How does docker run interpret dynamically generated --env arguments

问题

I am trying to provide a dynamically generated list of --env VAR1 --env VAR2 --env-file env.list environment variables to docker run.

Unfortunately it is not working.

for --env mapped variables, the variables are not visible in the container.
for --env-file provided file, docker complains that it cannot find the file: docker: open "env.list": no such file or directory.

Details

Running:

# env_params contains either --env or --env-file arguments
MY_VAR=123
env_params='--env "MY_VAR"'
echo ${env_params} 

docker run -it --rm \
    ${env_params} \
    my_docker_image env | grep MY_VAR

will not output anything. MY_VAR is not visible inside the container. But:

MY_VAR=123
docker run -it --rm \
    --env "MY_VAR" \
    my_docker_image env | grep MY_VAR

will work and 123 will be printed.

In a similar way --env-file will not work when provided through env_params but will work when provided directly to the docker run command.

What am I doing wrong?

回答1:

There are two issues here.

First, When you run, in your shell:

MY_VAR=123

You have not set an environment variable. You have set a local shell variable. When you use --env MY_VAR, you are telling Docker that you want to make the environment variable MY_VAR available inside the container, and since it doesn't exist you get nothing:

$ MY_VAR=123
$ docker run -it --rm -e MYVAR alpine env | grep MY_VAR
<crickets>

If you first export that to the environment:

$ export MY_VAR=123
$ docker run -it --rm -e MYVAR alpine env | grep MY_VAR
MY_VAR=123

Then it will work as you expect. Alternately, you can use the VARNAME=VARVALUE form of the --env option:

docker run -e "MY_VAR=${MY_VAR}" ...

The second issue has to do with how shell variable interpolation works. If you have:

env_params='--env "MY_VAR"'   
docker run -it --rm \
    ${env_params} \
    alpine env

Then the resulting command line is:

docker run -it --rm --env '"MY_VAR"' alpine env

That is, the argument you're passing to docker run includes literal double quotes. You can fix that through the use of the eval statement (keeping in mind that you'll need to modify your script to export MY_VAR):

eval docker run -it --rm \
    ${env_params} \
    alpine env | grep MY_VAR

Alternately (and I would argue preferably) you can use your env_params variable as an array, as long as you're using bash:

env_params=(--env MY_VAR)
env_params+=(--env SOME_OTHER_VAR)

docker run -it --rm \
    "${env_params[@]}" \
    alpine env | grep MY_VAR

Which would result in the correct command line:

docker run -it --rm --env MY_VAR --env SOME_OTHER_VAR alpine env

I guess the summary here is that your issues ultimately have nothing to do with "how docker run interprets dynamically generated arguments", but have everything to do with "how shell variables and interpolation work".

来源：https://stackoverflow.com/questions/48807931/how-does-docker-run-interpret-dynamically-generated-env-arguments

标签

bash

environment-variables