We have an Azure setup with two VM’s running SQL Server 2014. The servers are placed in defferent domains with one server in West Europe (Master) and the other in North Europe (DR). On the master we have an automated process that runs backup and copies the trn logs to an Azure storage account in North Europe. On the DR server we have an automated process that collects trn logs from the Azure storage account and restores on the DR server. The database is in standby mode allowing selects. Sometimes at random hours the restore fails with the error message below, putting the database in a restoring state. If I delete the database and do a restore of the last full backup and alle the preceeding tlog files, including the failed one, the restore is not reporting any error.
Image may be NSFW.
Clik here to view.
Image may be NSFW.
Clik here to view.
If I look at the content of the Tlog file and compare it with a working file for the identical hours bur a different day I get the picture below. The first one is the working TLOG and the second is the one that failes. The difference is that working one has 136 records and the failing has 9. They are identical except for transactionid 0000:00000000 up to the line LOOP_SHRINK_NOOP where the failing one stops and the working one has rows for LOP_BEGIN_CKPT, LOP_MODIFY_ROW and a lot of rows with LOP_COUNT_DELTA. For the two last records (not showing in the picture) there is a LOP_XACT_CKPT and a LOP_END_CKPT.
I have checked the files for most of the times the restore fails and this difference in the TLog is present at every time.
What is happening and how to debug?
Image may be NSFW.
Clik here to view.