José, thanks for the updates. I think most of my questions have been answered. I had just two more:
DA5: Do we have a way to quantify the metadata propagation latency? In the past, I think we have eye-balled the metadata LEO between brokers and the active controller to approximate the lag, but maybe you've found a better way? DA6: Will the MV be the only way to enable/disable this new behavior? -David A On Fri, May 2, 2025 at 10:47 AM José Armando García Sancio <jsan...@confluent.io.invalid> wrote: > Hi Jun, > > On Thu, May 1, 2025 at 12:38 PM Jun Rao <j...@confluent.io.invalid> wrote: > > We could probably just keep the new HighWatermark field as described in > the > > KIP, but limit it only for KRaft. Also, since HighWatermark is a tagged > > field, it probably doesn't need to be gated by an MV. > > Yes. Because the HighWatermark field is tagged and ignorable, all > version 18 FETCH requests can be serialized to versions 17 requests. > We still need an MV version because the replica fetcher restricts > possible version values using the MV and > FetchRequest.Builder#forReplica builder. > > Thanks, > -- > -José > -- -David
