N1QL Index on XDCR replication target

Couchbase Server
,
1

We have a production cluster that is replicated by XDCR to another cluster (for backup, DR, reporting, etc.). The replica is not modified independently.
Can we build N1QL indexes (CREATE INDEX) on the replica cluster that are not in the production cluster, and similarly can we drop these indexes?

Side question (yes I know it depends on hardware), what is the order of magnitude (seconds, minutes, hours) wall-clock time for creating an index on a bucket with ~100,000,000 complex documents.

2

@JonathanGoldberg, index definitions do not get replicated over XDCR. Therefore, create/drop on one cluster doesn’t impact the indexes on the other cluster. So you can create/drop index independently on the backup cluster.

Regarding creating an index on 100M documents, given sufficient resources(memory, CPU cores, SSD), it should be a few minutes. Large/complex documents can also degrade performance especially if the “indexed” field(s) is also big.

3

Thanks so much, this was very helpful