Debugging Kubernetes pods without centralized logs means SSHing into the master, running `kubectl get pods` and `kubectl logs` repeatedly, and wasting 30 minutes just to understand failures. With 10+ pods, this manual process becomes unbearable, highlighting the need for a centralized logging stack.

The solution combines Fluent Bit as a DaemonSet to collect logs from all pods, Loki to store them efficiently (like ElasticSearch but simpler), and Grafana for real-time search and visualization. This architecture provides immediate log access and historical data, replacing manual CLI digging with a unified web interface.

Deployment requires careful configuration: a StatefulSet for Loki to persist data, a DaemonSet for Fluent Bit with proper RBAC permissions to enrich logs with Kubernetes metadata, and a Grafana deployment with Loki pre-configured as a datasource. The author also shared an Ansible role for automated deployment and common pitfalls, like under-provisioning Loki's memory.