[fix](regression) fix test_fe_cached_partition_version#65158
Conversation
### What problem does this PR solve?
Issue Number: None
Related PR: None
Problem Summary: The cloud_p0/version test_fe_cached_partition_version case expected the follower FE to keep reading the cached Beijing partition version immediately after an insert when the cache TTL was 1000 ms. In ASAN docker runs, a single insert can take longer than 1000 ms, so the cache can expire naturally before the following select and the follower correctly reads the fresh version. Keep a large TTL around the immediate stale-cache assertions, then switch back to the short TTL before checking expiration refresh. The case also keeps the table version cache TTL aligned with the partition version cache TTL.
### Release note
None
### Check List (For Author)
- Test: Regression test
- ./run-regression-test.sh --run -d cloud_p0/version -s test_fe_cached_partition_version
- Behavior changed: No
- Does this need documentation: No
|
Thank you for your contribution to Apache Doris. Please clearly describe your PR:
|
|
run buildall |
|
/review |
There was a problem hiding this comment.
Pull request overview
This PR updates a cloud regression suite to reduce flakiness around FE cached table/partition version TTL behavior (notably in slower ASAN docker clusters), by explicitly controlling version-sync behavior and aligning both table + partition cache TTL settings throughout the test.
Changes:
- Disable
cloud_enable_version_syncervia docker FE configs for this suite’s cluster. - Set
cloud_table_version_cache_ttl_msalongsidecloud_partition_version_cache_ttl_msat each phase of the test. - Adjust TTL sequencing to keep caches alive while inserts run, avoiding timing-related flakes.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
There was a problem hiding this comment.
Automated review summary for PR 65158:
I reviewed the single changed file, regression-test/suites/cloud_p0/version/test_fe_cached_partition_version.groovy, plus the surrounding FE cloud version-cache/session-variable paths and docker cluster setup. I did not find any additional inline issue to raise in this pass. I also refreshed live PR comments before submission; the currently open Copilot comments already cover the pre-existing 1}) insert literal concern, so I did not duplicate it.
Critical checkpoint conclusions:
- Goal/test coverage: The patch is focused on stabilizing the cloud docker version-cache regression by disabling FE version syncer and keeping table and partition cache TTLs aligned. The test still exercises follower stale-cache reads and expiration refresh.
- Scope: The change is small and limited to one regression test.
- Concurrency/lifecycle/config compatibility: No new concurrency or lifecycle behavior is introduced. The new FE config is cluster-start configuration, and the TTL changes use existing session/global variables consumed by the cache paths.
- Parallel paths: Both table and partition cache TTLs are now set together at the relevant phases.
- Testing/validation: I verified GitHub PR metadata/patch, live PR comments, surrounding session/cache code, and
git diff --checkfor the changed file. I did not run the docker regression in this checkout because.worktree_initialized,thirdparty/installed, andthirdparty/installed/bin/protocare missing. - User focus: No additional user-provided focus points.
|
run nonConcurrent |
|
run buildall |
No description provided.