ci(test): increase xdist workers from 2 to 8#1501
Closed
ramakrishnap-nv wants to merge 7 commits into
Closed
Conversation
|
Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually. Contributors can view more details about this message here. |
Collaborator
Author
|
/ok to test 98a15f1 |
ramakrishnap-nv
added
non-breaking
Collaborator
Author
|
/ok to test 8fcd4fe |
CI Test Summary3 failed · 28 passed · 0 skipped
|
Collaborator
Author
|
/ok to test 233d1c5 |
Collaborator
Author
|
/ok to test 25af2f2 |
Collaborator
Author
|
/ok to test 3029f52 |
bdice notes RAPIDS generally uses -n 6 or -n 8; safe max is determined by GPU memory usage, not CPU cores (target: 16GB GPU). Testing -n 8 to measure speedup vs stability tradeoff. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
-n 8 produced wrong solver results (Objective nan, Unknown status, inf bounds) due to GPU memory pressure. Stepping down to -n 6 which is the lower end of bdice's recommended RAPIDS range. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
…ion) -n 6 produced same Unknown status / NaN objective failures as -n 8 in TestCuoptCliCPUOnly remote tests. Trying -n 4 as next step down. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
StreamLogs did a single extra getline after detecting job completion, which silently dropped any lines flushed between the last EOF poll and the job-complete marker. With multiple final log lines (e.g. "Optimal solution found." followed by "Best objective ..."), only the first was sent to the client, leaving the rest unstreamed. Replace the single read with a drain loop so all buffered lines are sent before the job_complete sentinel. Fixes test_cli_lp_remote / test_cli_mip_remote flakiness under xdist where increased I/O activity widens the race window. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
stop_log_streaming() calls TryCancel() which races against the server's final log drain: poll_for_completion() returns as soon as the job is marked complete, then TryCancel() kills the stream before the server has finished sending remaining lines (e.g. "Best objective …"). Add drain_log_streaming(): polls log_stream_done_ (set when the thread exits naturally after receiving the job_complete sentinel) for up to 5s, then falls back to stop_log_streaming(). Use it on the success path in solve_lp and solve_mip so all final log lines are received before the stream is torn down. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
ramakrishnap-nv
force-pushed
the
ci/pytest-xdist-n8
branch
from
July 6, 2026 15:11
3029f52 to
113de2b
Compare
Collaborator
Author
|
/ok to test 113de2b |
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Collaborator
Author
|
/ok to test ee61fef |
-n 6 triggers cudaErrorInvalidDevice on single-GPU runners (L4, A100). Drop back to -n 4 which avoids device exhaustion. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Signed-off-by: Ramakrishna Prabhu <ramakrishnap@nvidia.com>
Collaborator
Author
|
/ok to test c889f62 |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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
Bumps
-n 2→-n 8in bothrun_cuopt_pytests.shandrun_cuopt_server_pytests.sh, following bdice's suggestion on #1486:Base infrastructure (pytest-xdist dependency, worker-aware gRPC/server ports) landed in #1486.
What to watch in CI
cudaErrorInvalidDeviceor OOM would indicate too much GPU pressure-n 2baseline (~3m 48s cuopt, ~5m 34s server)🤖 Generated with Claude Code