EKS 1.26 end of standard support, June 2024

[sblair-metrostar]

Context

Need to upgrade EKS to 1.29 (current as of January 2024) before June 2024.

https://docs.aws.amazon.com/eks/latest/userguide/kubernetes-versions.html

Value and/or benefit

Saves additional costs associated with extended support.

https://aws.amazon.com/blogs/containers/amazon-eks-extended-support-for-kubernetes-versions-pricing/

Anything else?

No response

[Adam-D-Lewis]

BTW, Only version <= 1.28 is supported on Azure atm, and support for 1.26 expires end of March.

[Adam-D-Lewis]

I looked at upgrading to from 1.26 to 1.27 via Azure Console and it looks like the user-scheduler is using a deprecated API that is removed in k8s 1.27.

Also related: jupyterhub/mybinder.org-deploy#2587
stackabletech/demos#8 makes it look like it's already fixed.

[Adam-D-Lewis]

I went ahead with the upgrade to 1.27 and the upgrade failed due to my cloud Quotas not being high enough, but after I fixed that things seemed to work fine. I then upgraded to 1.28 and things seemed to work fine as well.
I went through the release checklist

• Log into keycloak as root user

  • Add a user

• Log into conda-store and create

  • a conda environment in a shared namespace and,
  • a conda environment in your personal namespace

• Launch dask-gateway cluster, test auto-scaler and

  • Validate that the dask-labextention is working

• Confirm that a notebook can be submitted via Jupyter-Scheduler - worked fine after adding papermill to the environment

• Open VS-Code extension

  • Add the Python extension
  • Create a .py file and run it

• dask compute failed likely due to version mismatch likely related to [BUG] - Default Dask environment gets out of date #2260

/home/conda/nebari-git/5475fd44-1708041785-4-dask/lib/python3.11/site-packages/distributed/client.py:1391: VersionMismatchWarning: Mismatched versions found

+-------------+----------------+-----------------+---------+
| Package     | Client         | Scheduler       | Workers |
+-------------+----------------+-----------------+---------+
| dask        | 2024.2.0       | 2022.11.1       | None    |
| distributed | 2024.2.0       | 2022.11.1       | None    |
| python      | 3.11.6.final.0 | 3.10.13.final.0 | None    |
| tornado     | 6.3.3          | 6.1             | None    |
+-------------+----------------+-----------------+---------+
  warnings.warn(version_module.VersionMismatchWarning(msg[0]["warning"]))

[Adam-D-Lewis]

Based on that, I think supporting at least up through k8s 1.28 will be very straight forward.

[dcmcand]

I have run 1.29 locally and everything seems to be working with some small changes to the local deployment that shouldn't impact cloud deployments. I think we should target 1.29 support.