We are having Active Active Kafka cluster setup with topic renaming using Mirror Maker 2.0 as specified in https://strimzi.io/blog/2020/03/30/introducing-mirrormaker2/. I believe topic such as us-email are setup as follows:
dc1
- us-email
- us-email-dc2 (mirror of dc2)
dc2
- us-email
- us-email-dc1 (mirror of dc1)
Producers can publish to their local DC's and both clusters would contain data of both the DC's. So far so good.
Consumer app would subscribe to wild card topic (us-email-*) to read data of both DC's. If that's the case, Do I setup a consumer to read from their respective DC's? In this case, there will be duplicate message read for reach message due to mirroring. OR it is recommended to point a single consumer group to a single DC only at a time to prevent duplication? If yes, if a single DC fails, how will the failover happen?