Apache Kafka Mirror Maker 2 Bi-directional for Disaster recovery?
We are planning to leverage MM2 Bidirectional for DR scenario for an intensive Kafka app.
Here is an eg.
Kafka Clusters: A, B
Cluster A Topic : t2
Cluster B Topic (Mirrored) : A.t2
Consumer Group : G1
Under normal circumstance we will be consuming t2 from Cluster A.
Once the disaster scenario occurs at Cluster A, we will consume the A.t2 from B. (Given data and offsets synced to B).
The offset from Cluster B syncs to A, once its up. (MM2 sync).
The Cluster B will stop consuming A.t2.
Cluster A will continue to consume t2 from now on. (Back to normal).
We are hearing few facts on watermark issues, data retention delta issues etc.. with bidirectional replication. (Details at the bottom of below page). Appreciate if someone could update, if you have leveraged MM2 Bidirectional for prod scenarios.
Read more here: Source link