Add metrics observer sidecar, NIXL handshake recorder, and signal-based profiling - #2
Draft
thameem-abbas wants to merge 1 commit into
Draft
Conversation
…ed profiling - metrics_observer.py: Prometheus metrics observer that detects throughput plateau via coefficient of variation, then writes signal file to trigger profiling at steady state. Runs as sidecar container, stdlib only. - sitecustomize.py: Add NixlHandshakeRecorder to time _nixl_handshake() calls via import hook. Add signal_mode that waits for signal file instead of fixed call ranges before starting torch.profiler. - webhook.py: Add observer sidecar injection (build_observer_sidecar_patch), new annotation mappings for nixl-handshake, signal-mode, observer config. Auto-detects vLLM port from pod spec. - kustomization.yaml: Add metrics_observer.py to ConfigMap generator. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> Signed-off-by: thameem-abbas <thibrahi@redhat.com>
thameem-abbas
marked this pull request as draft
May 1, 2026 12:09
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
/metrics, detects throughput plateau (coefficient of variation < threshold over sliding window), then writes/tmp/profiler_startsignal file. Stdlib only, no external deps.NixlHandshakeRecorderwith import hook to time_nixl_handshake()calls and export to JSON. Add signal-mode profiling that waits for signal file instead of fixed call ranges.vllm.profiler/observerannotation, new annotation mappings for nixl-handshake/signal-mode/observer config, auto-detect vLLM port from pod spec.metrics_observer.pyto ConfigMap generator.Test plan
vllm.profiler/observer: "true"annotation, verify metrics-observer sidecar is injected/tmp/nixl_handshake_timings.json🤖 Generated with Claude Code