As I'm continuing to work in docker-machine and Django, I'm trying to make a setup script for my project that auto-detects platform and decides how to set up Docker and the required containers. Auto-detection works fine. One thing I can't figure out is how to automatically set the environment variables needed for docker-machine to work on Mac OS X. Currently, the script will just tell the user to manually set the environment variable using the command
eval $(docker-machine env dev)
where dev is the name of the VM. This prompt happens after initial setup is successfully completed. The user is told to do this because the following subprocess call does not actually set the environment variables:
subprocess.call('eval $(docker-machine env dev)', shell=True)
If an error occurs during creating the VM because the VM already exists, then I use subprocess to see if Docker is already installed:
check_docker = subprocess.check_call('docker run hello-world', shell=True)
If this call is successful, then the script tells the user that Docker was already installed and then prompts the user to manually set the environment variables to be able to start the containers needed for the Django server to run. I had originally thought that the script behaved correctly in this scenario, but it turns out that it only appeared that way because I had already set the environment variables manually. Of course, I see now that the docker run command needs the environment variables to be set in order to work, and since the environment variables never get set in the script, the docker run test doesn't work. So, how am I supposed to correctly set the environment variables from Python? It seems like using subprocess is resulting in the wrong environment getting these variables set. If I do something like
subprocess.call('setdockerenv.sh', shell=True)
where setdockerenv.sh has the correct eval command, then I run into the same problem, which I'm guessing is rooted in using subprocess. Would os have something to do this properly where subprocess can't? It's important that I do this in the Python script, or else having the user manually set the environment variables and then manually test to see if docker is installed defeats the purpose of having the script.
evalworks because as a shell built-in, it doesn't fork another process, but executes in the current process. There is no way to modify the environment of your current shell by running a Python script.os.environand manually setting the environment variables that way. I think what this does is set the env variables for the Python process running the script, so that child subprocesses can use those variables.