[https://nvbugs/5508267][fix] Proper handling of inactive canceled requests #9280
Conversation
Signed-off-by: thorjohnsen <[email protected]>
📝 WalkthroughWalkthroughModified canceled request handling in PyExecutor to use set-based lookups for O(1) performance, introduced Changes
Sequence Diagram(s)sequenceDiagram
participant Executor
participant CanceledHandling as _handle_canceled_requests
participant CancelLogic as _try_cancel_request
participant Queue as executor_request_queue
Executor->>CanceledHandling: Process canceled requests
CanceledHandling->>CanceledHandling: Build set of canceled IDs (O(1))
loop For each canceled request
CanceledHandling->>CancelLogic: Attempt cancellation
alt Cancellation succeeds
CancelLogic-->>CanceledHandling: Success
CanceledHandling->>CanceledHandling: Mark CANCELLED, reset decoding_iter
else Cancellation fails
CancelLogic-->>CanceledHandling: Pending
CanceledHandling->>CanceledHandling: Record in still_pending_canceled_ids
end
end
alt enable_attention_dp is false
CanceledHandling->>Queue: Preserve only still_pending_canceled_ids
else enable_attention_dp is true
CanceledHandling->>Queue: Clear all canceled IDs
end
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes
Pre-merge checks and finishing touches❌ Failed checks (1 inconclusive)
✅ Passed checks (2 passed)
✨ Finishing touches
🧪 Generate unit tests (beta)
Tip 📝 Customizable high-level summaries are now available in beta!You can now customize how CodeRabbit generates the high-level summary in your pull requests — including its content, structure, tone, and formatting.
Example instruction:
Note: This feature is currently in beta for Pro-tier users, and pricing will be announced later. 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: 2
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
tensorrt_llm/_torch/pyexecutor/py_executor.py(1 hunks)
🧰 Additional context used
🧠 Learnings (1)
📚 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/pyexecutor/py_executor.py
🧬 Code graph analysis (1)
tensorrt_llm/_torch/pyexecutor/py_executor.py (1)
tensorrt_llm/_torch/pyexecutor/executor_request_queue.py (2)
get_canceled_req_ids(762-763)clear_canceled_req_ids(765-766)
🔇 Additional comments (2)
tensorrt_llm/_torch/pyexecutor/py_executor.py (2)
1770-1771: Good optimization: set-based lookup for O(1) performance.Converting the canceled request IDs list to a set improves membership testing from O(n) to O(1), which is beneficial when checking many active requests.
1776-1777: Improved clarity: explicit early exit for non-canceled requests.The explicit check and early
continuefor non-canceled requests improves code readability and performance by avoiding unnecessary processing.
| still_pending_canceled_ids = [] | ||
| for request in self.active_requests: | ||
| req_id = request.py_request_id if not request.is_child else request.parent_request_id | ||
| if req_id in self.executor_request_queue.get_canceled_req_ids(): | ||
| if req_id not in canceled_req_ids: | ||
| continue | ||
|
|
||
| is_cancelled = self._try_cancel_request(request) | ||
| if is_cancelled: | ||
| # Mark requests as finished, then, we reuse all existing code | ||
| # to clean up the KV cache resources. | ||
| request.finish_by_reason(FinishReason.CANCELLED) | ||
| request.decoding_iter = request.py_decoding_iter | ||
| else: | ||
| still_pending_canceled_ids.append(req_id) |
There was a problem hiding this comment.
🧩 Analysis chain
Critical: Undefined method _try_cancel_request called.
The method _try_cancel_request is invoked at line 1779 but is not defined in this file. This will cause an AttributeError at runtime when a canceled request is encountered.
Run the following script to verify whether this method exists elsewhere in the codebase:
If the method doesn't exist, it needs to be implemented in this PR. The method should:
- Accept a
LlmRequestparameter - Return a boolean indicating whether cancellation succeeded
- Handle any state-specific logic that might prevent immediate cancellation
🏁 Script executed:
#!/bin/bash
# Search for _try_cancel_request definition or import
ast-grep --pattern $'def _try_cancel_request($$$) {
$$$
}'
# Also search with ripgrep to catch any references
rg -n '_try_cancel_request' --type=pyLength of output: 111
---
🏁 Script executed:
```shell
#!/bin/bash
# Find the class definition for py_executor to check parent classes
rg -n 'class.*PyExecutor' --type=py -A 5
# Check if there's any method definition containing 'cancel' that might be related
rg -n 'def.*cancel' --type=py -C 2
Length of output: 7568
Critical: Method _try_cancel_request is undefined and must be implemented.
