We explored quite a few topics that are beyond "normal" Kubernetes usage. We learned how to scale Pods using HorizontalPodAutoscaler. We saw that scaling Pods does not provide enough benefits if we cannot scale cluster nodes as well. We explored how to do just that using Cluster Autoscaler. Unfortunately, it is currently available only for AWS, GKE, and AKS.
While scaling Pods and nodes is essential, we had to gather metrics as well. They give us insights into the behavior of our clusters and applications running inside it. For that, we adopted Prometheus. More importantly, we saw how we can leverage Alertmanager to create notifications that will alert us when there is an issue, instead of staring at a screen waiting for a graph to reach a "red line".
We learned that gathering metrics from exporters might not be enough, so we instrumented our...