We are planning for a multi data center HBase cluster for replication ( disaster recovery). The cluster 1 or Master cluster will replicate to the Slave cluster in another DC. In such a setup can both clusters use different zookeepers or should both point to a common zookeeper instance accessible from both data centers ? What is the best strategy.
Since you have mentioned Master and Slave cluster, I assume that the data replication will be done only from Master to Slave. Even though a single zookeeper quorum can be used for handling the replication, I would suggest to go with separate zookeeper quorum for each cluster. Because lets say the data center is one location is completely down due to some natural disaster, the other cluster won't be affected.
If you decide to go with a single zookeeper quorum, you need to mention different zookeeper.znode.parent to maintain data of both cluster. For more details you can also refer to Apache document on Hbase Replication deployment