How to run a script from file in another project using include in GitLab CI?

user_3600709 0
Viewed 17646

I'm trying to run a shell script from my template file located in another project via my include.

How should this be configured to work? Below scripts are simplified versions of my code.

Project A

template.yml

deploy:
  before_script:
    - chmod +x ./.run.sh
    - source ./.run.sh

Project B

gitlab-ci.yml

include:
 - project: 'project-a'
    ref: master
    file: '/template.yml'

stages:
  - deploy

Clearly, the commands are actually being run from ProjectB and not ProjectA where the template resides. This can further be confirmed by adding ls -a in the template file.

So how should we be calling run.sh? Both projects are on the same GitLab instance under different groups.

gitlab-ci
4 Answers

If you have access project A and B, you can use multi-project pipelines. You trigger a pipeline in project A from project B.

In project A, you clone project B and run your script.

Project B

job 1:
  variables:
    PROJECT_PATH: "$CI_PROJECT_PATH"
    RELEASE_BRANCH: "$CI_COMMIT_BRANCH"
  trigger:
    project: project-a
    strategy: depend

Project A

job 2:
  rules:
    - if: '$CI_PIPELINE_SOURCE == "pipeline" && $PROJECT_PATH && $RELEASE_BRANCH'
  script:
    - git clone -b "${RELEASE_BRANCH}" --depth 50 https://gitlab-ci-token:${CI_JOB_TOKEN}@${CI_SERVER_HOST}/${PROJECT_PATH}.git $(basename ${PROJECT_PATH})
    - cd $(basename ${PROJECT_PATH})
    - chmod +x ../.run.sh
    - source ../.run.sh
user_142147880

We've also run into this problem, and kinda wish Gitlab allowed includes to "import" non-yaml files. Nevertheless the simplest workaround we've found is to build a small docker image in repo A, which contains the script you want to run, and then repo B's job uses that docker image as the image, so the file run.sh is available :)

Minimal Dockerfile:

FROM bash:latest
COPY run.sh /usr/local/bin/
CMD run.sh

(Note: make sure you chmod +x run.sh before building your image, or add a RUN chmod +x /usr/local/bin/run.sh step)

Then, you'd just add this to your Project B's .gitlab-ci.yml:

stages:
  - deploy

deploy:
  image: registry.gitlab.com/...  # Wherever you pushed your docker image to
  script: run.sh
user_95877070

it's also possible to request a script by curl instead of copying a whole repository:

- curl -H "PRIVATE-TOKEN:$PRIVATE_TOKEN" --create-dirs "$CI_API_V4_URL/projects/$CI_DEPLOY_PROJECT_ID/repository/archive?path=pathToFolderWithScripts" -o $TEMP_DIR/archive.tar.gz
- tar zxvf $TEMP_DIR/archive.tar.gz -C $TEMP_DIR --strip-components 3
- bash $TEMP_DIR/run.sh
  • to make a curl request
  • to archive a folder with scripts
  • to unzip scripts in a temporary folder
  • to execute sh
user_34953210

As hinted by the answer above, multi project pipelines is the right approach for it.

Here's how it worked for me:

GroupX/ProjectA - contains reusable code

# .gitlab-ci.yml

stages:
  - deploy

reusable_deploy_job:
  stage: deploy
  rules:
    - if: '$CI_PIPELINE_SOURCE == "pipeline"'   # run only if triggered by a pipeline
  script:
    - bash ./src/run.sh $UPSTREAM_CUSTOM_VARIABLE

GroupY/ProjectB - job that will reuse a code

# .gitlab-ci.yml

stages:
  - deploy

deploy_job:
  stage: deploy
  variables:
    UPSTREAM_CUSTOM_VARIABLE: CUSTOM_VARIABLE   # pass this variable to downstream job
  trigger: groupx/projecta
user_168572980
Related