Course contentsShow
← Cloud and Infrastructure
Lesson 2049 of 2,98943. Observability - LoggingFree lesson

Kubernetes Logging Architecture

Understand how Kubernetes handles stdout/stderr, node-level logging, and cluster-level patterns.

Kubernetes Logging Architecture

What you'll learn: How Kubernetes collects, routes, and manages logs from containers and cluster components.

The Kubernetes Logging Model

Kubernetes doesn't provide a built-in logging solution, but instead offers logging architecture patterns you implement yourself. Think of it like plumbing: Kubernetes provides the pipes and outlets (stdout/stderr streams), but you choose the collection and storage system.

Application Logs: stdout and stderr

Containers should write logs to stdout (standard output) and stderr (standard error). Kubernetes automatically captures these streams. When you run kubectl logs <pod-name>, you're reading what the container wrote to stdout/stderr.

# View logs from a specific container
kubectl logs my-pod -c my-container

# Follow logs in real-time
kubectl logs -f my-pod

This works because the container runtime (like containerd or CRI-O) redirects these streams to files on the node.

Node-Level Logging

On each node, container logs are stored as files (typically in /var/log/containers/). The kubelet and container runtime manage log rotation to prevent disk overflow. However, when a pod is deleted, these logs disappear—they're ephemeral.

Cluster-Level Logging Patterns

For persistent logs, you need cluster-level logging:

  • Node-level agent: Deploy a DaemonSet (like Fluent Bit) that reads container logs from each node and ships them to a central system
  • Sidecar container: Add a logging container to pods that processes and forwards logs
  • Direct application export: Applications send logs directly to a logging backend

Most production clusters use the node-level agent pattern—it's resource-efficient and doesn't require modifying pod definitions.

Key Takeaway: Kubernetes captures container stdout/stderr automatically at the node level, but you must implement cluster-level logging (typically via DaemonSets like Fluent Bit) to aggregate, persist, and search logs centrally.