Observability is a critical component of any serious Kubernetes-based platform. It is the only way to ensure the reliable operation of cloud-native applications and the rapid debugging by developers of the trickiest problems that only occur in the production environment.
The essential pillars of good observability are logs, metrics and traces. There are a large number of commercial tools and SaaS providers that support the aggregation and analysis of relevant diagnostic data.
In this talk, however, we use a stack built entirely on open source building blocks: Promtail to forward logs to Loki, Prometheus to collect metrics, and Tempo to receive traces. We will also show how the new Exemplars storage function enables the fast transition from metrics to traces and logs.