authzed
How SpiceDb cache is being distributed among all nodes. #2209
-
|
As I understand , nodes are not aware of each other then how cache data is being distributed among each other. I read a old post which talks about cleint side gRPC loadbalancing. Is this still the case? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
@Anuragch yeah, this is described in this blog post. Strictly speaking, each SpiceDB node will be aware of the existence of each other through service discovery (e.g. in kubernetes, we use So yes, none of this behavior has changed and is there since day one, as it is an integral part of Zanzibar's scaling properties. |
Beta Was this translation helpful? Give feedback.
@Anuragch yeah, this is described in this blog post. Strictly speaking, each SpiceDB node will be aware of the existence of each other through service discovery (e.g. in kubernetes, we use
kuberesolver, but it also supports non-kube environments through XDS). So effectively, the gRPC client will load-balance requests across nodes using a consistent hash ring: each subproblem in the SpiceDB schema will have its hash computed and be dispatched to the node responsible for that hash range.So yes, none of this behavior has changed and is there since day one, as it is an integral part of Zanzibar's scaling properties.