC CNI/kube-proxy bumping is trivial task and easily automated via your gitops managment methods.

Is the target for the customers with more out of the box Kubernetes with a basic set of aws provided addons ?