Drone AWS Autoscaler and Agent Pools

I would like to experiment with the Drone AWS Autoscaler.
Our Drone setup is made of different Agent pools and each pipeline target the right pool by using Labels.

We would like to replace our Terraform script that creates the AWS Agents (with LaunchConfiguration, TargetGroups and Autoscalers) by the Drone autoscaler.

I have a few questions about the Done Autoscaler before I start setting it up:

  • if we want to maintain different Agent pools, we will have to deploy one Drone Autoscalers per pool, right ?
  • Then, can multiple Drone Autoscalers share the same Postgresql DB ?
  • And can Drone Autoscalers use the Server’s Postgresql DB ?
  • We have some projects where multiple pipelines running in parallel. Let’s say one commit on a repo triggers 6 pipelines in parallel, we need to have at least 6 Agents at any time. We can do it with DRONE_POOL_MIN=6. If another commit is made on the same repo, we will queue 6 more pipelines. Does that mean the Autoscaler will create 6 new agents right away (if we set DRONE_POOL_MAX=12) ? Is there a delay we can set to wait before creating (and then deleting) Agents ?

Thanks for your help

1 Like