[#10374][fix] fixed race condition in AutoDeploy's mp tests port acquisition #10366
Conversation
📝 WalkthroughWalkthroughThe changes introduce explicit port-aware, retry-capable distributed initialization with inter-process synchronization primitives for PyTorch setup, while removing four helix-related fake operation registrations from the custom operators module. Changes
Sequence Diagram(s)sequenceDiagram
participant R0 as Rank 0 Process
participant R1 as Other Ranks
participant SP as shared_port<br/>(mp.Value)
participant PB as port_ready_barrier<br/>(mp.Barrier)
participant DPG as dist.init_<br/>process_group
rect rgb(220, 240, 255)
Note over R0,DPG: Retry Loop (up to max_retries)
end
loop Retry attempt
R0->>R0: Try binding port
R0->>DPG: _try_init_process_group(port)
alt Success
DPG-->>R0: True
R0->>SP: Write port to shared_port
R0->>PB: Wait on port_ready_barrier
Note over R0: Rank 0 bound successfully
else Port in use
DPG-->>R0: False (warning logged)
R0->>R0: Continue loop
end
end
rect rgb(240, 255, 240)
Note over R0,PB: Synchronization Phase
end
par Rank 0 Path
R0->>PB: Signal barrier (port ready)
and Other Ranks Path
R1->>PB: Wait on port_ready_barrier
PB-->>R1: Barrier released
R1->>SP: Read port from shared_port
R1->>DPG: init_process_group(port)
DPG-->>R1: Success
end
R0->>PB: Barrier exit
Note over R0,R1: All ranks initialized with same port
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes The distributed initialization changes introduce multi-branch retry logic, inter-process synchronization coordination, and modified control flow across spawning and initialization paths. The fake op removals require verification that no external code depends on these registrations. Heterogeneous changes demand separate reasoning for synchronization correctness and removal impacts. Pre-merge checks and finishing touches❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✨ Finishing touches
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. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
tensorrt_llm/_torch/custom_ops/cpp_custom_ops.py (1)
755-775: Remove commented-out code instead of leaving it in place.These fake-op registrations appear to be intentionally disabled. Commented-out code is a maintenance burden and should be deleted. If these ops are no longer needed, remove them entirely. If they might be needed later, version control history will preserve them.
Additionally, these changes seem unrelated to the PR objective (fixing port acquisition race condition). Consider splitting unrelated changes into a separate PR for cleaner review history.
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
tensorrt_llm/_torch/auto_deploy/distributed/common.pytensorrt_llm/_torch/custom_ops/cpp_custom_ops.py
🧰 Additional context used
📓 Path-based instructions (2)
**/*.py
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.py: Code developed for TensorRT-LLM should conform to Python 3.8+
Indent Python code with 4 spaces. Do not use tabs
Always maintain the namespace when importing in Python, even if only one class or function from a module is used
Python files should use snake_case naming:some_file.py
Python classes should use PascalCase naming:class SomeClass
Python functions and methods should use snake_case naming:def my_awesome_function():
Python local variables should use snake_case naming:my_variable = ...
Python variable names that start with a number should be prefixed with 'k':k_99th_percentile = ...
Python global variables should use upper snake_case with prefix 'G':G_MY_GLOBAL = ...
Python constants should use upper snake_case naming:MY_CONSTANT = ...
Avoid shadowing variables declared in an outer scope in Python
Initialize all externally visible members of a Python class in the constructor
For Python interfaces that may be used outside a file, prefer docstrings over comments
Python comments should be reserved for code within a function, or interfaces that are local to a file
Use Google style docstrings in Python for classes and functions, which can be parsed by Sphinx
Python attributes and variables can be documented inline with type and description
Avoid using reflection in Python when functionality can be easily achieved without reflection
When using try-except blocks in Python, limit the except to the smallest set of errors possible
When using try-except blocks in Python to handle multiple possible variable types (duck-typing), keep the body of the try as small as possible, using the else block for logic
Files:
tensorrt_llm/_torch/custom_ops/cpp_custom_ops.pytensorrt_llm/_torch/auto_deploy/distributed/common.py
**/*.{cpp,h,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
All TensorRT-LLM Open Source Software code should contain an NVIDIA copyright header that includes the year of its latest meaningful modification
Files:
tensorrt_llm/_torch/custom_ops/cpp_custom_ops.pytensorrt_llm/_torch/auto_deploy/distributed/common.py
🧠 Learnings (5)
📓 Common learnings
Learnt from: lirundong
Repo: NVIDIA/TensorRT-LLM PR: 9725
File: tensorrt_llm/_torch/custom_ops/cuda_tile_custom_ops.py:110-178
Timestamp: 2025-12-12T10:07:36.866Z
Learning: In PyTorch custom operators registered with torch.library.custom_op, mutable operators that return None and specify mutates_args do NOT require a register_fake decorator. The mutation tracking is handled automatically without needing a FakeTensor kernel, as documented in the PyTorch tutorial on mutable Python custom operators.
