This is concerning index rebuild's impact on transaction replication.
During an index rebuilds with full recovery mode, transaction log entries will be created (normal).
If I switch to bulk logged recovery, will:
- less data be replicated ?
- less mining of the transaction log occur ?
What I'm seeing after an index maintenance where some large table indexes are rebuild, replication will report "Approximately 5000000 records have been scanned in pass #4, 0 of which were marked for replication".
This concerns me in that "real" data changes are being slowed down in getting to the subscriber.
Is there an actual proof that this is happening while indexes are rebuilt? Do you have enterprise Edition so you can rebuild indexes with ONLINE=ON? If yes then I believe only the Log reader agent does more work and as long as you don't see any queue in the Distribution agent because of this there should be not a delay in delivering transactions to the subscriber. You can check that under the "Undistributed Commands" tab in the Replication Monitor on the particular Subscription(s) where your concern is.