Details
-
Bug
-
Resolution: Duplicate
-
Critical
-
5.5.0
-
Untriaged
-
Yes
Description
While running rebalance related perf tests on my branch master toy build, I noticed some regressions in set and get latencies. This is compared with the ones reported by latest Vulcan run on show fast.
These regressions looked unrelated to my changes, so I rerun the test with a branch master build (without my ns_server changes). It shows the same regressions as seen below.
Link to the logs is available at the bottom of the console output page.
- set latency comparison: http://cbmonitor.sc.couchbase.com/reports/html/?snapshot=hestia_550-2807_rebalance_3c1e&snapshot=hestia_550-20000_rebalance_48f7&label=vulcan&label=master#6a41814c6439ddb2c676d7e3b479d3f3
- get latency comparison:http://cbmonitor.sc.couchbase.com/reports/html/?snapshot=hestia_550-2807_rebalance_3c1e&snapshot=hestia_550-20000_rebalance_48f7&label=vulcan&label=master#7a39345757f44ca44bd34b6af824b903
- Branch master run console output: http://perf.jenkins.couchbase.com/job/hestia/2658/console
- Vulcan run from show-fast: http://perf.jenkins.couchbase.com/job/hestia/2629/console
Attachments
Issue Links
- relates to
-
CCBC-939 Up to 25% drop in pillowfight throughput after upgrade to libcouchbase 2.9.0
- Open