Skip to content

Distribute large integration test files across workers - #10761

Merged
GarrettBeatty merged 1 commit into
build-perf-sphinx-parallelfrom
build-perf-integ-parallel
Oct 22, 2026
Merged

GarrettBeatty merged 1 commit into
build-perf-sphinx-parallelfrom
build-perf-integ-parallel

Conversation

@GarrettBeatty

@GarrettBeatty GarrettBeatty commented Oct 15, 2026 •

Copy link
Copy Markdown
Contributor

Description of changes:

Part 2 of 3 (stacked on #10760) of the CI build-perf changes from #10759. Speeds up the integration tests (the Catapult TestSource step).

With --dist=loadfile, customizations/test_generatecliskeleton.py (38,716 offline tests, ~40 min) and customizations/s3/test_plugin.py (~16 min) each ran on a single worker, so adding workers did not help.

  • scripts/ci/run-tests: add a --dist option (default unchanged: loadfile).
  • scripts/ci/run-integ-tests: run with --dist loadgroup. tests/integration/conftest.py groups tests per file as before, except the two large offline files (test_generatecliskeleton.py, botocore/test_utils.py), which are distributed per test, and s3/test_plugin.py, which is distributed per class.

Testing:

Measured on a c5.9xlarge with 8 workers (together with #10762): 42:45 → 11:37, same results before and after (58,564 passed; the same 2 failures in both runs are specific to the EC2 test environment). End-to-end dry run in a Catapult dev stack: TestSource 48.1 → 17.2 min compared with prod medians.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

@GarrettBeatty
GarrettBeatty added this pull request to stack #10763 October 15, 2026 14:29
@GarrettBeatty
GarrettBeatty marked this pull request as ready for review October 15, 2026 14:44
@GarrettBeatty
GarrettBeatty requested a review from a team as a code owner October 15, 2026 14:44
Add a --dist option to scripts/ci/run-tests and run the integration
tests with --dist loadgroup. Tests are grouped per file as before,
except the large offline skeleton tests, which are distributed per
test, and the S3 plugin tests, which are distributed per class.
@GarrettBeatty
GarrettBeatty force-pushed the build-perf-integ-parallel branch from 93b3608 to 18ba913 Compare October 22, 2026 16:52
@GarrettBeatty
GarrettBeatty merged commit bbcbab0 into feature/build-perf Oct 22, 2026
108 of 111 checks passed
GarrettBeatty added a commit that referenced this pull request Oct 23, 2026
* Parallelize docs builds with sphinx-build -j (#10760)

* Parallelize docs builds with sphinx-build -j

* Declare the guzzle theme parallel-safe and build sitemap.xml from the
  environment so it is complete and ordered like a serial build when
  writing in parallel.
* Add the parallel_chunks extension, which reads with one interleaved
  chunk per process so the main process does not repeatedly unpickle
  large environments, and restores the serial document order.
* Add the parallel_write extension, which loads, resolves and indexes
  doctrees in the worker processes and merges search index entries
  back in chunk order.

With sphinx-build -j 16 this builds the docs in ~9.5 minutes instead of
~57 minutes, with output identical to a serial build.

* Fail unit tests when Sphinx is upgraded without revalidation

The parallel docs extensions patch Sphinx internals, so add a unit test
that fails when the Sphinx version in requirements-docs-lock.txt no
longer matches the version they were validated against, with steps to
confirm parallel and serial builds produce identical output.

Also restore the release version in doc/source/conf.py, which was
accidentally reverted.

* Distribute large integration test files across workers (#10761)

Add a --dist option to scripts/ci/run-tests and run the integration
tests with --dist loadgroup. Tests are grouped per file as before,
except the large offline skeleton tests, which are distributed per
test, and the S3 plugin tests, which are distributed per class.

* Speed up slow S3 integration tests (#10762)

* Fix test_can_delete_single_sse_c_object waiting ~100s for the
  object_exists waiter to fail.
* Don't retry connection errors in test_verify_endpoint_url_is_used,
  which uses an endpoint with nothing listening.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants