Arvind Bright

Hi @k-raval -- apologies. We are a bit thin on the ground with multiple teammates on vacation. Hence the delayed turn back time. I would lean on to you for...

Thanks for the logs. The RouteAction looks correct to me with the hash_policy providing the header to use. ```json "hash_policy":{"header":{"header_name":"ULID"}} ``` Also, I should rephrase my previous comment about ring...

Let me followup with the team to get you better guidance for this usecase. Meanwhile, could you confirm that in the case with Envoy proxy you are also using the...

I was not able to find the language which talks about a deterministic ring in all client implementations either in [A42: xDS Ring Hash LB Policy](https://github.com/grpc/proposal/blob/master/A42-xds-ring-hash-lb-policy.md) or in the [Envoy's...

Thanks for the additional information. I will take a look at it shortly

@k-raval - Could you clarify if the ADS response dump that you provided us earlier was for the client that restarts or the one that doesnt?

Based on the ClusterLoadAssignment ADS response, I'm not able to reproduce the same ring on my side. I'm alternatively trying to see if the I can reproduce the issue if...

Apologies for the delay. Could you try with this [branch ](https://github.com/arvindbr8/grpc-go/tree/dbug_6927)?

Removing the `stale` label to give more time on for the reproduction, since there might be a potential bug here.

Thanks. > Somehow, no issue was observed Shows evidence of not being deterministic. I'm sort of inclined now to believe if there is an interaction with istiod ADS responses which...