OT: MS Exchange backups

[Ugo Bellavance]

Denis Beauchemin wrote:
> Gerry Doris a écrit :
>> shuttlebox wrote:
>>> On Nov 23, 2007 10:04 PM, Denis Beauchemin
>>> <Denis.Beauchemin at usherbrooke.ca> wrote:
>>>  
>>>>>> Manager (TSM) and Backup Exec but couldn't backup our 250GB data
>>>>>>         
>>>> I'm not sure I understand your question...  it takes 48 hours to backup
>>>> to disk!
>>>>
>>>> Of course, we can't afford to stop Exchange to either backup or restore
>>>> data.  Everything has to be done online...
>>>>     
>>>
>>> Something is very wrong. We use Legato Networker on Solaris and it
>>> backs up many TB during the night including at least six Exchange
>>> servers each with larger databases than yours. Of course we spread the
>>> full backups so most servers only gets a diff or inc backup but still
>>> the full backups never takes more than a few hours.
>>>
>>>   
>> We generally install either Legato or Veritas and backup many Exchange 
>> servers/TBs overnight.  Check to make sure that the servers are being 
>> kicked off at different times...don't try and start everyone at 
>> midnight!  250GB is nothing. 
> I agree with all of you. It should not take so much time...  disks are 
> on a local raid5 box.  I think the time is spent extracting people's 
> mailboxes from the db and copying them to disk...  but still, it takes 
> way too long...

You should check the queue length on all your disks when you start the 
backup.  How is this RAID5 array connected to the exchange server? On 
the same controller as the rest of the disks? Are they on the same disk? 
  If you have a dual NIC on this server, you can have a separate box 
connected with a crossovercable, running 2X 500 GB (or more) HDD in a 
RAID 1 config.  You'll have better write speed this way, or with a 
RAID10.  What I mean is that your bottleneck is maybe the reads on the 
live filesystem, maybe the write on the backup media.

> 
> How do you restore a single mailbox or some emails?

Using Backup Exec, with brick-level backups.

>  Do you have to 
> restore the hole db and then extract the relevant emails?  If so, it is 
> a waste of time and resources...  unless there isn't any other way...  I 
> think this is what they are trying to avoid but obviously it is not 
> working very well...