Exchange Using ISCSI SAN Defrag? 2

[tekinfmgr]

I thought I would drive this by and see how it looks to you all. It was recommended to us to disable Exchange defragmentation on the databases that are running on a ISCSI SAN. This is keep the changes from growing to fast on the snapshopts.

Is anyone else doing similar? I am scared really in not allowing Exchange defragmentation to occur. The databases will probably still grow because they cannot delete items and recover white space to write new translations

 

[xmsre]

No... you shouldn't disable online defragmentation. Out of curiosity, which consultant/misguided individual told you that? Disabling online defrag is a very bad thing. That said, overly aggressive online defrag does bloat the size of your snapshots by causing unnecessary churn. The iSCSI SAN part isn't important (nice, but not germane to the discussion). What is important is that you're using snapshots. The size of a snapshot is equal to the sum of the blocks changed since the last snapshot, or your change delta.

The MS guidance for Exchange 2003 is that online maintenance (including the online defrag) should complete at least once every week or two weeks. In Exchange 2007 SP1, MS added counters and a new method to determine the optimal frequency of online defrag. If you're truely curious, you'll find a description here:

http://msexchangeteam.com/archive/2007/12/06/447695.aspx.

What you should do is gradually cut back on your maintenance schedule (the time allocated each day, not the number of days) and watch the 1221s in the application log. Keep cutting back until you complete in the once every 7 - 14 (but not over 14) days range. I've seen the 24 hour change delta go from 30% to 6% in some cases simply by doing this. Online defragmentation isn't the sole cause of change delta (your users do send and receive mail, don't they?) so YMMV.

 

[58sniper]

A star for the excellent answer. I should aspire to answer things so completely.

Pat Richard MVP

 

[tekinfmgr]

Thanks for your quick response, it is my gut feeling he misguided or achieved some poor communication on this part. I am checking back with him and getting ready to turn on the defragmentation again. We put on 13 GB in the databases in just over a week, I assume that this is could be a factor. It just did not seem to me like the right thing to do. He is a storage expert, but maybe not so much on Exchange.

 

[xmsre]

The MS link states the 2003 guidance, the new 2007 SP1 guidance and goes into a little detail. You'll probably want to communicate that back. Your storage vendor may have a best practices doc as well.

Storage admins, and vendors for that matter, tend to take a very storage centric and myoptic view that often doesn't account for the nuances of the application workload placed upon the storage. It's not that storage admins are inherently evil or anything, it's just that there are thousands if not millions of applications out there running on scores of OS platforms. It's not possible to be an expert on every one of them. For that set of applications you've standardized on in your organization, it would be worth the effort to educate your storage team on them.