Since upgrading to elasticsearch v5.3.2, we see some performance degradation (+20ms on fulltext and +40ms on morelike, +2ms on comp suggest). This reboot also activated some new JVM options, to align them on the standard elasticsearch configuration. More context can be found on this upstream ticket. The problem seems to be related to the activation of Concurrent Mark and Sweep GC.
Description
Description
Related Objects
Related Objects
Event Timeline
Comment Actions
Removing UseConcMarkSweepGC seems to work on relforge1001. young gen size is back to something more reasonable.
Comment Actions
Change 358383 had a related patch set uploaded (by Gehel; owner: Gehel):
[operations/puppet@production] elasticsearch: remove UseConcMarkSweepGC