Docker - diagnosing issues

Diagnosing issues with Docker

Sometimes, things go awry. Fact. Happily, there’s an arsenal of tools available to help figure out what is going down with Docker, and although some of them are slightly cryptic it’s possible to get a vast body of data and information on the images, containers, and the environment in the mighty quest to make things run smoothly, or at the very least run correctly.

The first thing to do is to collect some basic information from Docker and the containers it is running.

docker version

This will tell you what version of Docker you are using, both client version and server (daemon) version. It will also tell you if you can connect - it fails if there’s a connection problem.

docker inspect

‘docker inspect’ is a powerful tool. It produces massive amounts of information on a container. In some ways, too much, which is why it’s important to learn how to find useful values from the output.

How to inspect specific values

The output from ‘docker inspect’ can be somewhat overwhelming. The in-built Go template system allows the extraction of a single key. For example, is the last run container still running?

docker inspect --format '{{.State.Running}}' $(docker ps -lq)  

How to find the path to the underlying volume

This example also inspects the last run container, but instead of run state it tells you the mapped volumes.

docker inspect --format '{{.Volumes}}' $(docker ps -lq)  

docker events

Docker will stream events in a format that tells you when it creates, starts, stops, and destroys containers and images. In it’s simplest form, it’s invoked without any options:

docker events 

It also keeps a history and can filter the output. To get all events since the epoch, use this command:

docker events --since="0" 

To filter the output to show container starts only:

docker events --since="0" --filter="event=start" 

This is a useful tool to track interactions with Docker and find out what is going on behind the curtain.

docker logs

Docker offers a mechanism to get the logs of the running container. Many applications take advantage of the fact that by default this is the stdout and stderr streams, making it trivial to send output to the Docker daemon for capture and later analysis. A lot of services do this by default, such as Apache Tomcat, and it’s an option in the logging framework log4j for example.

A better Docker ps format

If you pipe the Docker ps output to ‘less -S’ it prevents it from wrapping the table rows.

docker ps -a | less -S  

Watching the Docker logs

When using the ‘docker logs’ commaind, you can ask it to follow the output in the same way as the tail command using -F :

docker logs <containerid> -F  

docker exec

Docker has the facility to run additional processes in the container using ‘docker exec’ eliminating sshd installations from containers. This can invoke an interactive shell to inspect the container and be used to fix problems such as changing setting or adjusting log levels.