Alerting on failure
Two levels: wobbling (restarting, paused, health still starting) and down (exited or unhealthy).
Containers with state, health and restart count.
Docker monitoring
The agent reads the container list straight from the Docker socket — no second agent inside the container, no sidecar. The socket is mounted read-only and is explicitly opt-in.
- Access
- socket, read-only
- Opt-in
- yes
- Alert levels
- 2
What matters here
Per container
State, health check, restart count, image, CPU, memory and network throughput.
Off by default
Docker cannot say whether a stopped container was stopped on purpose. The rule is opt-in.
Honest about cost
Container CPU is derived from two agent samples instead of blocking the daemon for a second per container.
FAQ
Common questions about Docker monitoring
Is Docker monitoring enabled automatically?
No. You enable Docker monitoring per server because the socket grants broad access to the host.
Which container values are collected?
State, health check, restarts, image, CPU, memory and network throughput.
Why is a read-only socket still sensitive?
The Docker socket is inherently powerful. Uptimo calls this out explicitly and keeps the integration opt-in.