Monitoring Docker Containers with Grafana Alloy and cAdvisor

Short summary

A practical guide to monitoring Docker containers using cAdvisor scraped locally by Grafana Alloy, which forwards metrics via remote_write to a central Prometheus/Loki/Grafana stack. Covers cAdvisor deployment with read-only mounts, filtering unnamed cgroup series, and PromQL alert rules for restart loops, CPU spikes, and memory pressure. The author notes the hardest part is deciding which alerts matter, not collecting metrics.

• •Deploy cAdvisor as a privileged container with read-only mounts and scrape it locally via Alloy
• •Filter unnamed container series with {name!=""} to avoid dashboard clutter
• •Alert rules provided for restart loops, sustained high CPU, and container memory as percentage of host memory

Generated with AI, which can make mistakes.