Difference between VOLUME and RUN mkdir in Dockerfile

Question

After reading What is the purpose of VOLUME in Dockerfile , I'm still clueless on what is VOLUME ?

My observation is that, adding the following in Dockerfile

VOLUME /my_directory1

is having same behavior as

RUN mkdir /my_directory2

When I docker exec -it xxxxxx sh , I observe 2 directories /my_directory1 and /my_directory2 are created.

I believe VOLUME are more meaningful than mkdir . Just that, I can't understand it after reading What is the purpose of VOLUME in Dockerfile

Can someone explain to me in simpler term, with some hands-on example ? Thank you very much

Answer 1

First thing is you could not be able to compare VOLUMES with the RUN command even in your scenario you are getting those as directories. So, when you use this command in Dockerfile, as,

VOLUME /my_directory1

This will become the location for your volume. And, the image your included will program in a way that it tells when you start a new container from it, to actually create a new volume location and assign it to this directory in the container. Which means any file that you put in there, in the container will outlive the container until we manually delete the volume. So, that is one difference that volumes need manual deletion. You can't clean them up just by removing the container. The whole point of volume command is to say that this data is particularly important, at-least much more important than the container itself.

You can verify that by inspecting the container under mounts.

"Mounts": [
        {
            ....
            "Source": "/var/lib/docker/volumes/bdc5772c2e1e0575d3e2125a15eb46fd7d1690d251568919f0a599a2c53a1044/_data",
            "Destination": "/my_directory1",
            ....
        }
]

This is actually the running container getting it's own unique location on the host, to store that data, and, then it's in background, mapped or mounted to that location in the container, so that location in the container thinks that it's writing at /my_directory1 location, but, it's actually living in the location on the host.

On the other hand, RUN command executes the instruction you specify inside the container and as each commands that create a new layer on top of image.

RUN mkdir /my_directory2

So, it creates the directory inside your container. But, if you delete the container, this directory also gets deleted as the ephemeral in nature.

Hope this clears the concept. Thank you.

Answer 2

The VOLUME command defines meta-data on the image that cannot be unset, indicating where persistent data will be stored. At the time of the definition, docker will save a snapshot of the directory contents for the initial state of the volume. There are two significant downsides to creating a volume in the image:

1. There is a very limited/undefined ability to make future changes to the initial volume contents. Changes to the image under the volume at this location will not be seen, and anyone trying to extend your image with contents in this directory will not be able to use your image.

2. Anyone using your image that doesn't explicitly define the volume again at runtime at the same location will find anonymous volumes in their docker volume ls that are simply guids with no details about what container created them or what directory they are even used for. Keep in mind you can always define the volume at runtime even if not defined in the image, so there's no upside to the volume in the image but you have these downsides.

The RUN command makes a change to the image by creating a new filesystem layer containing the resulting diff of the filesystem from that command. This directory will exist inside the image that is part of the initial container state. If you simply want a directory to exist in your image, this is the option you should use.