New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[C10D] Fix nccl flightrecorder ignored dump timeout #118142
Conversation
Don't call future.get() unless it's ready, because it waits. Also, refactor the code a bit for simplicity. We should do a follow-on PR to clean up the timeouts further, but this should fix the glaring timeout bug. [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/118142
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (2 Unrelated Failures)As of commit 449e97f with merge base 0528602 (): FLAKY - The following job failed but was likely due to flakiness present on trunk:
BROKEN TRUNK - The following job failed but was present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
Don't call future.get() unless it's ready, because it waits. Also, refactor the code a bit for simplicity. We should do a follow-on PR to clean up the timeouts further, but this should fix the glaring timeout bug. ghstack-source-id: 70ff3ea3185f8b085917dccabadd706839bddf80 Pull Request resolved: #118142
Don't call future.get() unless it's ready, because it waits. Also, refactor the code a bit for simplicity. We should do a follow-on PR to clean up the timeouts further, but this should fix the glaring timeout bug. cc mrshenli pritamdamania87 zhaojuanmao satgera rohan-varma gqchen aazzolini osalpekar jiayisuse H-Huang kwen2501 awgu penguinwu fegin XilunWu wanchaol fduwjj wz337 tianyu-l yf225 [ghstack-poisoned]
Don't call future.get() unless it's ready, because it waits. Also, refactor the code a bit for simplicity. We should do a follow-on PR to clean up the timeouts further, but this should fix the glaring timeout bug. ghstack-source-id: 1b649b3586d377d18eb652d9a0e56ff7657c3338 Pull Request resolved: #118142
@@ -1012,29 +1012,31 @@ void ProcessGroupNCCL::waitForDumpOrTimeout( | |||
TORCH_CHECK(fut.valid(), "Expected a valid future"); | |||
|
|||
auto futStatus = fut.wait_for(std::chrono::seconds(timeout_sec)); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I am not sure the default times are set properly, e.g., the default timeout_sec is 30s, so we wait for a max of 30s for local dump here, which should long pass the wakeUpTime point (default 2 secs after init dumping)?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yea, we need to clean up the time variables.
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Don't call future.get() unless it's ready, because it waits. Also, refactor the code a bit for simplicity. We should do a follow-on PR to clean up the timeouts further, but this should fix the glaring timeout bug. Pull Request resolved: #118142 Approved by: https://github.com/shuqiangzhang ghstack dependencies: #118044, #118046, #118047
Don't call future.get() unless it's ready, because it waits. Also, refactor the code a bit for simplicity. We should do a follow-on PR to clean up the timeouts further, but this should fix the glaring timeout bug. Pull Request resolved: #118142 Approved by: https://github.com/shuqiangzhang ghstack dependencies: #118044, #118046, #118047
Don't call future.get() unless it's ready, because it waits. Also, refactor the code a bit for simplicity. We should do a follow-on PR to clean up the timeouts further, but this should fix the glaring timeout bug. Pull Request resolved: #118142 Approved by: https://github.com/shuqiangzhang ghstack dependencies: #118044, #118046, #118047
Don't call future.get() unless it's ready, because it waits. Also, refactor the code a bit for simplicity. We should do a follow-on PR to clean up the timeouts further, but this should fix the glaring timeout bug. Pull Request resolved: #118142 Approved by: https://github.com/shuqiangzhang ghstack dependencies: #118044, #118046, #118047
Don't call future.get() unless it's ready, because it waits. Also, refactor the code a bit for simplicity. We should do a follow-on PR to clean up the timeouts further, but this should fix the glaring timeout bug. Pull Request resolved: #118142 Approved by: https://github.com/shuqiangzhang ghstack dependencies: #118044, #118046, #118047
Stack from ghstack (oldest at bottom):
Don't call future.get() unless it's ready, because it waits.
Also, refactor the code a bit for simplicity.
We should do a follow-on PR to clean up the timeouts further, but this
should fix the glaring timeout bug.
cc @mrshenli @pritamdamania87 @zhaojuanmao @satgera @rohan-varma @gqchen @aazzolini @osalpekar @jiayisuse @H-Huang @kwen2501 @awgu @penguinwu @fegin @XilunWu @wanchaol @fduwjj @wz337 @tianyu-l @yf225