i have this problem. Is it really worth it to improve the UX and tooling of running etcd clusters? Wouldn't you rather just get rid of it? I have all the technology to just run k8s on a managed database.
-
-
This Tweet is unavailable.
-
Replying to @vielmettialt @vielmetti
no, because I've already replaced it in k3s. but.... k8s is heavily tied to etcd behavior. so it requires replicating etcd behavior on another data source. Which surprisingly is quite hard on nosql but quite simple on a proper ACID DB.
5 replies 0 retweets 4 likes -
-
We need to spec out desired watch semantics (e.g., the consistency model) and storage layout changes before we expose a pluggable storage API.https://github.com/kubernetes/kubernetes/pull/31622#issuecomment-327606218 …
1 reply 0 retweets 4 likes -
Unfortunately there just isn't enough motivation to do that. Everybody is just going to implement "Just enough etcd API" to make k8s work. Shortest path.
2 replies 0 retweets 2 likes -
Replying to @ibuildthecloud @bgrant0607 and
It is a shame that it was required at the start instead of going the plugable route. It is one of the biggest headaches of managing Kubernetes.
1 reply 0 retweets 1 like -
Replying to @gaunetes @ibuildthecloud and
What would manage then? Everything else is equally or even more complicated to run
1 reply 0 retweets 0 likes -
Replying to @pires_oss @ibuildthecloud and
K3s uses SQLite which may be good for that use case (IDK). It is more that without the plugable model no one can step in a build or try anything else. With CSI, CNI, CRI users can pick and choose what works best.
1 reply 0 retweets 2 likes
Some things need to be right from the beginning because they are infeasible to change. Other things can wait. None of CRI, CNI, CSI, API aggregation, CRD, admission webhooks, etc. existed at the beginning of the project. It's a gradual process of creating the plugin APIs.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.