📚 Learning: 2025-12-12T10:07:31.564Z
Learnt from: lirundong
Repo: NVIDIA/TensorRT-LLM PR: 9725
File: tensorrt_llm/_torch/custom_ops/cuda_tile_custom_ops.py:110-178
Timestamp: 2025-12-12T10:07:31.564Z
Learning: In PyTorch custom operators registered with torch.library.custom_op, mutable operators that return None and specify mutates_args do not require a register_fake decorator. Mutation tracking is handled automatically without needing a FakeTensor kernel. This applies to Python custom op definitions in tensorrt_llm/_torch/custom_ops that use mutates_args and return None; verify you are not relying on register_fake in these cases.
Applied to files:
tensorrt_llm/_torch/custom_ops/cpp_custom_ops.py
📚 Learning: 2025-08-27T16:22:10.695Z
Learnt from: Fridah-nv
Repo: NVIDIA/TensorRT-LLM PR: 7227
File: tensorrt_llm/_torch/auto_deploy/utils/quantization_utils.py:94-100
Timestamp: 2025-08-27T16:22:10.695Z
Learning: When there are inconsistent operator detection methods (like custom_op() vs target_op()), removing one method and standardizing on the other is often cleaner than supporting both methods simultaneously.
Applied to files:
tensorrt_llm/_torch/custom_ops/cpp_custom_ops.py
📚 Learning: 2025-11-14T11:22:03.729Z
Learnt from: nzmora-nvidia
Repo: NVIDIA/TensorRT-LLM PR: 9163
File: tensorrt_llm/_torch/auto_deploy/custom_ops/quant.py:107-113
Timestamp: 2025-11-14T11:22:03.729Z
Learning: In TensorRT-LLM AutoDeploy custom ops, when adding hardware capability checks to select between kernel implementations (e.g., cuBLAS vs. CUDA kernel), use descriptive variable names that identify the specific GPU architectures or families being targeted (e.g., `is_blackwell_geforce_or_ada`) rather than generic names like `enable_cuda_core`. This makes it clear that the code is selecting an implementation path based on hardware capabilities, not enabling/disabling hardware features.
Applied to files:
tensorrt_llm/_torch/custom_ops/cpp_custom_ops.py
📚 Learning: 2025-08-19T12:45:11.997Z
Learnt from: amitz-nv
Repo: NVIDIA/TensorRT-LLM PR: 7033
File: tensorrt_llm/_torch/pyexecutor/model_engine.py:0-0
Timestamp: 2025-08-19T12:45:11.997Z
Learning: In tensorrt_llm/_torch/pyexecutor/model_engine.py, DoRA (Delta Orthogonal Rank Adaptation) functionality was removed from the PyTorch flow to eliminate issues with inverted DoRA detection logic. The original is_dora condition was checking if scaling_vec_pointer == 0, which was potentially incorrect.
Applied to files:
tensorrt_llm/_torch/custom_ops/cpp_custom_ops.py
🧬 Code graph analysis (1)
tensorrt_llm/_torch/auto_deploy/distributed/common.py (1)
tensorrt_llm/_utils.py (1)
get_free_port(477-478)
🪛 Ruff (0.14.10)
tensorrt_llm/_torch/auto_deploy/distributed/common.py
142-142: Consider moving this statement to an else block
(TRY300)
211-211: Avoid specifying long messages outside the exception class
(TRY003)
217-217: Avoid specifying long messages outside the exception class
(TRY003)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
- GitHub Check: Pre-commit Check
🔇 Additional comments (5)
tensorrt_llm/_torch/auto_deploy/distributed/common.py (5)
88-103: LGTM!Clean signature update that forwards the new synchronization parameters to
initialize().
