[PG NCCL] catch cuda lib runtime error - driver shutting down #74258
Conversation
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
Summary: There is a case when PG cleanup thread checks cuda event status after cuda runtime library has been unloaded. When that happens, it would leads to a "driver shutting down" error. This issue usually happens when cuda API is called in global or static object destructor. Test Plan: wait for user Differential Revision: D34904896 fbshipit-source-id: a2846050f0f7b37742a9e0d79e13f3b7b05d1fad
mingzhe09088
requested review from
H-Huang,
mrshenli,
pritamdamania87,
rohan-varma and
zhaojuanmao
as code owners
CI Flow Status⚛️ CI FlowRuleset - Version:
|
Contributor
🔗 Helpful links
💊 CI failures summary and remediationsAs of commit 2f4ab77 (more details on the Dr. CI page):
🕵️ 2 new failures recognized by patternsThe following CI failures do not appear to be due to upstream breakages:
|
facebook-github-bot
added
oncall: distributed
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D34904896 |
osalpekar
approved these changes
Mar 17, 2022
facebook-github-bot
pushed a commit
that referenced
this pull request
Mar 18, 2022
Summary: Pull Request resolved: #74258 There is a case when PG cleanup thread checks cuda event status after cuda runtime library has been unloaded. When that happens, it would leads to a "driver shutting down" error. This issue usually happens when cuda API is called in global or static object destructor. Test Plan: wait for user Reviewed By: jiayisuse, osalpekar Differential Revision: D34904896 fbshipit-source-id: 705c0812132dae97ea55fcb22730557880ca35e1
Contributor
|
Hey @mingzhe09088. |
| } | ||
| } | ||
| } catch (const std::exception& e) { | ||
| if (std::string(e.what()).find("driver shutting down") == std::string::npos) { |
Contributor
There was a problem hiding this comment.
This seems really fragile, is there a better way to detect this. Maybe using cudaGetLastError?
Contributor
Author
There was a problem hiding this comment.
Since the cuda library has been unloaded at this point, so any cuda runtime API would fail with "driver shutting down" error. If there is a way to prevent cuda library from unloading, that would solve the issue here.
| if (std::string(e.what()).find("driver shutting down") == std::string::npos) { | ||
| throw; | ||
| } | ||
| LOG(INFO) << "[Rank " << rank_ |
Contributor
There was a problem hiding this comment.
Shouldn't this be LOG(WARNING) or LOG(ERROR)?
Contributor
Author
There was a problem hiding this comment.
Sure, which one do you prefer?
| if (std::string(e.what()).find("driver shutting down") == std::string::npos) { | ||
| throw; | ||
| } | ||
| LOG(INFO) << "[Rank " << rank_ |
Contributor
There was a problem hiding this comment.
Why do we swallow the exception and just continue here? Isn't this misleading to the user where they think the operation has completed but it actually has not?
Contributor
Author
There was a problem hiding this comment.
When this happens, the operation has completed so user code is not affected. We just want PG to peacefully exit without aborting.
shahofblah
pushed a commit
that referenced
this pull request
Mar 25, 2022
Summary: Pull Request resolved: #74258 There is a case when PG cleanup thread checks cuda event status after cuda runtime library has been unloaded. When that happens, it would leads to a "driver shutting down" error. This issue usually happens when cuda API is called in global or static object destructor. Test Plan: wait for user Reviewed By: jiayisuse, osalpekar Differential Revision: D34904896 fbshipit-source-id: 705c0812132dae97ea55fcb22730557880ca35e1 (cherry picked from commit ecb5f14)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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.
Summary: There is a case when PG cleanup thread checks cuda event status after cuda runtime library has been unloaded. When that happens, it would leads to a "driver shutting down" error. This issue usually happens when cuda API is called in global or static object destructor.
Test Plan: wait for user
Differential Revision: D34904896