autoheal

Autoheal

Docker Autoheal is a Docker HEALTHCHECK monitor. Whereas Docker will already run the specified healthchecks and mark containers that fail them as unhealthy, it does not take any action upon that check. It would be nice if Docker would (optionally) restart unhealthy services automatically. While Docker Swarm has this functionality builtin, it is not enabled for normal non-swarm Docker daemons. Docker Autoheal is a separate service that fixes this missing functionality. This service binds to the Docker socket, and so it receives health notices for all containers, and can automatically restart those that are unhealthy, based upon configuration applied in those container's Docker labels.

Setup

make config

You must configure autoheal to tell it which of your containers it should watch:

• Set AUTOHEAL_CONTAINER_LABEL=all to watch ALL containers indiscriminantly (but only for containers that actually have a HEALTHCHECK).
• Otherwise set this to a specific label name, eg. AUTOHEAL_CONTAINER_LABEL=autoheal, and then only the containers with the label autoheal=true will be watched (set this label on any other container on your Docker host.).

Install

make install

Read the logs

The logs are very quiet: nothing is printed on startup unless there is an error. When a watched container becomes unhealthy, and if it is configured to be restarted, you will see a message like this when it is restarted:

autoheal-autoheal-1  | 21-02-2023 21:50:42 Container /unhealthy-example (3cd71f557bc7) found to be unhealthy - Restarting container now with 10s timeout

Example HEALTHCHECK

An example service is included for how to create a healthcheck and to test inducing an unhealthy state, and having autoheal restart it.