Details
-
Bug
-
Resolution: Duplicate
-
Major
-
3.0.0
-
None
-
1
Description
I opened another bug here as well: https://issues.couchbase.com/browse/MB-37888
Currently our pillowfight tests run with 3.0.0-alpha.5 and throughput seems right for 7.0.0-1180. However, after running experiments with libcouchbase 3.0.0 on the same build, we are seeing a steep drop in throughput from 200k ops to around 6k ops. This is with replicate_to=1 and persist_to=0. This throughput is similar to the persist_to=1 or 2 cases of this test.
alpha.5 good run, 7.0.0-1180:
http://perf.jenkins.couchbase.com/job/ares/12014/
3.0.0 bad run, 7.0.0-1180:
http://perf.jenkins.couchbase.com/job/ares/12300/