fix: log multirun failures to job files

[suzunn]

Problem

Failed jobs in a basic multirun sweep were captured in JobReturn without being written while the job logging configuration was active. This left per-job log files empty and made failures other than the first one difficult to diagnose.

Solution

I added a focused file-handler path that records the active exception in the job log without emitting another traceback to the console. I also extended the partial-failure regression test and added a bugfix news fragment.

Fixes #3215

Testing

• .venv/Scripts/python -m pytest tests/test_basic_sweeper.py tests/test_basic_launcher.py tests/test_callbacks.py -q (42 passed)
• Black, isort, Flake8, and Pyrefly checks for the touched Python files
• git diff --check

[meta-cla]

Hi @suzunn!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at cla@meta.com. Thanks!

[omry]

Thanks. overall looks good. I'll take a closer look when it hits my queue.
In the mean time CLA please.

[Copilot]

Pull request overview

This PR ensures that exceptions raised during a basic multirun sweep are logged while the per-job logging configuration is still active, so each failed job’s *.log file contains actionable failure details.

Changes:

• Add a small internal helper in run_job() to emit the active exception to file handlers only (avoiding extra console tracebacks).
• Extend the partial-failure multirun regression test to assert the failed job log contains the exception.
• Add a bugfix news fragment for the user-visible behavior change.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
hydra/core/utils.py	Logs job exceptions to per-job file handlers inside the job’s logging context before storing the exception in JobReturn.
tests/test_basic_sweeper.py	Extends the multirun partial-failure test to validate per-job log content for the failing job.
news/3215.bugfix	Documents the bugfix in release notes.

[omry]

Thanks, this is in the right direction.
Please add coverage for the case where multiple jobs fail in the same multirun, and verify that each failed job’s own log file contains the traceback. The original issue is not only that one failed job lacks a log, but also that failures are captured and surfaced later, making later failed jobs hard to diagnose.

[omry]

Also, please accept the CLA, I cannot accept the change with it.

[omry]

Hi, are you planning to finish this?

[suzunn]

Yes — sorry for the delay. Just pushed an update:

• Rebased onto latest main (resolved the conflict with the new test_glob_uses_primary_config_searchpath test).
• Added test_multiple_failures_log_to_own_job_files: a multirun where two jobs fail with different errors (ZeroDivisionError and TypeError) and one succeeds. It verifies each failed job's own log file contains its traceback, and the successful job's log does not.

Both tests pass locally. Will sort out the CLA next.

[omry]

Thanks for following up and adding the multiple-failure coverage. This now covers the case I was concerned about. The change looks good.