container-slayer is a tool for restarting unhealthy Docker containers.

Installation & Usage

Docker container

Important: you need to mount the Docker socket inside the running container.

docker run --name slayer -v /var/run/docker.sock:/var/run/docker.sock --restart always --detach nsheridan/container-slayer

Install from source

go install nsheridan.dev/container-slayer@latest

Configuration

Configuration is controlled by flags:

% container-slayer -h
Usage of container-slayer:
  -filter string
    	Only restart containers with this label. 'all' means all containers will be considered. (default "all")
  -interval duration
    	Interval between checks (default 1m1s)
  -socket string
    	Path to Docker socket (default "/var/run/docker.sock")
  -timeout duration
    	Time to wait for Docker to respond (default 30s)
  -unhealthy_count int
    	Number of consecutive unhealthy probes before restarting the container (default 3)

Usage

container-slayer only operates on containers which have a HEALTHCHECK set in the image.

labels

By default container-slayer will consider any unhealthy container for restart. Add labels to your containers to restrict slayer's operations to a set of containers:

docker run -p 80:80 --name httpbin --label safe-to-reap kennethreitz/httpbin

Then use the -filter flag to limit restarts to just those labeled containers.

container-slayer -filter safe-to-reap

or

docker run --name slayer -v /var/run/docker.sock:/var/run/docker.sock --restart always --detach nsheridan/container-slayer -filter safe-to-reap