Deprecate/remove xpack.monitoring.ui.container.elasticsearch.enabled setting?

[ycombinator]

Today, users must set the xpack.monitoring.ui.container.elasticsearch.enabled setting in their kibana.yml if their Elasticsearch nodes are running in a containerized environment. Otherwise CPU stats in the Monitoring UI are not shown correctly.

I'm wondering if we can discover that Elasticsearch is running in a containerized environment from data in .monitoring-es-* itself and therefore get rid of this kibana.yml setting.

[elasticmachine]

Pinging @elastic/stack-monitoring

[cachedout]

I think @pickypg might have some thoughts here.

[pickypg]

I'd love to see us do that. We discussed doing something like this awhile ago, like from Metricbeat data just having it report the data correctly from the onset and avoiding the setting altogether (8.0 presumably).

The biggest issue with automatically determining it at query time is that:

1. Per node page, it's not so bad -- it's one query and adjustment.
2. Per the node listing page, you have to be sure that it's the same across all nodes because it's possible, albeit rare, for a cluster to have some nodes getting CPU throttled containers and some not.

[ycombinator]

like from Metricbeat data just having it report the data correctly from the onset

Meaning, this is not reported today by the internal node_stats ES collector?

[pickypg]

Exactly. The internal ES node stats report the CPU as a system value and, separately, as the Cgroup value.