I'll subscribe to those issues and try to provide our context on it. It feels a bit daunting as I don't have the academic background to discussing those in deep details, but a man's gotta learn 💪

Thanks for your help

thank you helping out!

How beefy is the spicedb image? If running it against RDS in an EKS cluster, what kind of CPU/IO/RAM specs should we expect to need to throw at it to provide reasonable responsiveness?

hi @User, there are a few factors at play

each SpiceDB node will contain a cache, which takes up some amount of memory (trying to get you the exact numbers now), but roughly 16MB per cache, and there are two per node

plus an additional 16MB cache for types

we usually see around ~100-200mb of memory per SpiceDB node with "steady state" traffic

but it can vary based on incoming traffic

as for CPU and IO, it really will depend heavily on your traffic

Will spicedb take advantage of multiple cores?

yes

it spins off work in goroutines

So giving each node 2 virtual cores should be a good starting point.

yes

but as I caveated above: it will heavily depend on your traffic

I don't imagine our initial traffic being very heavy. Being able to scale up and out is what matters.

yeah

the key parts of scaling up and out will be (a) ensuring your datastore scales and (b) ensuring that your SpiceDB nodes are configured to dispatch to one another

that way, you can add more SpiceDB nodes as your traffic increases

RDS backend... so it should scale just fine

Hello 👋 I have a small question regarding the SpiceDB schemas. I would need to be able to patch the schema to add dynamically some definitions, relations and permissions. From what I found in the documentation we can only use the

WriteSchema

method, however I doesn't seem to be able to call it if some relations already exists in the current schema. Is there a clean solution to do it ?

You can write to the schema as long as you’re not removing any relations which currently have relationships

Adding net new things should be fine

If you need to delete all data before removing a relation you can use the DeleteRelationships call and specify only the object type and relation.

@User Thank for your answer ! Currently I only tried using the `zed`client and when I try to write twice the exact same schema I receive an error stating that data already exist. If I well understand, using directly a gRPC call would work ?

It should work from zed, I will take a look when I get to a computer