We need monitoring outside the cluster for visibility both before it starts and when the cluster completely breaks. But since this setup precludes prometheus from talking directly to pods or services, we miss out on many of the useful features of prometheus's kubernetes support. We should consider running a prometheus instance inside the cluster too in order to collect metrics that make more sense from that perspective.
We need monitoring outside the cluster for visibility both before it starts and when the cluster completely breaks. But since this setup precludes prometheus from talking directly to pods or services, we miss out on many of the useful features of prometheus's kubernetes support. We should consider running a prometheus instance inside the cluster too in order to collect metrics that make more sense from that perspective.