The method self._try_cancel_request(request) is called at line 1779 but does not exist anywhere in the codebase. Verification confirms:
- No definition exists in PyExecutor or any parent class
- No imports provide this method
- This will cause an
AttributeErrorat runtime when a cancellation is attempted
The method needs to be implemented in PyExecutor. It should accept a LlmRequest parameter and return a boolean indicating whether cancellation succeeded.
| else: | ||
| # Only keep active requests that did not cancel in canceled req ids list | ||
| self.executor_request_queue.canceled_req_ids = still_pending_canceled_ids |
There was a problem hiding this comment.
🧩 Analysis chain
Verify direct assignment to canceled_req_ids.
The core fix for preventing unbounded growth looks correct: by retaining only still_pending_canceled_ids (requests that couldn't be canceled yet), the list won't accumulate indefinitely. However, ensure that directly assigning a Python list to executor_request_queue.canceled_req_ids is the correct approach.
Run the following script to verify the type and usage of canceled_req_ids:
Based on the relevant code snippet showing get_canceled_req_ids() returns List[int] and clear_canceled_req_ids() clears the list, direct assignment should work. However, consider whether calling a setter method would be more maintainable.
🏁 Script executed:
#!/bin/bash
# Find the definition and type of canceled_req_ids in ExecutorRequestQueue
ast-grep --pattern $'class ExecutorRequestQueue {
$$$
}'
# Check how canceled_req_ids is initialized and used
rg -n 'canceled_req_ids' tensorrt_llm/_torch/pyexecutor/executor_request_queue.py -A 3 -B 3Length of output: 1270
Now let me search for any cached references to canceled_req_ids in the py_executor.py file:
Now let me search for any potential cached references to canceled_req_ids in the py_executor.py file:
Based on my analysis of the code and Python semantics, I can now generate the rewritten review comment:
Use in-place mutation instead of direct assignment for canceled_req_ids.
Direct assignment (=) replaces the reference, breaking any other code that holds a reference to the original list object. In-place mutation keeps the same list object while updating its contents.
The class has a clear_canceled_req_ids() method (line 765-766) that mutates the list in-place, suggesting the API is designed to preserve the list object identity. For consistency and safety, replace line 1795:
self.executor_request_queue.canceled_req_ids = still_pending_canceled_idswith:
self.executor_request_queue.canceled_req_ids.clear()
self.executor_request_queue.canceled_req_ids.extend(still_pending_canceled_ids)This ensures any code that previously called get_canceled_req_ids() and cached the reference will see the updated contents.
🤖 Prompt for AI Agents
In tensorrt_llm/_torch/pyexecutor/py_executor.py around lines 1793-1795, the
code assigns a new list to self.executor_request_queue.canceled_req_ids which
replaces the list object and breaks any external cached references; modify it to
mutate the existing list in-place by clearing the current list and extending it
with still_pending_canceled_ids so the original list identity is preserved and
all cached references see the updated contents.
Signed-off-by: thorjohnsen <[email protected]>
|
/bot run |
|
PR_Github #25135 [ run ] triggered by Bot. Commit: |
Signed-off-by: thorjohnsen <[email protected]>
|
/bot run |
|
PR_Github #25139 [ run ] triggered by Bot. Commit: |
|
PR_Github #25135 [ run ] completed with state |
2 participants
Summary by CodeRabbit
Bug Fixes
Performance
Description
A client observed a gradual slowdown of server speed during normal operation. The slowdown was root caused to failed requests, each failed request was added to a list of canceled request ids and never removed from it, hence the list of canceled request ids would grow longer and longer with each failed request. The implemented fix is to retain only active requests that have not been canceled yet at the end of _handle_canceled_requests(...).
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.
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.