Hi all. We use the kubernetes operator in EKS. It seems that by default it does not request any resources. We were running into some unexplained speed issues that we fixed with a [patch](https://authzed.com/docs/spicedb/ops/operator#passing-additional-configuration) to request 1G for the spicedb pods -- is there anything to be concerned about with this approach or any recommended values we should use? Thanks for the help!

nope, that's explicitly what the patches are for

OK, great. Makes sense. Seems like this fixed the immediate issue. In the past we've scaled up by increasing replicas and database resources -- generally speaking when should we expect to have to scale up vertically vs horizontally? Is it just case by case or is it e.g. related to whether we're read or write heavy

it depends on a number of factors, but often, yes, read vs write (or check/LR/LS vs direct reads vs writes)

some things that i know to cause spikes in memory: - large LookupResources/LookupSubjects result sets - calling the watch API and then observing a large spike in write traffic - large spikes in general traffic we recently implemented a memory-based admission control that will prevent broad pressure from resulting in an OOM, but you can still trigger an OOM through the above

Got it. Thanks for the help.

sure thing!