Question posted in
The official documentation can be found here.

Docker not found in Gitlab CI

am trying to follow this link to deploy using Gitlab CI and CDK

my .gitlab-ci.yml looks like below …

image: node:18-alpine    
cache:
  paths:
  - cdk/node_modules/

stages:
- build
- deploy

build:
  stage: build
  script:
    - cd cdk
    - npm install

deploy:
  stage: deploy
  services:
    - name: docker
  variables:
    ENV: "dev"
    AWS_ACCESS_KEY_ID: $AWS_ACCESS_KEY_ID
    AWS_SECRET_ACCESS_KEY: $AWS_SECRET_ACCESS_KEY
    CDK_DEPLOY_ACCOUNT: $SANDBOX_ACCOUNT_ID
    CDK_DEFAULT_ACCOUNT: $SANDBOX_ACCOUNT_ID
    CDK_DEPLOY_REGION: $AWS_DEFAULT_REGION
  before_script:
  - npm install -g aws-cdk@latest
  - cdk --version
  script:
    - docker --version
    - cd cdk && cdk deploy ExpressAppStack --require-approval never --verbose

but it keeps failing at deploy with

npm install -g aws-cdk@latest
added 1 package in 1s
$ cdk --version
2.130.0 (build bd6e5ee)
$ docker --version
/bin/sh: eval: line 155: docker: not found
Cleaning up project directory and file based variables
00:00
ERROR: Job failed: exit code 127

I expected the YML to work as the one mentioned in AWS tutorial, can some one explain how and why to fix this ?

Tags:

2

Answers


  1. 0

    The problem is that the docker executable is not found in the CI runner machine’s paths to executable commands.

    Every job in your gitlab ci file will run in a docker container, as specified by the image keyword. At the top of your file you specified image: node:18-alpine which means all your jobs will run using that image. That means that the docker container executing your script will only have access to the executables that the image has pre-configured with it. And in your case, node:18-alpine does not have the docker cli installed.

    You can reproduce this behavior locally using a docker container.

    ❯ docker run --rm -it --entrypoint /bin/sh node:18-alpine docker --version
/bin/sh: can't open 'docker': No such file or directory

    If you switch your image to be something that has the docker cli installed, you’ll be able to access it within your script. For example, we can see with the docker:25.0.3-dind image.

    ❯ docker run --rm -it docker:25.0.3-dind docker --version
Docker version 25.0.3, build 4debf41

    This particular image is just an example. You’ll want to ensure that the particular image you choose to run your script with has all the necessary dependencies for your deployment process. It’s entirely possible that no such off-the-shelf image has everything you need. In which case you can either build a custom image, or install what you need within your script itself. For example, you might choose to modify your script to simply install the docker CLI before executing the rest of the script steps.

    script:
- # install docker CLI
- # run your deployment steps
    Login or Signup to reply.
  2. 0

    You can install required dependencies for the alpine based image as part of a before_script or similar:

    before_script:
  - apk add docker-cli bash

    You might also need to specify DOCKER_HOST=tcp://docker:2375 to make it actually target your sidecar service.

    bash is needed if you use esbuild for building lambdas and similar via CDK.

    You might also wanna consider creating this base image for repeated use instead of installing dependencies over and over again.

    Login or Signup to reply.
Please signup or login to give your own answer.

Back To Top
Search