Details
-
Bug
-
Resolution: Fixed
-
Major
-
7.6.0
-
Untriaged
-
0
-
Unknown
Description
We are seeing regression starting from 7.6.0-1814 where failure detection time is 2.5 seconds more for both ns_server and disk failure tests.
http://showfast.sc.couchbase.com/#/timeline/Linux/reb/failover/all#reb_failover_100M_dgm_kv_hestia
Two commits went into build 1814:
Commit: e6995f8c4b8d8850bced3140c6db135c7b0df8c5
MB-58762: Update tick period when enabling auto_failover Instead of waiting til the auto_failover_cfg config change to do so.
Commit: 57d4a376899f1a6a8b994856966d00346f02b9d1
MB-58762: Pull common code enabling auto_failover into one fun We currently duplicate code for the two paths in which we initially enable auto_failover and update settings after it has already been enabled. Pull that code into one common function.
Jenkins jobs:
Last good build (7.6.0-1813 - https://perf.jenkins.couchbase.com/job/hestia/9848/console): 2.3s
Regressed build (7.6.0-1816 - https://perf.jenkins.couchbase.com/job/hestia/9847/console): 4.7s
Latest build (7.6.0-1825 - https://perf.jenkins.couchbase.com/job/hestia/9849/console): 4.7s