An memory problem can occur if HVR has to capture changes from SQL Server 2008’s backup transaction files.
It seems to be due to a bug inside SQL Server 2008 which causes a serious memory leak.
This problem does not occur in Microsoft SQL Server 2012 and higher.
It only occurs if HVR is using the SQL read method (aka FN), and if the latency is so large that HVR must read changes from backup transaction files.
The SQL mode is used in HVR 5.0 when Capture /LogReadMethod=SQL is specified and also used in HVR 4.7 and 4.8 (as default).
The main workaround is to use DIRECT (aka FD) mode.
In HVR 5.0, define Capture /LogReadMethod=DIRECT.
In HVR 4.8 a special environment variable can be used to switch to this same method. Contact Technical Support for details.
Another workaround is to upgrade from SQL Server 2008 to 2012 or higher.
A disadvantage of DIRECT mode is that HVR needs higher (Windows system) privileges. However HVR does not support the “SQL” method with “minimal privileges” for SQL Server 2008 anyway (due a different issue).
The DIRECT transaction log read method is much faster, but requires that HVR agent is installed on the SQL Server source database machine.
It is strongly recommended to switch to DIRECT mode when capturing from a SQL Server 2008 database because the memory ballooning can kill a production DBMS system (not just HVR itself).