We have a 6 node cluster (1 ingest, 5 data, 8G jvm heap each)
Sometimes we have a peak cpu in the data nodes, going to 100% cpu and the ingest node refuse to process more querys.
One of the node has a GC at that time:
It's possible that just one GC in one node peaks up the rest of the nodes cpus? (not seems logical to me)
How can I solve the issue? more ram to that node could help?