You’re monitoring containers wrong
summary
Argues that the Docker stats API is too limited and that instrumenting every container adds prohibitive overhead, so a lightweight kernel module deployed as a monitoring container captures system calls and network activity with minimal cost and full visibility. Captures produce capture files that can be replayed and inspected via a curses UI to root-cause intermittent bugs, metadata from Kubernetes enables tier-level grouping, and built-in teams ACLs restrict sensitive views; experience covers hundreds of Docker customers.
AI-written summary · about
chapters (12)
- 0:00Start
- 2:30Breaking traditional model
- 2:53Traditional deployment
- 3:49The Docker revolution
- 4:35Containerized deployment
- 6:04Application workflow
- 6:53Container monitoring
- 7:42Metric collection
- 9:25Kernel instrumentation
- 10:55Information aggregation
- 16:51Real example
- 17:34Analysis & troubleshooting
related talks
37mYou’re Monitoring Kubernetes Wrong
44mThe Dark Art of Container Monitoring
29mKubernetes-Defined Monitoring [I]
50mMonitoring and metrology for containers
30mMix Kubernetes with Linux Kernel Tracing to Boost the Observability of Your Cluster
01h 00mMonitor your containers with the Elastic Stack
46mMonitoring and metrology for containers- 30mMix Kubernetes with Linux Kernel Tracing to Boost the Observability of Your Cluster