...
- Do we want to pursue moving Fabric minions to external hosting in the short term?
- Building raw resources on AWS or IBM cloud.
- What does this look like?
Minutes
- Recap of the conversation so far up to last week.
- Looking at Cloudbees
- They don't offer a managed solutions, they only have a support contract available.
- They have third parties they work with that can do managed services and pricing is on a per-user model.
- A user is every committer.
- Any special pricing for open source projects? We would have to get a quote and then negotiate with them.
- Tested Cloudbees with Fabric and it doesn't really give us any advantage.
- It's very similar to Jenkins and at the end of the day, it's not really something we want to pay for.
- Looked at CircleCI
- It does what Fabric wants to do but it doesn't integrate with Gerrit.
- The two problems:
- CI is one problem and is more concerned
- Moving forward ideas
- CircleCI
- Supports running both on VM's and container hosts.
- Supports AWS and other cloud platforms so it would support Sawtooth long-running tests.
- Does declarative pipelines that Fabric team wants.
- We already have a CircleCI account and we can run some tests on there.
- Bare AWS with teams doing what they want
- Kubernetes
- Tekton UI was unable to get it running.
- Couldn't find support for VM nodes.
- Kubevert can support VMs but we'd have to use rsync to get files onto the machine for testing.
- Tekton UI was unable to get it running.
- CircleCI
- Producing the GB report
- Three sections
- Introduction that describes the current state and the set of problems.
- Describe the potential solutions for solving the CI problem and also for solving the soak/test net problem.
- Itemize the most likely solution to solve the most problems for the most teams.
- Three sections
Action items
- David Huseby investigating Prow and Kubernetes
- Fabric team to narrow solutions for the report.
- Mark Wagner to look into Kubernetes with VM (kubevert)