Wikimedia Release Engineering Team/Local Dev Sync/2019-12-05
Agenda:
- Status on mw/core dev setup. What needs to be done?
- Spend time looking at / trying kosta's POC
- Think about putting it in a separate directory
- Complex dev, what to do next?
- collect metrics from vagrant?
- make some sort of scaffold dev environment individuals can add things to?
- ask teams individually what they need?
- continue with local-charts?
- could continue with helm charts but ditch minikube
- helm offers the customizability/flexibility that we need
- kubernetes requires some overhead
- would be nice to have a solution that could work with k8s and docker
- k3sup could be used instead of minikube. We could create an image of it and distribute
- consider that this adds tooling that people have to learn and understand when they might not need to
- don't want to maintain dependency tree
- maintain vagrant?
- don't want to, but there some(one?) who think we should
- maintain docker-dev?
- make it more user-friendly and integrated into core
- other?
- see what it would take to use k3sup instead of minikube
- container-based system is nice so we can use images generated in CI
- would be nice if you could dl an image of some sort and just start
- difficult because of the different OSs
- would be nice if you could dl an image of some sort and just start
- is a role-based system maintainable?
- The problem is vagrant got really big and no one who knows anything about it is on the hook to maintain it
- might be useful the just have a few "roles", like, frontend, backend, etc
- is a role-based system maintainable?
Decisions:
- For now try working on docker-dev for mw/core
- keep experimenting with local-charts
- Try k3s in earnest
- https://github.com/alexellis/k3sup
- Just make a basic vanilla vm and run k3sup from outside (give it the vm's IP)
- Boom, you've got a k3s cluster
- Try k3s in earnest