chrisjbawden/dependson repository overview

⁠dependson

A lightweight “sidecar” Docker container that watches a primary container and enforces ad‑hoc “depends on” behaviour - if your monitored container goes offline, any listed dependents are stopped (and prevented from starting until it’s back up).

⁠Features

• Watches Docker events (die / start) on your host’s daemon
• On startup: if the MONITOR container isn’t running, it immediately stops all DEPENDANT containers
• If MONITOR dies, waits a configurable grace period and then stops each DEPENDANT
• Any time a DEPENDANT tries to start while MONITOR is down, it’s shut down immediately
• Runs as PID 1 in its own container; supports Docker’s --restart policies

⁠Configuration

All other behaviour (listening to Docker events, parsing JSON) is built into the bundled monitor.sh.

⁠Usage

docker run -d \
  --name [container]-monitor \
  -v /var/run/docker.sock:/var/run/docker.sock \
  -e MONITOR="my-critical-container" \
  -e DEPENDANT="worker-1 worker-2 ui-service" \
  -e GRACE="10" \
  --restart unless-stopped \
  chrisjbawden/dependson:latest

-v /var/run/docker.sock:/var/run/docker.sock Grants the sidecar access to Docker events and CLI on the host.

-e MONITOR="…" Cofnigure which contained to monitor/watch

-e DEPENDANT="…" Configure which container/s to control in response to changes in the container being monitored.

--restart unless-stopped Ensures your watchdog will relaunch automatically if it crashes or the host reboots.

⁠Behaviour in action

1. On container start

• Sidecar checks if ${MONITOR} is running.
• If not, it stops any ${DEPENDANT} immediately.

1. When ${MONITOR} dies

• Emits a log message, waits ${GRACE} seconds.
• Stops each container in ${DEPENDANT}.

1. If someone (or Docker’s restart policy) tries to start a ${DEPENDANT}

• Sidecar detects the start event.
• If ${MONITOR} is still down, it shuts that container down again.

Development of this image can be found here - https://github.com/chrisjbawden/dependson⁠