Backup compression issue after failover to AG replica

Hi all,

We have a 2-node AG cluster with one primary replica and one non-readable secondary replica. To the best of my knowledge, these servers are identical in every way - same SQL version, Windows version, patches, hardware, drives etc.

Recently we failed over to the secondary node and after a few days found that our backup drive was filling up, which had not previously been the case on the original primary node. When comparing the backups on the two nodes, the ones from the original primary were roughly 5-20% smaller than those performed on the secondary node (it varied slightly for each DB). The issue appears to be that the backup compression is not quite as effective on the secondary node - the question is why would that be when they are identical servers?

My only thought was that the way the data was physically stored on disk may be slightly different between the two - i.e. the data pages may be arranged differently, which could affect the efficiency of the compression.

Does anyone have any other thoughts?


Are you using the same MAXTRANSFERSIZE parameter for the BACKUP command on both servers? Differences with this can be particularly noticeable when TDE is enabled.

Hi Ifor, thanks for your reply! No, unfortunately we're not using the MAXTRANSFERSIZE parameter and our databases are not using TDE either. :confused:

Can you confirm if the backup compression instance level setting is same on both the nodes?

Hi, yes the instance-level compression setting is the same for both nodes, although we're explicitly using the COMPRESSION option when running the backups regardless.