smengcl commented on code in PR #3981:
URL: https://github.com/apache/ozone/pull/3981#discussion_r1028726029
##########
hadoop-ozone/integration-test/src/test/java/org/apache/hadoop/ozone/freon/TestOMSnapshotDAG.java:
##########
@@ -194,27 +199,57 @@ void testZeroSizeKey()
DifferSnapshotInfo snap1 = getDifferSnapshotInfo(omMetadataManager,
volumeName, bucketName, "snap1");
- DifferSnapshotInfo snap3 = getDifferSnapshotInfo(omMetadataManager,
- volumeName, bucketName, "snap3");
+ DifferSnapshotInfo snap2 = getDifferSnapshotInfo(omMetadataManager,
+ volumeName, bucketName, "snap2");
// RocksDB does checkpointing in a separate thread, wait for it
final File checkpointSnap1 = new File(snap1.getDbPath());
GenericTestUtils.waitFor(checkpointSnap1::exists, 2000, 20000);
+ final File checkpointSnap2 = new File(snap2.getDbPath());
+ GenericTestUtils.waitFor(checkpointSnap2::exists, 2000, 20000);
+
+ List<String> actualDiffList21 = differ.getSSTDiffList(snap2, snap1);
+ LOG.debug("Got diff list: {}", actualDiffList21);
+ // Hard-coded expected result.
+ // The result is deterministic. Retrieved from a run.
+// final List<String> expectedDiffList21 = Arrays.asList(
+// "000066", "000080", "000087", "000073", "000095");
+// Assertions.assertEquals(expectedDiffList21, actualDiffList21);
Review Comment:
I'm going to remove this. It appears that once I set the DB column family
write buffer size to a small enough value like we have here, the flush and
compaction start to behave slightly differently on different architectures. It
gives a different set of result in the CI (x86, iirc) than on my Mac (arm64).
This indicates some other factors might be impacting the list of SSTs
generated here now that write buffer size might no longer be the dominant
factor triggering compactions. Could be some default RocksDB tunable
differences for each platform? (This might be possible as each platform does
have its own native JNI lib in the rocksdbjni jar.) Or simply because of the
detected system memory / CPU core causing some relevant defaults to be
different.
At least it seems the result here is stable if the environment stays the
same. Two runs from a previous commit:
https://github.com/smengcl/hadoop-ozone/actions/runs/3502672241/jobs/5867215890#step:5:3353
https://github.com/apache/ozone/actions/runs/3502681795/jobs/5867234494#step:5:3350
```
Error: TestOMSnapshotDAG.testZeroSizeKey:218 expected: <[000066, 000080,
000087, 000073, 000095]> but was: <[000067, 000060]>
```
--
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]
