Blog Post

Exchange Team Blog
2 MIN READ

Backup^H^H^H^H^H^HRestore best practices

The_Exchange_Team's avatar
Nov 07, 2006

There is a lot of emphasis on Exchange database backups. As a result most Exchange administrators plan regular backups with considerations for their specific environment. In the process of developing a backup solution, Exchange administrators have to answer multitude of questions: Shall I protect my databases with daily Full backups, or weekly Full complemented with daily Incrementals? Shall I use VSS technology with hardware clones? Shall I include multiple Storage Groups in a single backup set? How about the IO/CPU impact? Is there enough network bandwidth to complete backups within nightly backup window? Where shall I store backups? How often should I retire the backup media? How do I monitor backups? ...
 
Successful backups are all goodness and a must have, but do you know if your backups are good enough to restore, well at least good enough to bet your job on it?   One of the horror stories I remember hearing from PSS was an Exchange administrator thinking backups were succeeding daily came to realize the good backups were all overwritten by bad ones when excessive database corruptions finally took his database offline. There was no hope of restoring from backups, and repair was not an option either. Granted a good monitoring solution (and an administrator paying attention to monitoring reports) can go a long way, but unlike good wine backups don't get better with age. A restore process that was perfect six months ago can fail to work due to changes on the backup product, changes in configuration, changes in the process, or a simple change of personnel in charge of the restore.
 
Just like you schedule regular backups, you should also schedule regular testing of your restore procedures. I am not asking you to restore all your backups all the time, neither am I telling you to restore some of your backups some of the time, what I am advising is "to restore some of your backups all the time".  When it comes to being prepared for database failures, running weekly or biweekly restores to a Recovery Storage Group is a small step for an administrator but a giant step for your users' precious data.
 
Be prepared with backups, and don't be afraid to use them.

- Ayla Kol

Published Nov 07, 2006
Version 1.0
  • Your thoughts on NT Backup?  Good enough for a small shop?  
  • I use NTBackup even for large shops. When you do Disk-To-Disk-To-Tape Backups and do some tweaks, it works pretty good! Christian
  • NTBackup can be tweaked to increase throughput, we use it for over 300 local exchange sites, then out to tape as Christian mentione above -  no problems with scripting the schedules either !

  • We do not do tape.

    We have 2 backup generations (twice a week full backups) and the backups are mirrored offsite.

    In effect we have 4 backup bytes on disk for each byte of database (56K mailboxes and 2.5TB of production DB)
  • We use local (across campas) host-based asyncronus replication, then tape backups of production that are off site, weekly/monthy/yearly.  This allows for a high availability, and for a true disaster recovery plan should we experiance a hardware failure, or worse.