[https://nvbugs/5448767][fix] fix mpi4py deadlocks in pp event-loop #6976
Conversation
📝 WalkthroughWalkthroughAdds wait() calls on stored non-blocking send handles at the end of PyExecutor's per-pipeline-parallel loop. Updates an integration test to optionally prefix the MPI worker launch with NVIDIA NSYS profiling when NSYS_PATH and NSYS_FILE are set; otherwise worker invocation is unchanged. Changes
Sequence Diagram(s)sequenceDiagram
participant Prev as Prev batch & stats
participant Exec as PyExecutor._executor_loop_pp
participant Handles as send_handles
participant Next as Next iteration
Prev->>Exec: finish processing previous batch & stats
Exec->>Handles: iterate send_handles
loop for each handle
Exec->>Handles: handle.wait() if handle != None
end
Handles-->>Exec: all sends completed
Exec->>Next: continue loop
sequenceDiagram
participant Test as test_disaggregated
participant Env as Environment
participant NSYS as nsys
participant MPI as mpirun workers
Test->>Env: read NSYS_PATH & NSYS_FILE
alt profiling enabled
Test->>NSYS: build nsys profile command
NSYS->>MPI: prefix mpirun with nsys (nsys_cmd + mpirun ...)
MPI-->>Test: execute workers under NSYS
else profiling disabled
Test->>MPI: execute mpirun workers (no nsys prefix)
end
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Suggested reviewers
✨ Finishing Touches
🧪 Generate unit tests
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. CodeRabbit Commands (Invoked using PR/Issue comments)Type Other keywords and placeholders
Status, Documentation and Community
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (3)
tensorrt_llm/_torch/pyexecutor/py_executor.py (1)
852-855: Good fix: drain outstanding isends at loop exit to prevent PP hang. Add minor hardening.Waiting on all non-None send handles here is the right move to avoid leaving downstream ranks blocked in recv when an upstream rank exits early.
Consider a small robustness tweak to avoid leaving stale handles and to log MPI errors rather than failing the whole worker on shutdown:
Apply this diff:
- for req in self.send_handles: - if req is not None: - req.wait() + for i, req in enumerate(self.send_handles): + if req is None: + continue + try: + req.wait() + except Exception as e: + # Best-effort cleanup on shutdown; don't crash the worker + logger.warning(f"Error while waiting on pending send (slot {i}): {e}") + finally: + self.send_handles[i] = Nonetests/integration/defs/disaggregated/test_disaggregated.py (2)
159-165: NSYS_PATH is ignored; use it to locate the nsys binary.You read NSYS_PATH but always invoke "nsys" from PATH. Respect NSYS_PATH (allowing either a directory or full binary path) so profiling works in hermetic environments.
Apply this diff:
- nsys_path = os.getenv("NSYS_PATH", None) - nsys_file = os.getenv("NSYS_FILE", None) - nsys_cmd = [ - "nsys", "profile", "--trace", "cuda,cublas,nvtx", "--output", nsys_file, - "--force-overwrite", "true", "--duration", "180", - ] if nsys_path and nsys_file else [] + nsys_path = os.getenv("NSYS_PATH", None) + nsys_file = os.getenv("NSYS_FILE", None) + if nsys_path and nsys_file: + nsys_exe = os.path.join(nsys_path, "nsys") if os.path.isdir(nsys_path) else nsys_path + nsys_cmd = [ + nsys_exe, "profile", "--trace", "cuda,cublas,nvtx", "--output", nsys_file, + "--force-overwrite", "true", "--duration", "180", + ] + else: + nsys_cmd = []
169-173: mpirun -v enabled unconditionally; gate it to profiling-only.As written, mpirun’s verbose flag is always on. That’s noisy and can bloat CI logs. Enable it only when profiling is active.
