The problem
If you’re here, then you probably have a Node.js application running in Cluster mode, either through the native Node APIs or through a package manager like PM2. In this mode, however, there is usually a load balancer that switches between several child processes to do computational work. Each of these child processes has their own statistics for resource usage. If you’re using something like Prometheus, to collect custom metrics, they are also saved per process. This results in jagged or incorrect results, when trying to display and analyze the data in a tool such as Grafana. The question is, how do we collect all these metrics and aggregate them for easy consumtion at one place?