[joshuarli]

i've spent some effort getting the per-shard overhead down as much as humanly possible

with that, i'm comfortable sharding out more now

backend shards 22 -> 30 gets us ~25% faster on the full suite

will followup with adjusting selective testing to rightsize

---

Increase backend test shards from 22 to 30 and fix the shard balance data.

• .github/workflows/backend.yml static default shards 22 -> 30
• .github/workflows/scripts/calculate-backend-test-shards.py MAX_SHARDS 22 -> 30
• Refreshed backend-test-durations.json from the 30-shard colored full run (40755937953): old weights overestimated total test time 1.7x and missed new files, producing a 294-715s actual spread against 860s predictions.

[mchen-sentry]

Should be fine, but can we explain why we're splitting the files when there's a planner that breaks them apart already?

[github-actions]

Backend Test Failures

Failures on b1097f8 in this run:

tests/sentry/db/models/test_slug_concurrency.py::ConcurrentSlugTest::test_teams_get_different_slugs — log

[gw0] linux -- Python 3.13.17 /home/runner/work/sentry/sentry/.venv/bin/python3
tests/sentry/db/models/test_slug_concurrency.py:119: in test_teams_get_different_slugs
    first, second = create_concurrently(
tests/sentry/db/models/test_slug_concurrency.py:77: in create_concurrently
    raise errors[0]
tests/sentry/db/models/test_slug_concurrency.py:63: in run_second
    results["second"] = create()
                        ^^^^^^^^
tests/sentry/db/models/test_slug_concurrency.py:121: in <lambda>
    lambda: Team.objects.create(organization=organization, name="Platform"),
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
src/sentry/silo/base.py:157: in override
    return original_method(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
.venv/lib/python3.13/site-packages/django/db/models/manager.py:87: in manager_method
    return getattr(self.get_queryset(), name)(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
src/sentry/silo/base.py:157: in override
    return original_method(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
.venv/lib/python3.13/site-packages/django/db/models/query.py:665: in create
    obj.save(force_insert=True, using=self.db)
src/sentry/silo/base.py:157: in override
    return original_method(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
src/sentry/models/team.py:156: in save
    self._save_team(
src/sentry/models/team.py:167: in _save_team
    save_with_snowflake_id(
src/sentry/utils/snowflake.py:59: in save_with_snowflake_id
    with enforce_constraints(transaction.atomic(using=router.db_for_write(type(instance)))):
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
/opt/hostedtoolcache/Python/3.13.17/x64/lib/python3.13/contextlib.py:148: in __exit__
    next(self.gen)
src/sentry/db/postgres/transactions.py:115: in enforce_constraints
    with transaction:
         ^^^^^^^^^^^
src/sentry/testutils/hybrid_cloud.py:229: in new_atomic_exit
    maybe_flush_commit_hooks(connection)
src/sentry/testutils/hybrid_cloud.py:220: in maybe_flush_commit_hooks
    connection.run_and_clear_commit_hooks()
src/sentry/testutils/pytest/stale_database_reads.py:130: in run_and_clear_commit_hooks
    return old_run_and_clear_commit_hooks(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
.venv/lib/python3.13/site-packages/django/db/backends/base/base.py:769: in run_and_clear_commit_hooks
    func()
src/sentry/hybridcloud/models/outbox.py:236: in _drain_shard_with_metrics
    self.drain_shard()
src/sentry/hybridcloud/models/outbox.py:424: in drain_shard
    in_test_assert_no_transaction(
... (3 more lines)

[joshuarli]

Should be fine, but can we explain why we're splitting the files when there's a planner that breaks them apart already?

i've walked it back; it didn't make a difference and was cherry picked from an older experiment of mine