gitlab-autoscale-runner

Warning: This is an early alpha version without commitment for backwards compatibility.

gitlab-autoscale-runner is a docker container specialized in creating an autoscale runner for gitlab. The default settings are geared towards creating a runner that allows building docker images. The first runner is automatically created and registered to the gitlab server.

Environment Variables

Here are some of the relevant Environment Variables.

Name	Default	Description
CONFIG_FILE	/etc/gitlab-runner/config.toml	Path to config file
CI_SERVER_URL	https://gitlab.com/ci	The CI endpoint of gitlab
CONCURRENT	20	Hard limit of maximum number of machines running concurrently
REGISTRATION_TOKEN	-	The Runners Registration Token
RUNNER_NAME	Autoscale Runner	Name of Runner
RUNNER_LIMIT	1	Maximum number of machines fors this Runner
MACHINE_NAME	gitlab-autoscale-%s	Naming scheme for the VMs
MACHINE_MAX_BUILDS	100	Maximum builds until a machine is delete
MACHINE_IDLE_COUNT	0	Number of permanent idling hosts
MACHINE_IDLE_TIME	600	Seconds after which an idling host is deleted
MACHINE_DRIVER	-	Docker Machine driver e.g: 'digitalocean', 'google', 'aws', ...
DOCKER_IMAGE	docker:latest	Default image to run builds
DOCKER_PULL_POLICY	always	Image Pull Policy
DOCKER_PRIVILEGED	true	Privileged is required to run docker builds. Attention: Only allow trusted builds to run on this runner when this is true
CACHE_TYPE	-	Currently only s3 is supported for caching
S3_SERVER_ADDRESS	s3.amazonaws.com	Address of S3-compactible endpoint
S3_ACCESS_KEY	-	S3 Credentials
S3_SECRET_KEY	-	S3 Credentials
S3_BUCKET_NAME	-	Name of Bucket in S3
S3_BUCKET_LOCATION	-	Region of Bucket in S3
S3_CACHE_INSECURE	false	If true use http instead of https
DIGITALOCEAN_ACCESS_TOKEN	-	Access Token for creating VMs on DigitalOcean
DIGITALOCEAN_REGION	nyc3	Region to run Droplets in
DIGITALOCEAN_SIZE	512mb	Size of Droplets
DIGITALOCEAN_IMAGE	coreos-beta	Disk image of Droplets
DIGITALOCEAN_SSH_USER	core	SSH user to connect to Droplets
GOOGLE_APPLICATION_CREDENTIALS	-	If set docker machine will look for credentials at specified path, otherwise it will use credentials from metadata. Read more here.
GOOGLE_PROJECT	-	Google Project ID
GOOGLE_ZONE	us-central1-a	Zone to provision the VM
GOOGLE_MACHINE_TYPE	f1-micro	Type of machine to provision
GOOGLE_MACHINE_IMAGE	projects/coreos-cloud/global/images/family/coreos-beta	Machine Image ID
GOOGLE_SCOPES	https://www.googleapis.com/auth/logging.write	Scopes of provisioned machines
GOOGLE_DISK_SIZE	10	Size of Disk in GB
GOOGLE_DISK_TYPE	pd-standard	Disk type of ephermal disk
GOOGLE_USERNAME	core	Username to connect to the VM
GOOGLE_USE_INTERNAL_IP	false	Use internal IP when connecting to VM
GOOGLE_TAGS	gitlab-autoscale	Tags to be attached to the VMs
GOOGLE_PREEMPTIBLE	false	Use preemtible VMs

Persistance

For persistance you need to put the config file in a persistant volume. The config file if run as root is specified using the CONFIG_FILE Environment Variable otherwise it is saved to ~/.gitlab-runner/config.toml.

Examples

Start runner using GCE

docker run -e CI_SERVER_URL=https://gitlab.example.io/ci \
           -e REGISTRATION_TOKEN=Xz5waDAF4sgfADEXAMPLE \
           -e MACHINE_DRIVER=google \
           -e GOOGLE_PROJECT=my-project \
           -e GOOGLE_APPLICATION_CREDENTIALS=/etc/credentials.json \
           -v $(pwd)/credentials.json:/etc/credentials.json \
           wikiwi/gitlab-autoscale-runner

Start runner using DigitalOcean

docker run -e CI_SERVER_URL=https://gitlab.example.io/ci \
           -e REGISTRATION_TOKEN=Xz5waDAF4sgfADEXAMPLE \
           -e MACHINE_DRIVER=digitalocean \
           -e DIGITALOCEAN_ACCESS_TOKEN=2SBxEFaAWDe3AAWDEXAMPLE
           wikiwi/gitlab-autoscale-runner

Update Config

# When running as non-root
docker exec -it CONTAINER vi /home/gitlab-runner/.gitlab-runner/config.toml

# When running as root
docker exec -it CONTAINER vi /etc/gitlab-runner/config.toml

Add additional Runner

docker exec CONTAINER /bin/bash -c "RUNNER_NAME='My Runner' RUNNER_LIMIT=1 register.sh"

Docker Hub

Automated build is available at the Docker Hub.

TODO

• Integrate and document AWS driver.