02/20/2023, 11:47 PM
Hello everybody, we started seeing some of our spicedb calls taking too much time and upon some analysis came to understand that this happens to due things being reloaded from the datastore (postgres) every 5 seconds (the quantization interval) which with our schema and data size is unacceptably slow. Is there any recommended approach to solving such a problem? Are we expected to scale the datastore so that we have decent performance even when no valid cache is available? Can we alternatively implement some smarter cache invalidation (and warming on updates) so that the cache is kept being used at least while there's no changes or better while there's no changes affecting a particular context?