MicrosoftMicrosoftRegarding the assignment of parallel or single redo to databases, is this still the case in SQL 2019?
In our AG, we have 24 cores and 13 databases. So 8 of them get parallel redo (12 parallel redo threads per database * 8 databases = 96 of the available 100 threads) and the other 5 get single. In the instance we have a mix of very and active databases and small ones. Typically the small ones get parallel redo, persumably because they start up faster.
To work around this we drop the five smallest databases from the AG, restart the replica to get the 8 remaining databases parallel redo, then add those five databases back.
Having the ability to specify the number of parallel redo threads per database would address this for us.
