github-actions[bot] commented on code in PR #65226:
URL: https://github.com/apache/doris/pull/65226#discussion_r3533198228
##########
regression-test/suites/audit/test_audit_log_queue_time.groovy:
##########
@@ -110,10 +118,7 @@ suite("test_audit_log_queue_time", "nonConcurrent") {
auditResult = sql "${query}"
}
- auditResult.each { row ->
- assertTrue(row[1] >= sqlSleepTime * 1000)
- }
-
+ log.info("audit queue time result for marker ${testMarker}:
${auditResult}")
assertTrue(auditResult.size() >= queuedSqlCnt)
Review Comment:
This now only checks that at least one matching audit row has a positive
`queue_time_ms`, but that no longer proves the query actually waited behind the
`sleep(5)` query. `queue_time_ms` starts when `QueueToken` is constructed and
ends in `complete()`; `QueryQueue.getToken()` also calls `complete()`
immediately for the first admitted query, so a tiny positive value is enough to
satisfy the `where queue_time_ms > 0` filter. A regression that records only a
few milliseconds for the queued worker would still pass here. Since the new
latches synchronize both workers before issuing the `select`, the test can keep
a meaningful lower bound for the queued row, for example by checking the
maximum `queue_time_ms` is close to `sqlSleepTime * 1000` with a reasonable
tolerance or by measuring the observed admission interval.
--
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]
