Building a docker container
Using alpine as build image
The Containerfile receives VERSION as argument
It is set as image version
The same version is passed to dotnet publish command.
The passed version (which is a variable) in current context never gets substituted. Instead of substituted variable ${VARIABLE} is passed to dotnet publish command.
ARG VERSION
FROM mcr.microsoft.com/dotnet/sdk:6.0-alpine as build
ARG VERSION
RUN echo ${VERSION}
RUN echo ${VERSION} > image_version
WORKDIR /app
COPY . .
RUN dotnet restore
# TODO Pass the version as build command below.
RUN dotnet publish --version-suffix ${VERSION} --output /app/published-app
I tried different ways to do the variable substitution like below:
"$VERSION"
"$(VERSION)"
$VERSION ..... and many other ways I found in the internet. None of them worked. The argument passed is exactly what it is before the variable is substituted.
Thoughts on what may be causing this?
Related
My Dockerfile looks like the following:
from httpd:${bamboo.test.tag}
COPY index.html /usr/local/apache2/htdocs/
In Bamboo I have a task with the following script:
docker build --no-cache -t myproj/my .
When running the job, I get the following error:
build 26-Sep-2022 10:42:26 Step 1/2 : from httpd:${bamboo.test.tag}
error 26-Sep-2022 10:42:26 failed to process "httpd:${bamboo.test.tag}": missing ':' in substitution
How can I substitute the tag?
This is actually a problem with how you are using the dockerfile.
Docker will not expand environment variables inside your Dockerfile. You need to pass the environment value as a build argument in the docker build command then use the ARG keyword inside the Dockerfile.
Your Dockerfile would look like this:
ARG IMAGE_TAG
from httpd:${IMAGE_TAG}
COPY index.html /usr/local/apache2/htdocs/
And you would need to change you docker build command to:
docker build --no-cache --build-arg IMAGE_TAG=${bamboo.test.tag} -t myproj/my .
Check a more detailed explanation here
I have a singularity container built with a scientific filesystem app.
I run the container with singularity run --app myapp image.sif and, accordingly, I have the environment variable SCIF_APPDATA=/scif/data/myapp set inside the container.
I'd like to bind a path using this environment variable. Something like:
singularity run -B /my/path/on/host/:$SCIF_APPDATA/input/
Unfortunately this does not work. I didn't manage to make singularity use as a mount path the environment variable with its "internal" value.
I have to explicitly pass the value of the environment variable:
singularity run -B /my/path/on/host/:/scif/data/myapp/input
Does anybody know how to use container environment variables in bind paths?
I don't think it is possible to directly use environment variables from inside the container in the bind statement. However, you can do it in two steps, first a call to singularity exec to get the value of the variable, then use it with singularity run:
SCIF_APPDATA=$(singularity exec ascisoftApp.sif bash -c "echo \$SCIF_APPDATA")
singularity run -B /my/path/on/host/:$SCIF_APPDATA/input/ ...
Don't forget the backslash to escape the $ in the echo command.
Note: Not sure why but it doesn't work for me when I directly use echo, hence the wrapping with bash -c.
I know -P is used to execute a cmake script file. But I don't want create a file for a simple logic. I want to execute statement like this:
cmake "if(foo) do_something endif()"
The reason I want this feature is that I want to use if/else in add_custom_target and add_custom_command to execute some command according to whether a CMAKE variable is defined.
Finally I got a good solution to this. Now describe as follows.
Let's say I want to add two targets, whose aim are to create docker images and upload to remote repository when issuing make images and make publish. I use VERSION , GIT_COMMIT and PATCH to compose an image tag. GIT_COMMIT can be obtained from git log by using execute_process, and I need to pass newer VERSION and PATCH from command line every time I want to create images. (But if I don't plan to create, they won't be given)
So the complete statements are like this:
execute_process(COMMAND git log --pretty=format:%h -n 1 OUTPUT_VARIABLE GIT_COMMIT)
set(DOCKER_CERBERUS_URL "docker-registry.com/db/cerberus:${VERSION}-${GIT_COMMIT}.${PATCH}")
add_custom_target(images
COMMENT "Creating docker images ..."
# clean the cache
COMMAND rm -f CMakeCache.txt # important !
# if VERSION and PATCH are not given, don't build it
COMMAND /bin/sh -c "if [ x${VERSION} != x -a x${PATCH} != x ]; then docker build -t ${DOCKER_CERBERUS_URL} ${CMAKE_CURRENT_SOURCE_DIR}/docker/; fi"
DEPENDS cerberus
VERBATIM )
add_custom_target(publish
COMMENT "Uploading images ..."
COMMAND docker push ${DOCKER_CERBERUS_URL}
DEPENDS images )
Then every time I want to create or publish with newer version number and patch number, I use cmake .. -DVERSION=xxx -DPATCH=xxx, make images or make publish.
There's something which may be strange. When I first issue cmake .. -DVERSION=xxx -DPATCH=xxx, then make images, sh command will be executed. Then I issue make images again, sh command will not be executed. My guess is: in the first run of make images, cmake will first load variable from CMakeCache.txt into memory, then sh command can get variables's value from memory, and at this time CMakeCache.txt has been deleted. So in the second run, the variables have no value.
I am looking to pass an environment variable which should get a random id.
Something like below.
ENV SERVICE_TAG= $uuid
In short, every time I run the container, I should get a random id for this environment variable inside the container.
Can anyone please suggest the way forward?
Thanks and regards,
Prasanth.
Add uuidgen package to the image. In case you are using alpine add
RUN apk add --no-cache util-linux
to the Dockerfile
Then in the entrypoint of your Dockerfile add
Below is a sample Dockerfile
FROM alpine:latest
RUN apk add --no-cache util-linux
ENTRYPOINT export UUID=`uuidgen` && echo $UUIDFROM alpine:latest
You should pass the environment variable from a docker run -env parameter generated uuid or guid from your shell, eg:
bash:
docker run --env SERVICE_TAG=$(uuidgen) yourimage
See more details at the official docker docs:
https://docs.docker.com/engine/reference/commandline/run/#set-environment-variables--e---env---env-file
I've set an environment variable (NPM_TOKEN) for my repo in Docker Cloud to use when building my Dockerfile. However, the variable is always empty...
Tried both of these in Dockerfile:
RUN echo ${NPM_TOKEN}
and:
ARG NPM_TOKEN
RUN echo ${NPM_TOKEN}
Am I wrong in assuming that Docker Clouds environment variables for build does the same thing as --build-arg?
It took me along time, but you can use build hooks to set variables for automated builds!
https://docs.docker.com/docker-cloud/builds/advanced/#build-hook-examples