You are viewing a plain text version of this content. The canonical link for it is here.
Posted to derby-dev@db.apache.org by "Bergquist, Brett" <BB...@canoga.com> on 2011/12/02 20:14:25 UTC

Question about online backup

We currently use the SYSCS_UTIL.SYSCS_BACKUP_DATABASE procedure to perform an online backup of the database.  This has been working well.  We are wondering however, if we can control the relative priority of this backup to have less of an impact on the normal concurrent operation of the database.

Here is the use case.   The system is being accessed and updated 24/7 and there is no down time.   The customer wants to create a backup once a day.  There is no specific point in time that they are trying to capture so knowing exactly which transactions are included or excluded from the backup is not that important.  They would use the backup only for a catastrophic failure to get "close" to what something was on a day and a higher level system interacting with this one would replay anything necessary after restoral of one of these backups.  In the 5 years that the system has been running a restoral of a backup has never been necessary.

What is important is that the online backup being performed does not adversely affect the normal operation of the system.   There are plenty of CPU cycles and should be plenty of I/O bandwidth.  This is an Oracle M5000 system with 32 cores and 32 Gb of memory with mirrored  drives using ZFS as the file system.

So what I was wonder is, is it possible to perform the backup at a lower relative priority than normal database operations.  It is okay if the backup takes longer, that is not a problem.

Any guidance will be greatly appreciated.

Brett