Kibana to run more than one process using PM2


(Colin Cunningham) #1

Hi. Has anyone successfully run Kibana using PM2 or have recommendations to do so for a system w/, say, k hardware threads available? I have a 96vCPU VM available.

Thanks. - Colin


(Christian Dahlqvist) #2

The Kibana server process is typically not very CPU intensive, so I have no idea how it would ever need or be able to use such a large number of CPU cores. What does your overall cluster architecture look like?


(Colin Cunningham) #3

Christian. While I have 96vCPUs (48 cores) on the server, it supports the other components of the ELK stack as well. We are using it to support analysis of data collected on other VMs in the cloud from which we'll pipe the data in using Kafka. The cluster architecture is AWS, this is on a r5d instance. You're right that it shouldn't take too much compute--I don't want to give it all to it. At the same time, it should be able to perform should multiple users connect to Kibana. For that, or to support a longer time span queries, we thought it might benefit from separate processes.
I was also looking for a recommendation on setting --max-old-space-size= as a relative share of overall memory the system has available. We have it at 16GB but could allocate quite a bit more if that is thought to make kibana more resilient.

Thanks, - Colin


(Christian Dahlqvist) #4

When you are querying lots of data, the Kibana server process does very little of the work. The vast majority of work is done by Elasticsearch, which will require much more resources and generally be the component limiting performance.


(system) closed #5

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.