- Ref read the EPIC (aws decrease costs)
- Get the most of the resources at our disposal (e.g. using machines that are standing dormant)
- Improve resiliency of our services by using multiple cloud providers
Add another Kubernetes cluster hosted in the 2 OSUSL VMs (that used to host JIRA and Confluence) to provide more compute resource for the "container agent" workload in ci.jenkins.io.
These 2 OSUSL machines are:
- edamame: 16 vCPUs, 12 Gb of memory, 45 Gb of disk, currently hosting a container "butlerbot", used to host JIRA
- lettuce: 8vCPUs, 12 Gb of memory, 38 Gb of disk, used to host confluence
The idea is to start k3s cluster (running as a systemd unit), either a single one or one per VMs (the easier is the best).
- Define a role(s?) in Puppet (jenkins-infra/jenkins-infra) for k3s
- Apply these roles to the VM after an overall cleanup
- Add this (these) cluster(s) to ci.jenkins.io