I’m running docker containers for all side projects. Usually one single container. Sometimes multiple containers started by docker-compose[1].
If containers get issues, I want to get alerts.
What a typical monitoring requirement! But if you do some research, you will find the information is just overwhelming. cAdvisor, Prometheus, InfluxDB, etc. Excuse me? Can’t we have a simple solution for this simple requirement? Here is my answer. Try it and discuss with me.
This solution works for docker-compose deployment. Please check Prometheus[2], for more complex docker environments.
BASIC IDEA
Firstly I admit: Prometheus can do the job. Yes, it’s capable, but …
Hint: Want to use prometheus in a relatively easy way, try dockprom [3]
Frankly speaking, learning Prometheus took much longer than what I had expected. And it’s a bit over-killing for my scenario. Prometheus solution itself will start multiple containers. But my envs only have very few containers to monitor.
Ironically “docker ps -a” can easily answer most of my questions.
root@denny-blog:~# docker ps -a CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES 2936e1d7ec34 denny_wordpress "/bin/sh -c /docke..." 3 weeks ago Up 13 days (healthy) 0.0.0.0:443->443/tcp blog-wordpress 87665b7a76df mysql:5.7 "docker-entrypoint..." 3 weeks ago Up 3 weeks 3306/tcp blog-mysql 4aa821fa9d37 denny/ss:v2 "ssserver -c /etc/..." 3 weeks ago Up 3 weeks (healthy) 0.0.0.0:6187->6187/tcp shadowsock ea96f4dbefaa ubuntu:14.04 "/bin/bash" 14 seconds ago Exited (0) 6 seconds ago nginx-test
- To get stopped container: just check “STATUS” of each container.
- To get unhealthy containers: check status for “unhealthy” output.
Hint: To enable docker healthcheck[4], you need update your Dockerfiles properly
- Send alerts, whenever any of above occurrences have happened. Should be easy, Right?
So how about we start a container, which keeps polling status of all containers. Then send out alerts, if necessary?
You may think: we won’t be able to run “docker ps” in containers. It only runs in docker host.
Yes, you’re right. But there is one way out. You can query status via docker unix socket file. Try below locally, you will understand.
docker run -v /var/run/docker.sock:/var/run/docker.sock \
-t -d --privileged --name test-socket \
--entrypoint=/bin/sh byrnedo/alpine-curl
docker exec -it test-socket \
curl -XGET --unix-socket /var/run/docker.sock \
http://localhost/containers/json
HOW TO USE
Check code in Github. Follow @dennyzhang
Try it, if you find it’s useful. And discuss with me.
This is very useful. I will implement on my system and update soon – working on it in some time.
Ritu, that’s great!
Share me with your feedback or experience, once you have the chance to try this approach.