AKS Learn feedback: Set new cordon behavior on AKS for upgrades

[ragatgen]

Type of issue

Other (describe below)

Feedback

Bug Report: undrainableNodeBehavior: Cordon causes persistent upgrade failure in AKS
Environment:

AKS Cluster Name: mynewakscluster
CLI Version (WSL/PowerShell): 2.60.0
Azure Portal CLI Version: 2.71.0
OS: WSL (Ubuntu 18.04) and Windows PowerShell
❗ Issue Summary
When attempting to update an AKS node pool using the undrainableNodeBehavior: Cordon setting, the cluster becomes stuck in an invalid state, and any further updates or scaling operations fail with the following error:

Failed to save Kubernetes service 'mynewakscluster'. Error: The value of parameter agentPoolProfile.upgrade.undrainableNodeBehavior is invalid. Error details: Undrainable node behavior 'Cordon' is not GAed for now.
Even after removing the flag from CLI commands, the error persists, indicating the value is persisted in the cluster configuration.

🔁 Steps to Reproduce
Attempt to update a node pool using:

az aks nodepool update
--cluster-name mynewakscluster
--name nodepool1
--resource-group myresourcegroup
--undrainable-node-behavior Cordon

Observe that the command fails or partially applies the setting.

Try to update or scale the node pool again (even without the flag), and the following error appears:

Failed to save Kubernetes service 'mynewakscluster'...
Attempting to register the feature via:

az feature register --name UndrainableNodeBehavior --namespace Microsoft.ContainerService

results in:

(FeatureNotFound) The feature 'UndrainableNodeBehavior' could not be found.
🧼 Workaround
The only way to recover from this state is to:

Delete the affected node pool:
az aks nodepool delete --cluster-name mynewakscluster --name nodepool1 --resource-group myresourcegroup

Recreate it without the unsupported setting:
az aks nodepool add --cluster-name mynewakscluster --name nodepool1 --resource-group myresourcegroup --node-count 3 --node-vm-size Standard_DS2_v2 --kubernetes-version

📌 Expected Behavior
The CLI should either prevent setting a non-GA feature like undrainableNodeBehavior: Cordon, or
There should be a supported way to remove/reset the setting once applied.
📎 Additional Notes
The CLI version mismatch between local environments (2.60.0) and the Azure Portal (2.71.0) may contribute to inconsistent behavior.
The feature is not listed in az feature list and cannot be registered manually.

Page URL

https://learn.microsoft.com/en-us/azure/aks/upgrade-cluster

Content source URL

https://github.com/MicrosoftDocs/azure-aks-docs/blob/main/articles/aks/upgrade-cluster.md

Author

@schaffererin

Document Id

1ce12b78-c666-18d2-dee6-b58465c81cf5

Platform Id

fb60b4a5-5ea6-bcd6-aea2-e0d1cc1fd347

[palma21]

There's a few things (likely) in play here

The new cordon behavior recently went GA, but that is only present in the latest CLI version (listed in the doc), you need 2.74 minimum version for this. Feature Flags are required for features while in preview, and keep in mind that for preview features an aks-preview CLI extension is required and the extension takes priority over the main one (it overrides it), which can explain differences in environment. az --version should give you version of the main CLI and any extension.
https://learn.microsoft.com/en-us/azure/aks/upgrade-cluster#set-new-cordon-behavior

If you had an issue with the feature after running the command (can't see that from your output, just that the CLI version didn't know the feature) do please open a support ticket or reference in the AKS GH issues repo.