The goal of this document is to give you enough technical specifics to configure and run the autoscaler. When properly configured, it will automatically provision and terminate Google Compute Engine instances based on your Drone server’s build volume.

Prerequisites

Create a Drone Token

You need to create a user Drone token with administrative privilege that you can provide to the autoscaler. The autoscaler will use this token to remotely access the Drone build queue. If you do not know how to create a token please see our tutorial.

Create Google Credentials file

If you running you autoscaler in a GCP instance then create application_default_credentials.json by executing next command in the GCP instance:

gcloud auth application-default login

And credentials will be saved in /home/$(whoami)/.config/gcloud/application_default_credentials.json.

Otherwise, create service account, generate credentials file and save it somewhere in the machine where plan to run autoscaler.

Download

The autoscaler is distributed as a lightweight Docker image. The image is self-contained and does not have any external requirements.

docker pull drone/autoscaler

Start the Server

The autoscaler container can be started with the below command. The container is configured through environment variables. For a full list of configuration parameters, please see the configuration reference.

$ docker run -d \
  -v /var/lib/autoscaler:/data \
  -v /path/to/application_credentials.json:/root/.config/gcloud/application_default_credentials.json \
  -e DRONE_POOL_MIN=${DRONE_POOL_MIN} \
  -e DRONE_POOL_MAX=${DRONE_POOL_MAX} \
  -e DRONE_SERVER_PROTO=${DRONE_SERVER_PROTO} \
  -e DRONE_SERVER_HOST=${DRONE_SERVER_HOST} \
  -e DRONE_SERVER_TOKEN=${DRONE_SERVER_TOKEN} \
  -e DRONE_AGENT_TOKEN=${DRONE_AGENT_TOKEN} \
  -e DRONE_GOOGLE_PROJECT=${DRONE_GOOGLE_PROJECT} \
  -e DRONE_GOOGLE_ZONE=${DRONE_GOOGLE_ZONE} \
  -p 8080:8080 \
  --restart=always \
  --name=autoscaler \
  drone/autoscaler

Reference

Environment

Configuration parameters are set using environment variables. This section defines a subset of recommended configuration parameters. For a full list, please see our configuration reference.

DRONE_POOL_MIN

An integer defining the minimum number of instances the autoscaler should keep alive. The default value is two instances.

DRONE_POOL_MIN=2

DRONE_POOL_MAX

An integer defining the maximum number of instances the autoscaler can provision. The default value is four instances.

DRONE_POOL_MAX=4

DRONE_SERVER_PROTO

A string containing your Drone server protocol scheme. This value should be set to http or https.

DRONE_SERVER_PROTO=https

DRONE_SERVER_HOST

A string containing your Drone server hostname or IP address.

DRONE_SERVER_HOST=drone.domain.com

DRONE_SERVER_TOKEN

A string containing your Drone user token. This token must grant administrative access to your Drone server.

DRONE_AGENT_CONCURRENCY

An integer defining the maximum number of concurrent builds an agent can execute. This value is used to configure new agents.

DRONE_AGENT_CONCURRENCY=2

DRONE_AGENT_TOKEN

A string containing the shared secret (DRONE_RPC_SECRET) that authorizes agent to server communication. This value is used to configure new agents.

DRONE_AGENT_TOKEN=IiOiIxMjM0NTY3ODkwIiwibmFtZSI6Ikxvb2

DRONE_GOOGLE_PROJECT

A string containing your Google Cloud Platform project ID.

DRONE_GOOGLE_PROJECT=my-sample-123

DRONE_GOOGLE_ZONE

A string containing the zone in which new instances are provisioned.

DRONE_GOOGLE_ZONE=us-central1-a

Network

The autoscaler exposes REST endpoints so that you can extract runtime information from the system. The autoscaler listens on port 8080 inside the container, and should be published on the host machine:

--publish=8080:8080

Volumes

The autoscaler creates a sqlite database and persists to a container volume at /data. To prevent dataloss, we recommend mounting the data volume to the host machine.

--volume=/var/lib/autoscaler:/data