Details
-
Bug
-
Resolution: Done
-
Critical
-
4.0.0
-
Security Level: Public
-
Build 4009
-
Untriaged
-
Unknown
Description
I ran the Q2 (singleton unique loop query) with GSI and stale=false and the indexer could not keep up with the updates per second done on the KV node. I changes the test to run with only 4.8k updates/sec and the indexer can still not keep up with the front-end workload (2.4k updates/sec worked fine though). See the test run with 4.8 updates/sec below.
http://ci.sc.couchbase.com/job/hades-dev/598/
For some reason when I run these tests and the query latency is really high the html reports don't get generated correctly. I can re-run the test if needed and you guys can see the behavior while the cluster is live.
Attachments
For Gerrit Dashboard: MB-15934 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
54101,2 | MB-15934 indexer: Change stable snapshot interval to 3s | master | ns_server | Status: ABANDONED | -2 | 0 |
54102,2 | MB-15934 indexer: Change default wal_size | unstable | indexing | Status: MERGED | +2 | +1 |
54172,1 | Revert "MB-15934 indexer: Change default wal_size" | unstable | indexing | Status: MERGED | +2 | +1 |