Fix for Node Feature Discovery Operator in OCP 4.17

[carlmes]

GitOps projects that use the nfd operator have been failing in OpenShift 4.17, it looks like the base image has been changed - we verified this by manually installing the operator and observing the new image name.

Update base image in node-feature-discovery.yaml fixes the issue.

We've tested this on an OpenShift 4.17 and 4.15 cluster, and NFD is now working great. More details in one of the downstream projects at: redhat-ai-services/ai-accelerator#80

[strangiato]

@carlmes is this still backwards compatible with older versions?

I'm wondering if we need to do a seperate overlay for 4.17 to maintain backwards compatibility.

[strangiato]

Something else to be aware of, we have been testing composer on 4.17 and NFD does not like wake up from a cluster shutdown in demo.redhat.com.

We noticed this image was different in 4.17 and we weren't sure if that was an issue related to that specific image, or if it was something specific to 4.17.

That would be something good to test in 4.16 with that same image.

This is the bug Anthony filed for NFD for reference:
https://issues.redhat.com/browse/OCPBUGS-44471

[carlmes]

@carlmes is this still backwards compatible with older versions?

I'm wondering if we need to do a seperate overlay for 4.17 to maintain backwards compatibility.

We tested it on OpenShift 4.17 as well as 4.15 clusters to check for backwards compatibility, screenshots in the downstream project update: redhat-ai-services/ai-accelerator#80

Do you think we should try an even earlier version of OpenShift? The lowest version we can provision on demo.redhat.com is 4.14.

[carlmes]

Tested this patched version on my temporary cluster by stopping and then restarting the cluster through admin UI:

NFD Operator seems to start up just fine, so this might fix the other issue as well.

[strangiato]

Nice!

Should have known you would have tested everything so thoroughly!