Fly-Style commented on code in PR #19188:
URL: https://github.com/apache/druid/pull/19188#discussion_r2986452340
##########
embedded-tests/src/test/java/org/apache/druid/testing/embedded/server/FaultyClusterTest.java:
##########
@@ -92,12 +93,21 @@ public void test_overlord_skipsCleanupOfPendingSegments()
cluster.callApi().postSupervisor(supervisorSpec);
final int recordCount = publish1kRecords(topic, true);
- waitUntilPublishedRecordsAreIngested(recordCount);
+ Assertions.assertEquals(expectedRecords, recordCount);
- cluster.callApi().postSupervisor(supervisorSpec.createSuspendedSpec());
+ waitUntilPublishedRecordsAreIngested(expectedRecords);
+
+ // Additionally, confirm that rows were placed into Druid.
+ // Caution: "ingest/rows/output" does not have a 'DATASOURCE' dimension.
+ indexer.latchableEmitter().waitForEventAggregate(
+ event -> event.hasMetricName("ingest/rows/output"),
Review Comment:
@gianm unfortunately, the intermediate stage of this test is flaky.
Specifically, `waitUntilPublishedRecordsAreIngested` check is passing, but
`verifyRowCount` check fails, suggesting that the `SELECT` query sees ~half of
the ingested rows. What's interesting - locally it passes even with processor
throttling I am using to detect flaky tests with timeouts.
Some conspiracy theory: after your reply, I am starting to have suspicions
that the flakiness may signalize that something is wrong with either metrics or
ingestion itself.
--
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]
