When developing using docker on OS X, you’ll currently¹ have to use docker-machine to spin up a virtual machine that is actually running the docker containers. Running a virtual machine takes up a bit more in the way of resources than just the docker containers, so if you’re not actually developing at the moment, it’s helpful to be able to start up the virtual machine only when you need it.

The current way I have to do that:

$ docker-machine start default
$ eval $(docker-machine env default)

What’s worse, the latter command has to be run for every shell that you start up. It’s by no means a hard pair of commands and you could easily wrap them in an alias or put them in your .profile equivalent (which is what I used to do). But unfortunately, I found a completely unrelated bug in tmuxp: if the shell takes too long to start up, tmuxp essentially won’t work. The above eval command took long enough to hit this limit.

So how do we fix it? Essentially (using zsh, my current shell of choice, although others should be similar):

assert-docker() {
    command docker ps 2> /dev/null > /dev/null
    if [ $? -ne 0 ]; then
        echo "Starting docker..."
        docker-machine start default
        eval $(docker-machine env default)
        echo
    fi
}

docker () { assert-docker && command docker $@ }
docker-compose () { assert-docker && command docker-compose $@ }

The basic idea is that assert-docker first checks if docker is running by trying to run docker ps. $? contains the status code, which will be non-zero if docker ps failed, so check that. If that’s the case, assume docker ps failed because docker-machine wasn’t running, so start it up. This will run docker-machine start default more often than needed, but it turns out it’s a [[wiki:NOP]]() if it’s already running.

The only interesting part here is the use of the keyword command that prefixes docker or docker-machine within the functions. Basically, this tells ZSH to use the system version of docker or docker-compose rather than the one that I defined, thus preventing an infinite loop. Whee!

How’s it work?

$ docker ps

Starting docker...
(dev) OUT | Starting VM...
Started machines may have new IP addresses. You may need to re-run the `docker-machine env` command.

CONTAINER ID        IMAGE               COMMAND             CREATED             STATUS              PORTS               NAMES

$ docker ps

CONTAINER ID        IMAGE               COMMAND             CREATED             STATUS              PORTS               NAMES

Switching to another terminal:

$ docker ps

Starting docker...
Machine "default" is already running.

CONTAINER ID        IMAGE               COMMAND             CREATED             STATUS              PORTS               NAMES

$ docker ps

CONTAINER ID        IMAGE               COMMAND             CREATED             STATUS              PORTS               NAMES

Neat. That should eventually save me a fraction of the time it took to get it right. :)