Implementing Modern DevOps

Monitoring Kubernetes clusters

Kubernetes covers 99 percent of the needs for any software company, but one part where it does not really shine is in embedded monitoring, leaving a space to be filled by third-parties. The main problem with Kubernetes comes from Docker: containers are ephemeral, so a common practice is to dump the logs into the standard output/error and use syslogd to gather them in a centralized location.

With Kubernetes, we have an added problem: the orchestrator on top of Docker needs to know how to fetch logs in order to make them available via the API or dashboard, so it is possible for the user to access them when required. But then there is another problem. Usually, logs are rotated on the basis of time and archived ...