The last two items should be done with the non-active DC going first.
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| knative: Bump all images to latest release 1.12.x | operations/docker-images/production-images | master | +60 -778 |
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | JMeybohm | T341984 Update Kubernetes clusters to 1.31 | |||
| Open | None | T391943 Q4 24-25 Goal: Operational Excellence - LiftWing Platform Updates & Improvements | |||
| Open | None | T398948 Q1 FY2025-26 Goal: Operational Excellence - LiftWing Platform Updates & Improvements | |||
| Declined | klausman | T380723 Update knative-serving+net-istio to v1.12.x on ML clusters |
Change #1097345 had a related patch set uploaded (by Klausman; author: Klausman):
[operations/docker-images/production-images@master] knative: Bump all images to latest release 1.16.0
V1.16.0 requires Istio 1.20, so I have backed down the version to the v.1.12 series. This would still make k8s v1.28 and 1.29 the recommended versions to run this with, as per https://kserve.github.io/website/0.13/admin/serverless/serverless/:
| Kubernetes Version | Recommended Istio Version | Recommended Knative Version |
| 1.27 | 1.18,1.19 | 1.10,1.11 |
| 1.28 | 1.19,1.20 | 1.11,1.12.4 |
| 1.29 | 1.20,1.21 | 1.12.4,1.13.1 |
The recommended version matrix for kserve 0.15 looks like this:
| Kubernetes Version | Recommended Istio Version | Recommended Knative Version |
| 1.30 | 1.22,1.23 | 1.15,1.16 |
| 1.31 | 1.24,1.25 | 1.16,1.17 |
| 1.32 | 1.25,1.26 | 1.17,1.18 |
Since we are aiming to upgrade to kserve 0.15 we could jump to one of the latest versions in order to be able to upgrade to k8s 1.31 T341984: Update Kubernetes clusters to 1.31
This is a big bump both for istio and knative so the scope of this task would change completely. I see that istio 1.24 is already available in our stack https://docker-registry.wikimedia.org/istio/proxyv2/tags/
Istio 1.24 is already bundled with the 1.31 k8s upgrade, so that part should be ok :)
We run knative 1.7, the jump to 1.16 is big but I think that the worst part was the migration to PSS (already done, so the config is now compatible with newer versions). We should test it to make sure that everything works of course, but in theory we should be able to upgrade the Docker images and test what config changes are needed.
I'm closing this task as it is outdated. There is work being done in T419722: Experiment with new kserve version on ml-staging-codfw to move directly to kserve 0.17 where the compatible knative version is being discussed