For many versions now I've noticed an issue with my 2-node cluster. After 2 weeks of running CPU spikes from about 30% to 70% and will remain there even if all query load is removed. This happens on both nodes in the cluster. This cluster is only getting input from one logstash server that gets about 1-1.5 million documents inserted per day with minimal read queries. Each node in the cluster is hosted on an ESXi server. Configuration for each node:
OS: Ubuntu 14.04
CPU: 2x2.9Ghz
RAM: 12GB of RAM.
/etc/defaults/elasticsearch:
ES_HEAP_SIZE=8g
MAX_LOCKED_MEMORY=unlimited
DATA_DIR=/var/lib/elasticsearch
PID_DIR=/var/run
/etc/elasticsearch/elasticsearch.yml:
cluster.name: redacted
path.data: /var/lib/elasticsearch
gateway.type: local
Thanks,
Ryan