AWS Deployments

I’m interested to hear about those deploying concourse to AWS.

We have a very docker heavy pipeline (and make use of dcind) and are seeing quite slow builds (relative to how long they would take on a developer’s machine).

We are now experimenting with c5d.xlarge instances and there seems to be some improvement. I’m also interested to know if people are managing BTRFS volumes themselves or leaving concourse to set up the loop back interface.

hi andy, we are using Concourse currently on vSphere, AWS and Azure as a clustered BOSH deployment.

Our cluster on AWS has one web instance t2.large and three workers also t2.large.

So far i can’t complain about the performance but we have to do a bosh -d concourse recreate frequently when we hit some problems with the worker.

Thanks that’s useful information. Are you doing much Docker stuff in your builds? Pushing and pulling of images etc.?

Thanks that’s useful information. Are you doing much Docker stuff in your builds? Pushing and pulling of images etc.?

Since every Concourse task is scheduled in a docker container we do a lot with docker. But we are using the same docker image over and over again since our ci scripts are using more or less the same tools.

The good thing with Concourse is that it caches very efficient when it comes to your resource docker images or task docker images.

We also have a pipeline which is building and pushing new versions of our dockerfiles which reside in a github repository.

Hi, we keep our workers separate from the master.
We find that a t3a.small runs the concourse master just fine (t3a is better value than t3 or t2). For workers we use m5a.xlarge, these get very busy during the week.
I’m playing with having workers call the land function on the master when shutting down so that I can do some sort of scaling by load, and have workers shut down during evenings and weekends.