morningman opened a new pull request, #63646:
URL: https://github.com/apache/doris/pull/63646
## Summary
Both tests have been muted on the External Regression pipeline due to
long-standing flakiness (analysis based on TeamCity build #92687 / id 953050).
Neither is a real product bug — both are test-side robustness issues.
### `test_file_cache_query_limit` (~50% pass rate)
After `POST /api/file_cache?op=clear&sync=true` the test waited exactly one
`file_cache_background_monitor_interval_ms` window and then asserted
`normal_queue_curr_size == 0` once. The counters surfaced by
`information_schema.file_cache_statistics` are republished by the background
monitor on its own cadence, so a single fixed-time wait races the refresh and
the assert fails roughly half the time even when the cache really is empty.
- Replace the four wait-then-assert blocks (`size == 0` after clear, `size >
0` after a query) with `Awaitility`-based polling (already imported) on the
relevant metric until the predicate holds, with a `max(30s, 6 ×
monitor_interval)` timeout.
- The original `assertFalse(...)` calls with their metric-specific messages
are kept as the final guard, so real failures still surface a precise reason.
- The two waits for BE config propagation (`enable_file_cache_query_limit`
flip) are left untouched — not in the failure path.
### `test_hive_query_cache` (~20–25% fail rate)
The `test { sql ...; time 20000 }` block at L122 ran TPC-H Q9 against
containerized hive parquet with `enable_sql_cache=false` set above, so the 20s
upper bound was timing a cold 6-table join, not a cache hit. The query
routinely exceeds 20s under cluster load.
- Drop the time guard; the `qt_tpch_1sf_q09` above already validates
correctness, and the cache behavior is exercised in the subsequent blocks that
explicitly enable sql cache.
## Test plan
- [ ] Run External Regression pipeline on this PR and confirm both cases
pass.
- [ ] After 5+ consecutive green runs, follow up to unmute these cases in
TeamCity.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
