Gitlab CI/CD fail with "bash: line 132: go: command not found"

We have installed Gitlab on our custom server. We are looking to use the gitlab CI/CD pipeline to build and release our software for that I'm working on a POC. I have created a project with the following .gitlab-ci.yml

variables:
  GOOS: linux
  GOARCH: amd64

stages:
  - test
  - build
  - deb-build

run_tests:
  stage: test
  image: golang:latest
  before_script:
    - go mod tidy
  script:
    - go test ./...

build_binary:
  stage: build
  image: golang:latest
  artifacts:
    untracked: true
  script:
    - GOOS=$GOOS GOARCH=$GOARCH go build -o newer .

build deb:
  stage: deb-build
  image: ubuntu:latest
  before_script:
    - mkdir -p deb-build/usr/local/bin/
    - chmod -R 0755 deb-build/*
    - mkdir build
  script:
    - cp newer deb-build/usr/local/bin/
    - dpkg-deb --build deb-build release-1.1.1.deb
    - mv release-1.1.1.deb build
  artifacts:
    paths:
      - build/*

TLDR: I have updated the gitlab-ci.yml and the screenshot of the error.

What I have noticed, the error is persistent if I use the shared runner(GJ7z2Aym) if you register a runner (i.e Specific Runner)

gitlab-runner register  --non-interactive  --url "https://gitlab.sboxdc.com/"   --registration-token "<register_token>" --description ""  --executor "docker" --docker-image "docker:latest" 

I see the build passing without any problem

Failed case.

https://gist.github.com/meetme2meat/0676c2ee8b78b3683c236d06247a8a4d

One that Passed

https://gist.github.com/meetme2meat/058e2656595a428a28fcd91ba68874e8

The failing job is using a runner with shell executor, that was probably setup when you configured your GitLab instance. This can be seen on logs by this line:

>
> Preparing the "shell" executor
> Using Shell executor...
>

shell executor will ignore your job's image: config. It will run job script directly on the machine on which the runner is hosted, and will try to find go binary on this machine (failing in your case). It's a bit like running go commands on some Ubuntu without having go installed.

Your successful job is using a runner with docker executor, running your job's script in a golang:latest image as you requested. It's like running docker run golang:latest sh -c '[your script]'. This can be seen in job's logs:

>
> Preparing the "docker" executor
> Using Docker executor with image golang:latest ...
> Pulling docker image golang:latest ...
> Using docker image sha256:05e[...]
> golang:latest with digest golang@sha256:04f[...]
>

What you can do:

1. Make sure you configure a runner with a docker executor. Your config.toml would then look like:

[[runners]]
  # ...
  executor = "docker"

  [runners.docker]
  # ...

It seems you already did by registering your own runner.

2. Configure your jobs to use this runner with job tags. You can set tag docker_executor on your Docker runner (when registering or via Gitlab UI) and setup something like:

build_binary:
  stage: build
  # Tags a runner must have to run this job
  tags:
  - docker_executor
  image: golang:latest
  artifacts:
    untracked: true
  script:
    - GOOS=$GOOS GOARCH=$GOARCH go build -o newer .

See Runner registration and Docker executor for details.

Since you have used image: golang:latest, go should be in the $PATH

You need to check at which stage it is failing: run_tests or build_binary.
Add echo $PATH in your script steps, to check what $PATH is considered.

Check also if the error comes from the lack of git, used by Go for accessing modules remote repositories. See this answer as an example.

From your gists, the default GitLab runner uses a shell executor (which knows nothing about Go)

Instead, the second one uses a Docker executor, based on the Go image.
Registering the (Docker) runner is therefore the right way to ensure the expected executor.