Apply this diff:
- workers_cmd = nsys_cmd + [ - 'mpirun', '-v', '--allow-run-as-root', '--oversubscribe', '-n', - str(num_ranks), 'trtllm-serve', 'disaggregated_mpi_worker', '-c', - config_file - ] + workers_cmd = ( + nsys_cmd + [ + 'mpirun', '-v', '--allow-run-as-root', '--oversubscribe', '-n', + str(num_ranks), 'trtllm-serve', 'disaggregated_mpi_worker', '-c', config_file + ] + ) if nsys_cmd else [ + 'mpirun', '--allow-run-as-root', '--oversubscribe', '-n', + str(num_ranks), 'trtllm-serve', 'disaggregated_mpi_worker', '-c', config_file + ]
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
💡 Knowledge Base configuration:
- MCP integration is disabled by default for public repositories
- Jira integration is disabled by default for public repositories
- Linear integration is disabled by default for public repositories
You can enable these sources in your CodeRabbit configuration.
📒 Files selected for processing (2)
tensorrt_llm/_torch/pyexecutor/py_executor.py(1 hunks)tests/integration/defs/disaggregated/test_disaggregated.py(1 hunks)
🧰 Additional context used
📓 Path-based instructions (2)
**/*.py
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.py: Python code must target Python 3.8+
Python indentation: 4 spaces, no tabs
Maintain module namespace in imports (from package.subpackage import foo; then use foo.SomeClass())
Python file names use snake_case
Python class names use PascalCase
Python functions/methods and local variables use snake_case; variables starting with a number get k_ prefix (e.g., k_99th_percentile)
Global variables use G_ prefixed UPPER_SNAKE_CASE (e.g., G_MY_GLOBAL)
Constants use UPPER_SNAKE_CASE in Python
Avoid shadowing variables from outer scopes in Python
Initialize all externally visible members of a Python class in init
Prefer docstrings for interfaces used outside a file; comments for local code
Use Google-style docstrings for classes and functions (Sphinx-parsable)
Document attributes/variables inline with short docstrings
Avoid reflection when simple alternatives exist (e.g., prefer explicit parameters over dict(**locals()))
In try/except, catch the narrowest exceptions possible
For duck-typing with try/except, keep try body minimal and put logic in else
Files:
tensorrt_llm/_torch/pyexecutor/py_executor.pytests/integration/defs/disaggregated/test_disaggregated.py
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh,py}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
Prepend NVIDIA copyright header (current year) to all source files
Files:
tensorrt_llm/_torch/pyexecutor/py_executor.pytests/integration/defs/disaggregated/test_disaggregated.py
reasonsolo
force-pushed
the
bug5448767
branch
3 times, most recently
from
9aba239 to
f04fed7
Compare
|
/bot run |
|
PR_Github #15588 [ run ] triggered by Bot |
|
PR_Github #15588 [ run ] completed with state |
|
/bot run |
|
/bot run |
|
/bot run |
|
PR_Github #15878 [ run ] triggered by Bot |
|
PR_Github #15879 [ run ] triggered by Bot |
|
PR_Github #15880 [ ] completed with state |
|
PR_Github #15878 [ run ] completed with state |
reasonsolo
changed the title
|
/bot run |
|
PR_Github #15883 [ run ] triggered by Bot |
|
PR_Github #15879 [ run ] completed with state |
There was a problem hiding this comment.
I'm not sure the changes to synchronize the sampler right after _sample_async make sense. @Tabrizian @amukkara do you agree?
|
PR_Github #16389 [ run ] completed with state |
…eadlock; synchronize sampler right after async calls to avoid hang Signed-off-by: Lizhi Zhou <[email protected]>
|
/bot run |
|
PR_Github #16428 [ run ] triggered by Bot |
|
PR_Github #16428 [ run ] completed with state |
…eadlock; synchronize sampler right after async calls to avoid hang Signed-off-by: Lizhi Zhou <[email protected]>
|
/bot run |
|
PR_Github #16497 [ run ] triggered by Bot |
|
PR_Github #16497 [ run ] completed with state |
|
/bot run --reuse-test |
|
PR_Github #16617 [ run ] triggered by Bot |
|
PR_Github #16617 [ run ] completed with state |
…6976) Signed-off-by: Lizhi Zhou <[email protected]>
…VIDIA#6976) Signed-off-by: Lizhi Zhou <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
…VIDIA#6976) Signed-off-by: Lizhi Zhou <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
…VIDIA#6976) Signed-off-by: Lizhi Zhou <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
…VIDIA#6976) Signed-off-by: Lizhi Zhou <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
…VIDIA#6976) Signed-off-by: Lizhi Zhou <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
…VIDIA#6976) Signed-off-by: Lizhi Zhou <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
…VIDIA#6976) Signed-off-by: Lizhi Zhou <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
8 participants
…eadlock
Summary by CodeRabbit
Bug Fixes
Tests
Chores
Description
Test Coverage
GitHub Bot Help
/bot [-h] ['run', 'kill', 'skip', 'reuse-pipeline'] ...Provide a user friendly way for developers to interact with a Jenkins server.
