Re: Offline defrag

John Smith wrote:
I have stopped making brick-level backup, and to backup 600gig (cross 8 mail stores) databases took 5 hrs.
this is backing up disk staging system.
So what is different from the time when you were getting 6MB/s?
but I still want to get to bottom of what is causing message level backup to be soooooooo slow. any ideas on things to investigate to resolve this issue would be much appreciated.
Take a look at system utilization during BLB, specifically disk IO/ps
my largest mail store database is 210gig with 150 gig whitespace, there are some mail boxes with corrupt items preventing mailbox move in this store. what is the best way to reclaim this much space.
If BLB reports corrupt item, it indicates problem with backup software, not with mailbox.
What happens when you move mailbox?
I have also notices the online defrag always terminated because of back up interference, so I moved the online defrag back to 2am - 6 am.

Set online defrag to end later, just before you start business day, so that it has as much time as possible.


"Kirill Palagin" <kpalagin@xxxxxxxxxxxxxxxxx> wrote in message news:uqPLq$a1GHA.4748@xxxxxxxxxxxxxxxxxxxxxxx
See inline.

John Smith wrote:
it is a company policy that we are able to restore message / folders as quick as possible, that is why we have invested a lot of money on disk staging systems. we keep the exchange backup on disk for a week for the images are expired. I am trying to trace what is causing the backup performance hit. at the moment we are think it has something to do with exchange.
"As quickly as possible" is not very specific. Exact time-frame would be much better. Besides, brick-level backup strategy apparently not working.

however there is noway to prove this, if there was overall system performance, a lot of users would have complained. at the moment the throughput is less then 1 meg for brick level backup and 6meg for IS backup. backing up 680gig database at the speed will never finish in time.
Both values seem to be low to me - I expect at least 3 times faster. How your hardware and all setup looks like? Also, you do not have to backup 680GB all at once - if you split data into, say, 6 databases, you will be able to backup each in approx 12 hours.

we have imposed storage limits, but people already have huge mailboxes.

how can I do some test to make sure everything in exchange - disk speed, network link throughput, read/writes of disk system - is working fine without dismounting the stores.
ExBPA.
http://www.microsoft.com/downloads/details.aspx?familyid=DBAB201F-4BEE-4943-AC22-E2DDBD258DF3&displaylang=en

I didn't quite understand this "(but white space is not backed up during brick-level backup)." but are they backup during IS backup, I think they are backup on full IS backup.
Brick-level backup reads each individual item (message, contact, calendar item etc) in each mailbox and saves this item (which is slooooow, as it requires many index page reads and sequential reading of the item parts, like attachments). Since white space in the database does not belong to any item BLB can not read it.
OTOH, Store-level backup just reads every page in the database sequentially and saves this stream (which is so much faster). It does not skip unused pages (white space) of the database.

Online defragmentation is always terminated according to the event log, this may have something to do with backup schedule times, can I run online defrag during the day and how much performance hit am I looking if I was to attempt this.
This could be a problem too - one of InfoStore maintenance (aka online defrag) tasks is to make freed pages ready for reuse. Not allowing it to finish will cause white space to grow. You can allow it during business hours (or just set it for full day).

"Kirill Palagin" <kpalagin@xxxxxxxxxxxxxxxxx> wrote in message news:%23XpmodZ1GHA.4264@xxxxxxxxxxxxxxxxxxxxxxx
Brick-level backup is infamous for being slow, redundant and troublesome (but white space is not backed up during brick-level backup).

Switch to Store backup and you may get 30GB/h and up transfer rates. After performing successful online backup you can worry about reducing white space (you can only reduce it, but never eliminate). You can reduce it either with the help of ESEUTIL or by moving mailboxes to the other database, whacking .edb+stm, moving mailboxes back. I strongly recommend ESEUTIL approach.

I also would suggest imposing storage limits.

File-system defrag will not make a difference because database access is random by nature, not sequential.

John Smith wrote:
we have roughly about 500 users, the last backup took 58hrs to complete and it didn't backup all the storage groups. we are performing a brick level backup. To give some idea on how our backup works, we backup to disk over night, and in the morning this is staged off to tape. however we have not had a successful backup for couple days because the throughput is sooo slow. sometimes we see multiple backup schedule running because the last backup didn't complete the allocated backup window.

just to make things clear, the total sum of our database are 680gig, our biggest database 210gig with 150gig whitespace ( Ouch ). the rest of the database are between 50-100gig and each database has roughly 12gig of whitespace. Going through the event log and filtering for event id 1221 and adding up all the whitespaces, we can roughly save 230gig of space and reduce the databases of 380gig total.

