A couple of days ago we started seeing build-sanity failures on Centos7 only (
MB-20215 and MB-20216). This was strange because
- it passed on other platforms and
- same tests passed when QE ran them.
I was really not sure what was causing this, when Wayne mentioned yesterday that the machine that is hosting a bunch of build sanity VMs were consistently running at 100% CPU. This led me to investigate this a little further and it turns out that while running build 904 (or greater)the memcached process gets to 350-380% of CPU. And pretty much stays that way for the entirety of test suite.
I ran this multiple times on both centos and ubuntu (builds 903 vs 904) and was able to see memcached going crazy with build 904.
Changelog for build 904: http://172.23.120.24/builds/latestbuilds/couchbase-server/spock/904/CHANGELOG
- relates to
MB-20286 DCP observe regression between 4.5.1 build 2727 and 2756
|For Gerrit Dashboard: MB-20235|
|66147,2||MB-20235: Fix wake's incorrect work counting||master||ep-engine||Status: MERGED||+2||+1|