Admiral Lighthouse is (aimed to be) an extremely lightweight alternative to other container monitoring tools. Requires Python 3.5. Built with aiohttp.
Essentialy, the monitoring part is realized by parsing the output of docker stats (Docker Engine >=13.0.0). It thus fully relies on the accuracy of docker-provided metrics about its own (running) containers. It can be thought of as an extension to Docker Remote API which, as of time of writing, lacks support for some of the nifty novelities introduced with Docker Engine version 13.0.0.
Each time docker stats outputs new data (every second), the stats of all running containers get summed up and put put on the queue. Future version (if any), will provide some additional aggregation tools, i.e. average the results over arbitrary timeframes.
On top of that, you can also interact with container's shell via docker exec command. The official docker API client for Python falls short of providing
an async way to communicate with the pseudo-tty. By parsing
It sports a simple web server listening on port 1988.
# build the image
docker build --rm -t admiral_lighthouse:latest .
# run it
docker run -d \
-v $(which docker):/bin/docker \
-v /var/run/docker.sock:/var/run/docker.sock \
-p 1988:1988 \
--rm \
superfunnel/admiral_lighthouse
# query it
curl http://0.0.0.0:1988/statwill query the server for the queued usage record, returning a JSON response, specifically a list of the following:
- summed cpu usage from all containers [%]
- memory usage [KB]
- incoming network traffic [KB]
- outcoming network traffic [KB]
- block writes [KB]
- block reads [KB]
- a total number of processes