I can handle the whitespace issue if the backup performance is quicker, but at the moment backup is slower than snail with arthritis.

can I also ask if file system defrag would improve anything, if so, would this require the stores to be dismounted.


"Kirill Palagin" <kpalagin@xxxxxxxxxxxxxxxxx> wrote in message news:ODv0vFZ1GHA.1288@xxxxxxxxxxxxxxxxxxxxxxx
Having 10GB of whitespace in databases this big should is expected and should not be cause for offline defrag.
How many users you have and how long backup takes? Hopefully, you are not doing brick-level backup.


John Smith wrote:
Hi all,

I am new to exchange, we have recently migrated from lotus notes, where users had unquoted mail files. When we have moved to exchange 2003, we have decided to introduce quota with the mail system. so we've created couple mail stores - 8 in total - and moved users about.

since we've moved users about, out database grew from 370gig to 680gig. I have checked the event log and there is a lot 1221 event id, it seems that each mail store has approximately 10gig of free space.

the other thing, our backup has also taking hit, it is extremely slow, it takes forever to backup. the exchange through put for the backup has takes a huge performance hit. I think the backup time interferes with online defrag, I know the obvious thing to do is run the backup another time, but because of the size of the databases being backed it is continuously backing 24/7.

I have heard about eseutil, but alot of people are advising against it, I don't have a recent backup either, I want to fix all this issues before it becomes a nightmare.

please advice.
--

Text from most Windows dialogs can be copied to clipboard with Ctrl-INS.

Free productivity applications suit - www.openoffice.org
Free Internet calling - www.skype.com
Free SQL database Firebird - full support for transaction control, triggers, stored procedures, partial SQL-99 compliance
http://www.ibphoenix.com/main.nfs?a=ibphoenix&s=1142758270:704186&page=what_is_interbase
--

Text from most Windows dialogs can be copied to clipboard with Ctrl-INS.

Free productivity applications suit - www.openoffice.org
Free Internet calling - www.skype.com
Free SQL database Firebird - full support for transaction control, triggers, stored procedures, partial SQL-99 compliance
http://www.ibphoenix.com/main.nfs?a=ibphoenix&s=1142758270:704186&page=what_is_interbase
--

Text from most Windows dialogs can be copied to clipboard with Ctrl-INS.

Free productivity applications suit - www.openoffice.org
Free Internet calling - www.skype.com
Free SQL database Firebird - full support for transaction control, triggers, stored procedures, partial SQL-99 compliance
http://www.ibphoenix.com/main.nfs?a=ibphoenix&s=1142758270:704186&page=what_is_interbase
Free graphic image manipulation program - www.gimp.org



--
Text from most Windows dialogs can be copied to clipboard with Ctrl-INS.

Free productivity applications suit - www.openoffice.org
Free Internet calling - www.skype.com
Free SQL database Firebird - full support for transaction control, triggers, stored procedures, partial SQL-99 compliance
http://www.ibphoenix.com/main.nfs?a=ibphoenix&s=1142758270:704186&page=what_is_interbase
Free graphics image manipulation program - www.gimp.org
.

Relevant Pages

  • Mit isql mehrere DBs sichern und wiederherstellen
    ... Script habe ich soweit geschrieben aber nach dem ersten Backup steigt ... Hier mal mein Script: ...
    (microsoft.public.de.sqlserver)
  • Re: performance degradation during backup
    ... disk array died you would loose the database and your most recent backups. ... >>> Whenever I run database backup during normal operating ... The database size is 40GB and the backup ...
    (microsoft.public.sqlserver.server)
  • RE: How do I restore from mdf and ldf files?
    ... Tasks -> Backup Database ... Tasks -> Restore Database ... Windows 2003 Server with Latest Service Pack ... Pre-requisites for Sharepoint Backup and Restore: ...
    (microsoft.public.sharepoint.portalserver)
  • Re: Uncommitted Transaction Log limit before stores dismount
    ... I do a backup on EX1 and the backup takes several hours and the backup ... locks the database as per design. ... when the number of UNCOMMITTED logs reaches 1008 Exchange IS ... If you are using Exchange Enterprise you can have stores as big as you ...
    (microsoft.public.exchange.admin)
  • Re: RMAN restore of full backup fails completely
    ... You are not using a catalog database. ... You might have autobackup controlfile on, ... Your autobackup has been made *prior* to the backup! ... plus archivelog', when i try a restore on the other server however, the ...
    (comp.databases.oracle.server)