Details
-
Bug
-
Resolution: Fixed
-
Critical
-
7.6.0
-
Untriaged
-
0
-
Yes
Description
Compared to build 7.6.0-2005, GSI file-based rebalance out time is at least 10 times higher on 7.6 RC4 build (7.6.0-2119). Note that, we saw regression in file-based rebalance swap on RC3 build (7.6.0-2090) in MB-60672. The fix of MB-60672 has been merged to 7.6.0-2107.
Shard Rebalance-out, 1 bucket x 100M x 1KB, 2K KV ops/sec, 1000 indexes, 20% RR, Plasma, s=1 c=100
Build | Rebalance Time (mins) | Job |
---|---|---|
7.6.0-2005 | 5.7 | https://perf.jenkins.couchbase.com/job/aether/3838/ |
RC3 (7.6.0-2090) | 112.1 | https://perf.jenkins.couchbase.com/job/aether/3942/ |
RC4 (7.6.0-2119) | 166.9 | https://perf.jenkins.couchbase.com/job/aether/4003/ |
RC4 (7.6.0-2119) | https://perf.jenkins.couchbase.com/job/aether/4012/ |
I have re-submitted job 4012 to confirm the regression, and the regression is reproducible. I see the same behavior in job 4003 and 4012 that rebalance didn't make progress for hours.
Job 4003
2024-02-11T02:53:00 [INFO] Rebalance progress: 92.53257142857146 %
....
2024-02-11T05:31:09 [INFO] Rebalance progress: 92.53257142857146 %
Job 4012
2024-02-12T12:26:13 [INFO] Rebalance progress: 92.78714285714288 %
...
2024-02-12T14:17:13 [INFO] Rebalance progress: 92.78714285714288 %
Attachments
Issue Links
- relates to
-
MB-60900 High heap in use for indexer [CLONE of GSI file-based rebalance-out time is 10+ times more on RC4 build]
- Resolved