Hello. If I switch Kibana to cgroup stats with "xpack.monitoring.ui.container.elasticsearch.enabled: true" config, I see just 0% CPU usage for all elasticsearch nodes. Kibana, Logstash and Elasticsearch are running in docker containers, cgroup path overrides are in place for kibana (--cpu.cgroup.path.override=/ --cpuacct.cgroup.path.override=/) and elasticsearch (es.cgroups.hierarchy.override=/). Am I missing some config?
What version of the stack are you on?
5.5.0
If you run this in your Kibana Dev Tools > Console;
GET .monitoring-es-*/_mapping
and then see if that mapping includes cfs_quota_micros
Yes, there is:
...
"os": {
"properties": {
"cgroup": {
"properties": {
"cpu": {
"properties": {
"cfs_quota_micros": {
"type": "long"
},
...
Hi @vb4t,
Are you using any type of throttling within the ES containers? If not, then I suggest just disabling this setting (going back to the default) and using the process-level CPU percentage.
We are in the middle of updating our documentation, so this should hopefully be clearer in the future, but unless you have a quota and therefore throttle the process, then there is no advantage to monitoring Cgroup CPU usage.
Hope that helps,
Chris
Hello. Currently I do not use throttling. But in the near future I want to use it.
This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.