<@UB2EH889X> Thanks for mentioning RunsOn again :w...
# office-hours
c
@Erik Osterman (Cloud Posse) Thanks for mentioning RunsOn again 😉 Regarding the license check possibly breaking the install: this is a soft check. Even if the licensing server goes down, the RunsOn install continues to work. Same for the telemetry ping. And to answer a question you had during the session: definitely not a SaaS, no centrally managed component or interface on my side except the license and telemetry endpoints. Also regarding the page about RunsOn you were not able to find in the CloudPosse documentation: you mention it at the bottom of https://docs.cloudposse.com/layers/github-actions/design-decisions/decide-on-self-hosted-runner-architecture/#option-5-managed-runners (option 5), but agree with you that it would be great to have a dedicated page in https://docs.cloudposse.com/layers/github-actions/ 😇
👍 1
z
hey @Cyril (runs-on.com) , we currently use ARC with Kubernetes, and we don’t love our current implementation and are always looking for ways to improve
but I don’t fully understand what problems runs-on solves. we’re already using self hosted nodes on our k8s cluster with karpenter
what does runs-on do for us?
e
a) you don't need a k8s cluster
b) it scales to zero
c) it's waaaaay easier to configure a lot of different runner groups
When you use ARC, it wills scale runner pods to zero, but k8s doesn't scale to zero 🙂
When you use ARC, you have 2 problems: 1) you have to optimize and configure ARC for scale 2) you have to optimize and configure karpenter (or your autoscaler) for scale
c
What @Erik Osterman (Cloud Posse) said, and you also get proper VMs with compatible images
z
c) it's waaaaay easier to configure a lot of different runner groups
I think this is the main thing I’d be interested in. Why is it easier?
also scaling to zero is cool from a costs perspective, but you lose out on local cache right? we use local nvme disks and it makes quite a difference
e
It’s easier due to its brilliant use of labels https://runs-on.com/configuration/job-labels/
Copy code
job:
  runs-on:
  - runs-on # required so that RunsOn knows it needs to process the workflow
  - cpu=4
  - ram=16
  - family=m7a+m7i-flex
  - image=ubuntu22-full-x64
z
thanks for the info! will take a look with my team