A long time ago (like 2 years) I had installed node exporter, prometheus, and grafana on my K3S cluster. This was useful to visualise resource usage in all my nodes, and served as a constant reminder of how underpowered the control plane was.
With my new cluster having run for a while, I would quite like to get this setup back, so I can avoid having to ssh into the servers and running top on them. This wasn't an urgent matter for me anyway as the nodes possess more than enough power for the few applications running, and my blog won't have a surge in popularity anytime soon.
Either way, I was pleasantly surprised to find that Prometheus has a helm chart with the whole stack combined, making running my monitoring super simple. That means just plugging in the helm repo and chart into ArgoCD, and clicking sync.
After a good few minutes, I have a massive number of resources running in my monitoring namespace and the following are just the "core" parts.
I created an ingress for Grafana, then logged onto the website. With the helm chart, a number of graphs are already created for me to monitor so many things including the nodes themselves and the containers residing in Kubernetes.
Finally, if you want, you can setup alerts to make sure you wake up at 2am when your servers are on fire, but I'll leave that for a future episode.