how are you running your schema creation, and what preshared-key are you using?

I am using

SPICEDB_TOKEN="random_key"

running the schema creation as part of the globalSetup function of JEST, via

client.writeSchema(schemaRequest)

🤔

but is the

random_key

static, or is it generated by some function on each invocation?

static, via env var

in

serve-testing

mode, each different preshared key is functionally its own sandbox, so if it's different between the setup function and the calls that would explain what you're seeing

the tests only fail on first run after spining up a new spicedb serve-testing container

the other thing would be that the memory DB still uses the same consistency semantics as a "real" instance

all requests are using

fully_consistent

, if not some tests become flaky. My concern here is with the first run only where few tests! a theory: it seems that the

client.writeSchema(schemaRequest)

actually triggers the schema creation in async fashin, the response do not mean that it was fully created, no? 🤔 thus when the jest runner carry on, it starts throwing queries against SpiceDB, too fast, before SpiceDB manages to create parts of the schema 🤔

that would make some sense. does the test setup await that call, or is it firing it off and then continuing with test running?

schemas are revisioned like everything else in SpiceDB

it was a concurrency issue with some paralell tests in Jest. I did refactor it ensuring not to run until writeSchema() finishes. and it works now. ✅ thanks for the swift support 🙌