Logging in Docker

The more services you have, the more logging data you will have to manage. And with Docker it has become even simpler to run services.

Luckily, it comes with lots of options when it comes to managing your logging data. The main mean of controlling how logging works inside Docker is through so-called “logging drivers”. There are a couple of built-in drivers and since 17.05 you can also provide custom drivers through plugins.

By default, logs are written as JSON and can be accessed using the docker logs command for each container. If you want to use a journald installed on the host-system, you can do so by setting --log-driver journald when starting a container.

That’s probably also the most interesting option for the environment I’m using Docker in right now. The driver exposes a handful of fields with each log statement that help you to find them:

• CONTAINER_NAME
• CONTAINER_ID and CONTAINER_ID_FULL
• CONTAINER_TAG
• CONTAINER_PARTIAL_MESSAGE for handling large log statements

Now I could, for instance, get all the log statements of the “my-app” container using journalctl:

$ journalctl CONTAINER_NAME=my-app

There are also drivers available GELF, Syslog etc. You can find out more on docs.docker.com.

Custom logging plugin

In case there is a target system that is not yet supported by Docker, you can also create your own logging plugin which is specified again inside the docker documentation. Such plugin is itself just another Docker image that offers a simple HTTP API.

Disabling logging and more…

If you set --log-driver none for a container, you can suppress logging for that container. This might be useful for systems where just too much log data is generated and you cannot control it at the source.

In general, Docker offers pretty much anything you’d want when it comes to logging. Luckily, the documentation is well written. Now I just have to find out, what system I should reconfigure first 🙂