Developing With Docker - The Debug Container

December 9, 2015 · docker developing-with-docker

One technique that’s proven to be useful over and over again when working with Docker is a Debug Container. The debug container isn’t a proper application, nor is it a one off job. It’s simply used for debugging problems.

Lets have a look at a common problem.

Problem

You want to connect to get a URL from an application listening on a port, but no connected containers or hosts have curl installed.

Solution

Create a Debug Container and link it to the problem container.

I’m going to assume the container running the app is called foo here, but change it to whatever you want. Firstly, create the following Dockerfile:

FROM centos:centos7

RUN yum install -y curl

CMD ["/bin/sleep", "9999"]

Build and tag the image:

docker build -t debug:1 .

Now we can test it out to run a one off command:

docker run -ti --rm --link foo:foo debug:1 curl -v http://foo

Discussion

You can use the image in a number of ways. You can start it up in the background, name it, and then use docker exec:

docker run --rm --link foo:foo --name debug debug:1
docker exec debug curl http://foo

You can run bash in the container:

docker run --rm -ti --link foo:foo debug:1 bash
$ curl -v http://foo
$ exit

Or use docker exec:

docker run --rm --link foo:foo --name debug debug:1
docker exec -ti debug bash
$ curl...

When you’re pairing, we call this “docker bash” :)

I hope you’ve gotten the outline. Here are some thing to consider:

I’ve used CentOS as my base image, naturally you can use whatever you’re comfortable with.
You probably want to install more packages. For example, I find telnet and nslookup useful.
Creating the debug container can be done just when you need it, but you’ll save time by doing it up front.
The sleep command keeps the conntainer running, but doing nothing.
The --rm option to docker run makes sure the container is removed when no longer in use, this prevent resource leakage.
This can also be used in a cluster. As the container is within the cluster’s subnet, it’ll be subject to all the networking rules and restrictions a normal container would be.

More posts about developing with Docker.