Run
/bot [-h|--help]to print this help message.See details below for each supported subcommand.
run [--reuse-test (optional)pipeline-id --disable-fail-fast --skip-test --stage-list "A10-PyTorch-1, xxx" --gpu-type "A30, H100_PCIe" --test-backend "pytorch, cpp" --add-multi-gpu-test --only-multi-gpu-test --disable-multi-gpu-test --post-merge --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx" --detailed-log --debug(experimental)]Launch build/test pipelines. All previously running jobs will be killed.
--reuse-test (optional)pipeline-id(OPTIONAL) : Allow the new pipeline to reuse build artifacts and skip successful test stages from a specified pipeline or the last pipeline if no pipeline-id is indicated. If the Git commit ID has changed, this option will be always ignored. The DEFAULT behavior of the bot is to reuse build artifacts and successful test results from the last pipeline.--disable-reuse-test(OPTIONAL) : Explicitly prevent the pipeline from reusing build artifacts and skipping successful test stages from a previous pipeline. Ensure that all builds and tests are run regardless of previous successes.--disable-fail-fast(OPTIONAL) : Disable fail fast on build/tests/infra failures.--skip-test(OPTIONAL) : Skip all test stages, but still run build stages, package stages and sanity check stages. Note: Does NOT update GitHub check status.--stage-list "A10-PyTorch-1, xxx"(OPTIONAL) : Only run the specified test stages. Examples: "A10-PyTorch-1, xxx". Note: Does NOT update GitHub check status.--gpu-type "A30, H100_PCIe"(OPTIONAL) : Only run the test stages on the specified GPU types. Examples: "A30, H100_PCIe". Note: Does NOT update GitHub check status.--test-backend "pytorch, cpp"(OPTIONAL) : Skip test stages which don't match the specified backends. Only support [pytorch, cpp, tensorrt, triton]. Examples: "pytorch, cpp" (does not run test stages with tensorrt or triton backend). Note: Does NOT update GitHub pipeline status.--only-multi-gpu-test(OPTIONAL) : Only run the multi-GPU tests. Note: Does NOT update GitHub check status.--disable-multi-gpu-test(OPTIONAL) : Disable the multi-GPU tests. Note: Does NOT update GitHub check status.--add-multi-gpu-test(OPTIONAL) : Force run the multi-GPU tests in addition to running L0 pre-merge pipeline.--post-merge(OPTIONAL) : Run the L0 post-merge pipeline instead of the ordinary L0 pre-merge pipeline.--extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx"(OPTIONAL) : Run the ordinary L0 pre-merge pipeline and specified test stages. Examples: --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx".--detailed-log(OPTIONAL) : Enable flushing out all logs to the Jenkins console. This will significantly increase the log volume and may slow down the job.--debug(OPTIONAL) : Experimental feature. Enable access to the CI container for debugging purpose. Note: Specify exactly one stage in thestage-listparameter to access the appropriate container environment. Note: Does NOT update GitHub check status.For guidance on mapping tests to stage names, see
docs/source/reference/ci-overview.mdand the
scripts/test_to_stage_mapping.pyhelper.kill
killKill all running builds associated with pull request.
skip
skip --comment COMMENTSkip testing for latest commit on pull request.
--comment "Reason for skipping build/test"is required. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.reuse-pipeline
reuse-pipelineReuse a previous pipeline to validate current commit. This action will also kill all currently running builds associated with the pull request. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.