k8s v1.28.2 and NFD updates

[mythi]

No description provided.

[mythi]

@uniemimu PTAL 5cca930

[tkatila]

$ kubectl get node -o json | jq .metadata.labels | grep gpu.intel.com
  "gpu.intel.com/cards": "card0",
  "gpu.intel.com/device-id.0300-9a49.count": "1",
  "gpu.intel.com/device-id.0300-9a49.present": "true",
  "gpu.intel.com/gpu-numbers": "0",
  "gpu.intel.com/memory.max": "0",
  "gpu.intel.com/millicores": "1000",
  "gpu.intel.com/tiles": "1",
$ kubectl get node -o json | jq .status.allocatable
{
  "cpu": "8",
  "ephemeral-storage": "450548187210",
  "gpu.intel.com/i915": "8",
  "gpu.intel.com/i915_monitoring": "1",
  "gpu.intel.com/memory.max": "0",
  "gpu.intel.com/millicores": "1000",
  "gpu.intel.com/tiles": "1",
  "hugepages-1Gi": "0",
  "hugepages-2Mi": "0",
  "memory": "32474604Ki",
  "pods": "110"
}

I recall the extended resources vanished from the node labels previously. Now they seem to stay in labels. I wonder if it's by design?

[eero-t]

As an occasional project contributor... I think it would be good to add some note about how NFD is intended to be used to DEVEL.md.

[mythi]

As an occasional project contributor... I think it would be good to add some note about how NFD is intended to be used to DEVEL.md.

@eero-t GPU plugin documentation talks about it and we also have similar steps documented as part of the operator deployment.

[mythi]

I recall the extended resources vanished from the node labels previously. Now they seem to stay in labels. I wonder if it's by design?

Fixed by adding a custom nfd-worker.conf where "local" (and all other feature sources) are disabled as "label sources".

[codecov-commenter]

Codecov Report

Merging #1539 (d7b5db8) into main (ba3ded1) will decrease coverage by 0.11%.
Report is 8 commits behind head on main.
The diff coverage is 20.00%.

❗ Current head d7b5db8 differs from pull request most recent head 8687d2c. Consider uploading reports for the commit 8687d2c to get more accurate results

@@            Coverage Diff             @@
##             main    #1539      +/-   ##
==========================================
- Coverage   49.58%   49.48%   -0.11%     
==========================================
  Files          42       42              
  Lines        4917     4923       +6     
==========================================
- Hits         2438     2436       -2     
- Misses       2339     2345       +6     
- Partials      140      142       +2     

Files Changed	Coverage Δ
pkg/controllers/dlb/controller.go	17.35% <0.00%> (ø)
pkg/controllers/dsa/controller.go	6.77% <0.00%> (ø)
pkg/controllers/fpga/controller.go	16.66% <0.00%> (ø)
pkg/controllers/gpu/controller.go	47.70% <0.00%> (ø)
pkg/controllers/iaa/controller.go	6.77% <0.00%> (ø)
pkg/controllers/qat/controller.go	9.72% <0.00%> (ø)
pkg/controllers/sgx/controller.go	10.74% <0.00%> (ø)
pkg/controllers/reconciler.go	4.56% <28.57%> (+0.30%)	⬆️
cmd/qat_plugin/dpdkdrv/dpdkdrv.go	83.27% <100.00%> (ø)

... and 1 file with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[eero-t]

@eero-t GPU plugin documentation talks about it and we also have similar steps documented as part of the operator deployment.

I did not mean end user documentation of installing NFD rules (updated in this PR), but some mention in internal documentation on how (NFD) labels are supposed to be configured for the plugins, e.g. in the DEVEL.md "Checklist for New Device Plugins" section.

[mythi]

@eero-t GPU plugin documentation talks about it and we also have similar steps documented as part of the operator deployment.

I did not mean end user documentation of installing NFD rules (updated in this PR), but some mention in internal documentation on how (NFD) labels are supposed to be configured for the plugins, e.g. in the DEVEL.md "Checklist for New Device Plugins" section.

Added: "+9. Plugin NodeFeatureRules added for Node Feature Discovery labeling."

[mythi]

@hj-johannes-lee PTAL