- Running as unprivileged user
- Selecting your shell
- Terminating and killing processes
The Shell executor is a simple executor that you use to execute builds locally on the machine where GitLab Runner is installed. It supports all systems on which the Runner can be installed. That means that it’s possible to use scripts generated for Bash, Windows PowerShell and Windows Batch (deprecated).
git lfscommand if Git LFS is installed on the machine, so ensure Git LFS is up-to-date when GitLab Runner will run using the shell executor.
The scripts can be run as unprivileged user if the
--user is added to the
gitlab-runner run command. This feature is only supported by Bash.
The source project is checked out to:
The caches for project are stored in
<working-directory>is the value of
--working-directoryas passed to the
gitlab-runner runcommand or the current directory where the Runner is running
<short-token>is a shortened version of the Runner’s token (first 8 letters)
<concurrent-id>is a unique number, identifying the local job ID on the particular Runner in context of the project
<namespace>is the namespace where the project is stored on GitLab
<project-name>is the name of the project as it is stored on GitLab
To overwrite the
cache_dir options under the
If GitLab Runner is installed on Linux from the official
packages, the installer will try to use the
user if found. If it is not found, it will create a
gitlab-runner user and use
All shell builds will be then executed as either the
In some testing scenarios, your builds may need to access some privileged
resources, like Docker Engine or VirtualBox. In that case you need to add the
gitlab-runner user to the respective group:
usermod -aG docker gitlab-runner usermod -aG vboxusers gitlab-runner
GitLab Runner supports certain shells. To select a shell, specify it in your
config.toml file. For example:
... [[runners]] name = "shell executor runner" executor = "shell" shell = "powershell" ...
Generally it’s unsafe to run tests with shell executors. The jobs are run with
the user’s permissions (
gitlab-runner) and can “steal” code from other
projects that are run on this server. Use it only for running builds on a
server you trust and own.
The shell executor starts the script for each job in a new process. On UNIX systems, it sets the main process as a process group.
GitLab Runner terminates processes when:
- A job times out.
- A job is canceled.
On UNIX systems
gitlab-runner sends a
SIGKILL to the process to
terminate it, because the child processes belong to the same process
group the signal is also sent to them. Windows sends a
taskkill /F /T.
On UNIX system
SIGTERM to the process and its
child processes, and after 10 minutes sends
SIGKILL. This allows for
graceful termination for the process. Windows don’t have a
equivalent, so the kill process is sent twice. The second is sent after
If for some reason this new termination process has problems with your
scripts but works with the old method you can
set the feature flag
true, and it will use the old method. Keep in mind that this
feature flag will be removed in GitLab Runner 14.0 so you still need to
fix your script to handle the new termination.