Jvm GC overhead

(Frédéric Renan) #1


I've just start a new cluster with following specs :

Under VMWARE, dispatched on 2 datacenters :
2 data / master / ingest nodes : 16Go RAM / 8Go Heap / 4 vCPUs
2 master nodes : 8Go RAM / 2Go Heap / 2 vCPUs

With very few documents i'm starting to see GC overheap messages :

[2017-09-01T12:11:11,231][INFO ][o.e.m.j.JvmGcMonitorService] [xxxxxxxxxxx] [gc][62272] overhead, spent [311ms] collecting in the last [1s]

the request is a simple discover over 2 days (2 indices) from kibana for approx. 400000 hits.

the heap is

I'm aware that GC troubles can't find a simple answer, but I'm very surprised to see them come very soon in my integration process.

What kind of measurements could be significant for this cluster ?
Is it a "classic" behaviour ?
Is there particularities around clusters on VMware VMs ?

(andy_zhou) #2

use GCG1 not CMS try.

(Frédéric Renan) #3

Not recommended by Elastic

(system) #4

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.