Dear community,

we are investigating use of ElasticSearch Cluster Monitor connector for monitoring of our Opensearch cluster.

It seems to be working fine except the CPU utilization monitoring.

We run our 3 openesearch nodes as containers. We monitor the container CPU utilization on host OS level and with above mentioned connector too. The results differ significantly. When an indexing DB demanding operation is run, we see 100% CPU utilization on OS level, but only couple % (e.g. 3%) in the DMA. This disproporition is present in an idle state as well (0% in DMA while some percentage in OS).

Looking closer inside the opensearch container I found there are 3 opensearch processes running:

Note, the %CPU in the screenshot above is an long term average, not the current usage. Taking into consideration this fact and the values of %CPU it seems the ElasticSearch connector is monitoring the 2nd process. But, the real work is done by the 3rd process, which is also obvious by the command it represents (the real database configuration).

Is there a way to choose the process in the Opensearch instance (be it bare metal, VM or container) to be monitored by the Dataminer ElasticSearch connector?