Sharding using server sided sharding in 1.36+ k8s #1944
Replies: 1 comment 1 reply
|
Hi thanks @nickexported This new feature is different compared to Sveltos sharding. With Sveltos, users decide when a new set of Sveltos controllers are needed by creating a new cluster shard (by adding the annotation sharding.projectsveltos.io/key to the managed clusters that should belong to it). The shard controller detects the new key and deploys a dedicated set of Sveltos controllers for that shard alone. Which clusters land in which shard is an explicit, operator driven decision: by team, by region, by blast radius. The new Kubernetes feature solves a different problem: automatically splitting one resource type into even hash based ranges (of metadata.uid or metadata.namespace) across replicas of the same controller, so each replica gets a slice of. We can keep an eye on this feature though and see how it evolves. Thanks again for sharing it. |
Uh oh!
There was an error while loading. Please reload this page.
Hi,
Starting with 1.36 k8s release we now have server sided sharded list and watch for controllers.
It looks like it solves exactly the same problem as shard-controller does. Do you have any thoughts or plans on that?
https://kubernetes.io/blog/2026/05/06/kubernetes-v1-36-server-side-sharded-list-and-watch/
All reactions