feat(profiling): add GC observability collector - #19940
Draft
vlad-scherbich wants to merge 7 commits into
Draft
Conversation
Adds GCCollector that hooks gc.callbacks to measure per-generation pause durations, emits alloc samples for collected objects, and records explicit gc.collect() call counts per flush interval. Controlled via DD_PROFILING_GC_ENABLED (default: true).
Contributor
|
Contributor
Author
|
@codex review |
Codeowners resolved asResolved from the full PR diff against |
Circular import analysis
|
Dependency direction analysis
|
Contributor
There was a problem hiding this comment.
Pull request overview
Adds a new profiling “GC observability” collector that hooks gc.callbacks to emit profiling samples around GC pauses, plus a new configuration flag (DD_PROFILING_GC_ENABLED) to control enablement, and wires the collector into the main Profiler.
Changes:
- Introduces
GCCollectorthat patchesgc.collect, subscribes togc.callbacks, emits wall-time samples for GC pauses, and emits a periodic “gc.config” snapshot sample. - Adds
DD_PROFILING_GC_ENABLEDto profiling settings and supported configuration registry, and enables the collector inddtrace.profiling.profiler.Profiler. - Adds unit + integration tests validating callback lifecycle and that GC samples appear in exported profiles.
Reviewed changes
Copilot reviewed 6 out of 6 changed files in this pull request and generated 5 comments.
Show a summary per file
| File | Description |
|---|---|
| tests/profiling/collector/test_gc.py | New unit/integration tests for GCCollector behavior and emitted samples. |
| supported-configurations.json | Registers new DD_PROFILING_GC_ENABLED configuration key. |
| ddtrace/profiling/profiler.py | Wires GCCollector into the profiler collector initialization flow. |
| ddtrace/profiling/collector/gc.py | New collector implementation using gc.callbacks + ddup.SampleHandle. |
| ddtrace/internal/settings/profiling.py | Adds ProfilingConfigGC (profiling_config.gc.enabled). |
| ddtrace/internal/settings/_supported_configurations.py | Updates generated allowlist to include DD_PROFILING_GC_ENABLED. |
Suppressed comments (1)
ddtrace/profiling/collector/gc.py:79
- push_alloc(size, count) treats the first argument as alloc-space (bytes) and the second as alloc-samples. Passing the number of collected objects as the size and a constant count of 1 will misrepresent the data in the allocation profile (object count will appear as bytes).
collected = info.get("collected", 0)
if collected > 0:
handle2 = ddup.SampleHandle()
handle2.push_alloc(collected, 1)
handle2.push_frame(frame_name, "gc", 0, gen)
handle2.push_monotonic_ns(time.monotonic_ns())
handle2.flush_sample()
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 5fff59f9fc
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
2 tasks
…ngs reload Config-sample uniqueness raced with gc.callbacks walltime, and importlib.reload replaced ProfilingConfigHeap so later isinstance() checks failed.
Restore gc.collect only when our wrapper is still installed, drain the explicit-call counter under a lock, and reset in-flight state after fork. Align help/docstrings with emitted samples and add a release note.
Keep the new GC pause collector off until customers set DD_PROFILING_GC_ENABLED=true.
Comment on lines
+2880
to
+2884
| { | ||
| "implementation": "C", | ||
| "type": "boolean", | ||
| "default": "true" | ||
| } |
Comment on lines
+94
to
+95
| # TODO: If we later want collected-object counts on the profile, add a | ||
| # dedicated libdatadog/pprof sample type. Do not reuse push_alloc. |
Comment on lines
+88
to
+92
| handle: ddup.SampleHandle = ddup.SampleHandle() | ||
| handle.push_walltime(pause_ns, 1) | ||
| handle.push_frame(frame_name, "gc", 0, gen) | ||
| handle.push_monotonic_ns(time.monotonic_ns()) | ||
| handle.flush_sample() |
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
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.
Adds GCCollector that hooks gc.callbacks to measure per-generation pause durations, emits alloc samples for collected objects, and records explicit gc.collect() call counts per flush interval. Controlled via DD_PROFILING_GC_ENABLED (default: true).
Description
Testing
Risks
Additional Notes