How to restart an existing Docker container in restart="always" mode?

Question

When you initially run a Docker container from an image you can specify the option:

--restart="always" 

This ensures that the container is always restarted by the Docker daemon if for some reason it stops. So you could run a container like so:

docker run --restart="always" <IMAGE> 

Also you can restart an existing Docker container by specifying its container ID, i.e.:

docker start <CONTAINER ID> 

However I can't determine if it's possible to change an existing container, that originally was not run with the --restart="always" option, to convert it to always restart in future.

Currently the only way I can think to do this is to save the container as a new image and then run that image as a new container with the --restart="always" option. Would this in fact be the correct way to do this?

EDIT: What I perhaps didn't make clear enough originally is that I am thinking about the situation where there have been changes in the container since it was originally run, which need to be persisted. So just running a new container from the original image would not be sufficient.

Answer 1

We now have docker update, which allows changing the restart policy of a running container.

docker update --restart=always <CONTAINER ID> 

There are three other options:

• no (default)
• on-failure
• unless-stopped

Please refer to the link for details.

Answer 2

Ok, so to answer my own question, it seems that it's not possible just to restart the same container with --restart=always, because that's something you have to do when you run a container for the first time and not a parameter that you can use when you start an existing container.

There are three possible work-arounds to this:

1. As @user2915097 stated, you can abandon the original container (stopping it and then deleting it with docker rm <CONTAINER ID>to tidy up). Then just run a new container from the original image specifying the -restart=always option this time.

2. If no volumes were used, so the changes are internal to the container, you need to commit the container to a new image and then run a new container from that image.

   docker commit <CONTAINER ID> <NEW IMAGE NAME>

   docker run -d --restart=always ... <NEW IMAGE NAME>

3. If volumes were used and all changes are restricted to the volumes, then you can run a second container with the --volumes-from parameter without having to commit a new version of the image. i.e.

   • docker stop <CONTAINER 1 NAME>
   • docker run -d --restart=always --volumes-from <CONTAINER 1 NAME> ... <ORIGINAL IMAGE NAME>

   It would then be safe to delete Container 1, as the volumes will not be deleted whilst another container continues to use them.

I guess there is a fourth possibility too; if you used a volume(s) and you know that there have been changes to the container that aren't on the volume, then you'll have to use a combination of (2) and (3).