yep! it's

grpc_server_handling_seconds_bucket

So I generated this histogram and it's doing a simple test - querying for the same permission for the same subject and resource over and over again. My question is - is the latency of the permission service includes the latency for dispatch. For example if pod1 receives a request but dispatches to pod2, is the permission latency from the perspective of pod1 includes the time to dispatch as well?

if cache is used, dispatches won't be made

I'm a bit confused here, - isn't the cache distributed itself, i.e it's only cached on the pod to which the query hashes to ? In my simple setup I have 2 pods. So if the requests are load balanced - half of the requests goto the pod 1 and half to pod2 and if say pod1 only caches it because it's the only home of the query it will answer the requests directly from the cache, but pod2 will need to do a dispatch to get to pod 1. Is that understanding correct?

if you're making the same call