add failure injector for monarch training script #270
Closed
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
meta-cla
bot
added
the
CLA Signed
d4l3k
approved these changes
Sep 25, 2025
| f"{self.uid} Injecting failure ({failure_type}) into random trainer" | ||
| ) | ||
|
|
||
| await self.failure_actors.fail.choose(failure_type) |
There was a problem hiding this comment.
.choose picks an arbitrary training?
There was a problem hiding this comment.
.choose picks an arbitrary training?
yup, choose will send it to one random trainer in the replica mesh
hm failing test is unrelated/flaky, I see it on #268 |
amirafzali
changed the title
|
@amirafzali has imported this pull request. If you are a Meta employee, you can view this in D83601242. |
Summary: This introduces a failure injector with 5 failure modes: - SEGFAULT: Triggers a SIGSEGV on the process - DEADLOCK = Deadlocks the GIL, resulting in ProcessGroupNCCL timeout and terminal failure - KILL_PROC: Immediately kills the process with non-zero exit code - COMMS = Forcefully aborts the ProcessGroup and NCCL communicator - KILL_SLURM = Kills a random replica SLURM job It can be enabled with the flag `--with--failure`, and it runs async every 120 seconds. Reviewed By: tushar00jain Differential Revision: D83601242 Pulled By: amirafzali
facebook-github-bot
force-pushed
the
add-failure-injector
branch
from
911abd4 to
fce2791
Compare
|
@amirafzali has exported this pull request. If you are a Meta employee, you can view the originating Diff in D83601242. |
|
@amirafzali merged this pull request in d596ec7. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
This introduces a failure injector with 5 failure modes:
It can be enabled with the flag
--with--failure, and it runs async every 60 seconds.