chore(profiles): sync GPU profiles with pcie_topology/NUMA from upstream (RUN-40173)#212
Merged
Merged
Conversation
eliranw
force-pushed
the
eliranw/RUN-40173-sync-gpu-profiles-numa
branch
from
4565e19 to
b9f3198
Compare
…eam (RUN-40173) Re-sync built-in GPU profiles from upstream NVIDIA/k8s-test-infra main (commit 497fa04), which adds the pcie_topology block (PCI root complexes + per-device numa_node) plus the gb300 profile -- what the mock backend needs to surface GPU->NUMA for NodeResourceTopology / topology-aware scheduling. hack/sync-profiles.sh now defaults to tracking main and accepts a tag, branch, OR commit SHA (blobless clone + checkout instead of clone --branch); the generated header records the exact synced commit for provenance. Also fix the sync-gpu-profiles workflow version extraction (head -2 -> head -3) so it captures the Source line. Signed-off-by: Eliran Wolff <eliranw@nvidia.com>
eliranw
force-pushed
the
eliranw/RUN-40173-sync-gpu-profiles-numa
branch
from
b9f3198 to
7a30881
Compare
iris-shain-runai
approved these changes
Jun 4, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Re-syncs the built-in GPU profiles via
hack/sync-profiles.shfrom upstream NVIDIA/k8s-test-inframain(synced commit497fa04), and improves the sync tooling.The previous pin was the
v0.1.0tag, which predates upstream'spcie_topologyblock (PCI root complexes + per-devicenuma_node). Without it, the mock backend's device-plugin can't report GPU→NUMA, soNodeResourceTopologyzones never get GPU resources and NUMA/topology-aware scheduling can't be exercised on mock GPUs.pcie_topologycurrently exists only on upstreammain(no tagged release yet).Changes
hack/sync-profiles.sh:DEFAULT_VERSIONnow tracksmain(the committedbuiltin.yamlis the pinned artifact; each sync produces a reviewable PR diff). Override with a tag/commit once upstream cuts a release.git checkout <ref>(the oldgit clone --branchonly accepted tags/branches, though the workflow advertises "tag or commit").# Source: NVIDIA/k8s-test-infra main (commit 497fa04).builtin.yaml(regenerated): all 7 profiles now carrypcie_topology(a100/b200/gb200/gb300/h100/l40s/t4);gb300is new..github/workflows/sync-gpu-profiles.yml: fix version extraction (head -2→head -3) — the# Source:line is line 3, so it previously captured nothing.CHANGELOG.md: entry under[Unreleased] → Changed.Verification
hack/sync-profiles.sh(default →main) ran clean → 7 profiles, header pinned to497fa04.builtin.yamlparses as valid YAML (7 ConfigMap docs); each profile containspcie_topology.Context
Found while testing the mock backend on a real-NUMA EKS node: the real NFD topology-updater produces a populated NRT (CPU zone
cap=8 alloc=7), but GPUs don't appear in the zones because the rendered mock profile had nopcie_topology. This sync is the prerequisite for mock-backend GPU↔NUMA.RUN-40173