Hi! We're currently evaluating SpiceDB for our permission storage and I've got a few things that I'm still a bit unclear on: * What is the "canonical" way to handle schema migrations? The guides only briefly mention this as "you will want to write data migrations and apply those changes using a schema migration toolchain". Is there some prior art on this somewhere? * We want to use the SpiceDB operator to manage our instance. Is there documentation on the

config

part of a `SpiceDBCluster`'s spec? * Are there recommendations on how many resources to allocate for the SpiceDB server?

hi! Today is US holiday so the team will be slow to respond, apologies! I'll do my best to answe your questions >* What is the "canonical" way to handle schema migrations? The guides only briefly mention this as "you will want to write data migrations and apply those changes using a schema migration toolchain". Is there some prior art on this somewhere? - I don't think we've written about schema migrations extensively. We like to draw parallels with relation database migrations in that you may need to do a phased execution: - You need to make sure your application stops writing relationships of any relation you are about to drop, and make sure all relationships defined after a given relation are removed. SpiceDB will prevent writing a new schema that drops a relation for which relationships exist. - Nothing immediately actionable if you are adding a new relation. It's a backward compatible change. >* We want to use the SpiceDB operator to manage our instance. Is there documentation on the config part of a SpiceDBCluster's spec? SpiceDB Operator example: https://github.com/authzed/spicedb-operator/blob/main/examples/cockroachdb-tls-ingress/spicedb/spicedb.yaml#L8-L20 In general what goes into

config

is typically everything that SpiceDB accepts as an CLI flag / environment variable, except you turn the name as described in the CLI help output from

kebab-case

into

camelCase

. For example

--log-level

turns into

logLevel

>* Are there recommendations on how many resources to allocate for the SpiceDB server? Not 100% sure we have a "formula" for this. It largely depends on what you schema looks like. It's better to experiment on a case by case basis, but we always recommend starting with SpiceDB clusters of 3-5 nodes depending on the workload. What datastore are you going to use?

Awesome, thanks a bunch 🙂