How does kine with mysql handle conflicting writes from two API servers concurrently

[friggitello6]

Using mysql as external storage, how does kine handle conflicting writes from two different API servers concurrently?

[brandond]

It handles it quite well!

#106 (comment)

[friggitello6]

It handles it quite well!

#106 (comment)

great, so we rely on DB MVCC. This means if I write a controller and try to update a resource, the controller is responsible to handle the "resource already modified error" and potentially reload and retry, right? If that's the case, do we need to do the same for controller running on K8s where the write always goes through the etcd leader?

On the similar topic, is there any container app that can run on k8s, but not K3s? In other words, can someone on purposely write a container app that doesn't work on k3s, but works on k8s?

[brandond]

Kine implements all the required bits for Kubernetes to function properly, including guarantees around consistency of sequential writes. Dealing with retries is part of writing any Kubernetes controller.

K3s is a Kubernetes distro. It is built from the same upstream code as all the other distros and passes all the standard conformance tests. It is not anything other than Kubernetes and does not behave differently, regardless of what datastore you use.