127-148: Good implementation of retry-capable initialization.The helper correctly catches port-in-use errors and returns False for retry logic. The exception string matching is reasonable given PyTorch doesn't expose specific exception types for this case.
Minor: Static analysis suggests moving the success
return Trueto anelseblock for the try statement (TRY300), but this is stylistic and doesn't affect correctness.
225-233: Existing path retained for OMPI/torchelastic/single-process cases.This path correctly relies on the launcher (OMPI/torchelastic) having already set the required environment variables. No issues for the intended use cases.
244-261: LGTM!Properly forwards synchronization primitives to
initialize_or_skip.
300-323: Good use of explicit spawn context for synchronization primitives.Using
mp.get_context("spawn")ensures thatValueandBarrierobjects are properly compatible with spawned processes. The synchronization pattern is correct: rank 0 attempts port binding and always signals via the barrier—settingshared_port.valueto the successful port or-1on failure—while other ranks wait at the barrier before reading the shared port value and raising an error if rank 0 failed.
|
/bot run |
|
PR_Github #30289 [ run ] triggered by Bot. Commit: |
|
PR_Github #30289 [ run ] completed with state
|
MrGeva
changed the title
|
/bot run --add-multi-gpu-test |
|
PR_Github #30325 [ run ] triggered by Bot. Commit: |
|
PR_Github #30325 [ run ] completed with state
|
|
/bot run --add-multi-gpu-test --disable-fail-fast |
|
PR_Github #30336 [ run ] triggered by Bot. Commit: |
galagam
left a comment
galagam
left a comment
There was a problem hiding this comment.
LGTM, added some small comments.
|
PR_Github #30336 [ run ] completed with state
|
Signed-off-by: Eran Geva <[email protected]>
Signed-off-by: Eran Geva <[email protected]>
Signed-off-by: Eran Geva <[email protected]>
Signed-off-by: Eran Geva <[email protected]>
MrGeva
force-pushed
the
egeva/fix_race_port
branch
from
b085bfb to
91e2b27
Compare
|
/bot run --add-multi-gpu-test |
MrGeva
changed the title
|
PR_Github #30348 [ run ] triggered by Bot. Commit: |
|
PR_Github #30348 [ run ] completed with state
|
3 participants
Problem
Failure seen in CI:
https://prod.blsm.nvidia.com/sw-tensorrt-top-1/blue/organizations/jenkins/LLM%2Fmain%2FL0_Test-x86_64-Multi-GPU/detail/L0_Test-x86_64-Multi-GPU/3663/pipeline/3347
The spawn_multiprocess_job function had a TOCTOU (Time-Of-Check-To-Time-Of-Use) race condition when selecting ports for distributed process group initialization. The original flow was:
Between steps 1 and 3, another process could claim the port, causing EADDRINUSE errors and test failures in CI.
Solution
Implemented a retry mechanism with shared state synchronization:
Rank 0 retries: If init_process_group fails with EADDRINUSE, rank 0 picks a new port and retries (up to 5 attempts)
Shared port communication: Use multiprocessing.Value to share the working port with other ranks
Barrier synchronization: Use multiprocessing.Barrier so other ranks wait until rank 0 successfully binds
Explicit spawn context: Use mp.get_context("spawn") to create synchronization primitives, ensuring they work correctly with spawned processes
Summary by CodeRabbit
✏️ Tip: You can customize this high-level summary in your review settings.
Description
Test Coverage
PR Checklist
Please review the following before submitting your PR:
PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.
PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.
Test cases are provided for new code paths (see test instructions)
Any new dependencies have been scanned for license and vulnerabilities
CODEOWNERS updated if ownership changes
Documentation updated as needed
Update tava architecture diagram if there is a significant design change in PR.
The reviewers assigned automatically/manually are appropriate for the PR.
Please check this after reviewing the above items as appropriate for this PR.
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.
Details
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.