How to distinguish different nodes with cpu/mem

5/7/2019

Figure an EFK helm chart in a kubernetes cluster.

Fluentbit is configured to provide cpu and mem informations, it's very nice.

Data come from all nodes, but without any per-node distinguish info. So the graph gives an average for all machines:

beautifull but useless graph

Is there a way at any level to distinct nodes from each other?

At the fluentbit level, I've tried to use kubernetes filter without success, because it work only with tail filter, neither cpu nor mem.

Any argument appreciated (even "change your stack", or "go to prometheus")

-- Slim
elasticsearch
fluent-bit
kibana
kubernetes
kubernetes-helm

1 Answer

5/7/2019

I would definitely recommend using Prometheus and Grafana to see cpu/mem and many other monitoring graphs. A very good place to start with many builtin graphs and alerts is here: https://github.com/coreos/kube-prometheus

-- Vasily Angapov
Source: StackOverflow