| Author |
 Topic  |
|
|
anthony
Moderator
USA
2373 Posts
Status: offline |
 Posted - 05/14/2012 : 3:48:34 PM
|
We are a small Exchange Organization of about 100 users on one server. However, due to a previous Exchange failure, Managment wants the email accounts to be distributed across multiple databases.
Almost a year ago, we had an Exchange failure, and due to the DB's large size, it took forever to do restores/rebuilds and all those other things that happen when Exchange bites the dust. Replaying logs took a long time, restores took a long time, EVERYTHING took a long time. Because it was all in one big DB.
Even though we are in a single server environment (at least we will end up that way in a week or so), they want to divide the mailboxes into 4 databases in hopes that this stuff will not be so cumbersome if it happens again.
We are wanting to distribute the DBs so that they are semi-equal in size, and my hope is that while we create new users - all things being equal (DB is healthy, in same AD domain, same site, same server, etc.) it will choose the smallest DB...
Any such luck? 
According to what I have read, it is just "random" after the above is determined. What have you guys seen?
Thanks,
|
anthony
There should be only one World's Greatest Dad shirt. And you should have to kill the previous owner to wear it.
|
|
|
Jazzy
Administrator
Netherlands
1929 Posts
Status: offline |
Posted - 05/14/2012 : 4:30:18 PM
|
Man, that sucks. Anyway, you can use a script to balance the mailboxes across the databases: http://www.stevieg.org/2010/09/balancing-exchange-databases/
On the other hand, you have 4 databases with 25 mailboxes each. How hard is it to check the sizes now and then and move a some mailboxes to another database? I mean, your management prefers to increase the overhead because it gives them a (false?) sense of security anyway. :) |
Jetze Mellema
Exchange specialist
Former MVP (2005-2012)
My blog: http://jetzemellema.blogspot.com (Dutch)
My company: http://www.imara-ict.nl/ |
 |
|
|
anthony
Moderator
USA
2373 Posts
Status: offline |
Posted - 05/14/2012 : 4:35:54 PM
|
| How much more overhead will it create? Is it something I should be concerned about? |
|
|
|
Jazzy
Administrator
Netherlands
1929 Posts
Status: offline |
Posted - 05/15/2012 : 02:32:55 AM
|
| No, don't worry. I meant stuff like we're talking about now, usually there's no need to worry about equal database load or management because you can fit 1000 mailboxes in a database easily. Also I think that you're planning for a single LUN per database and one per LUN logfiles so you can do a proper backup and restore, this means you need to keep sufficient free space in all LUNs in stead of LUNs for a single database. That's all. |
Jetze Mellema
Exchange specialist
Former MVP (2005-2012)
My blog: http://jetzemellema.blogspot.com (Dutch)
My company: http://www.imara-ict.nl/ |
|
|
|
NMDANGE
Honorable But Hopeless Addict
USA
2054 Posts
Status: offline |
Posted - 05/15/2012 : 09:42:34 AM
|
| If management wants a more reliable system, tell them you should have 2 Exchange servers! |
Michael D'Angelo
(former)MVP-MIIS, Pace University Senior Systems Administrator (Windows)
(MS)NMDANGE
PhoeniX WorX Systems Administrator. If you play Total Annihilation, please join us. http://www.phoenixworx.org |
|
|
|
Endaar
Old Timer
USA
567 Posts
Status: offline |
Posted - 05/15/2012 : 11:31:00 AM
|
| I'm curious...how big was the DB and what type of backup was in place? |
|
|
|
anthony
Moderator
USA
2373 Posts
Status: offline |
Posted - 05/15/2012 : 1:11:03 PM
|
When we initially had the failure? It was probably around 80GB, and we were using StorageCraft for backups. StorageCraft is an image based backup software much like DPM, or Acronis. We had it taking a snapshot every 15 minutes.
This failure occurred about 3 days before my start date - so I walked into it. But apparently, the database was corrupt, and had been for a while. We were restoring the DB from the last backup and replaying logs, and it would fail to mount, and fail to mount, and fail to mount. We did this about 10 times. This whole time we had Microsoft PSS assisting. Each time going back another backup interval. The first few times we went back 15 minutes, 30 minutes, then an hour, then 6 hours. Then, we went back a day, then 2 days, then 4 days. still no joy. We ended up going back almost 3 weeks! Yes, we lost 3 weeks of email...
That was a painful period, and not the way one wants to start a job. If you guys remember, this was a while ago, and this is what caused a cascade of events that led us to Exchange 2010, and a complete forest migration/domain change. As you can see, we are still cleaning up from the mess it created.
|
|
|
|
Endaar
Old Timer
USA
567 Posts
Status: offline |
Posted - 05/15/2012 : 1:19:48 PM
|
Wow, sounds like a total mess. It's interesting that Exchange ran with a corrupted DB for three weeks, but once it got to the point where it was completely non-functional, it wouldn't mount a DB backup that in theory should have been no more corrupt than the DB that was being used previously.
That being said, how does splitting mailboxes across different DBs prevent this? Is management just thinking that if a DB corrupts in the future it would only be that one DB, not everything? The delays in getting back up and running seemingly stemmed from the corruption, not the size of the DB...
Sorry I don't have any suggestions. Good luck with whatever way you choose to go.
James
|
|
|
|
anthony
Moderator
USA
2373 Posts
Status: offline |
Posted - 05/15/2012 : 1:27:35 PM
|
Right.... it only helps one thing. During our outage, every time we would do a restore, then the subsequent log replay, it took like 6+ hours. Especially once we were going way back 2 or three weeks. So the logic was that in the event a database got corrupted, that process would one - take less time, and 2 - not have the whole org down during the process (only the folks that were in that database would be down).
Of course my argument was that next time, corruption might not be the issue. So it only protects us from that ONE single thing. Nothing else really. But we also are no longer running anything until the hardware dies either (which was happening too before I got here). We were pretty suspect of the hardware all this was running on at the time too. It was an "Intel Whitebox" server built by a local consulting company. Nobody had a firm answer on its age. |
|
|
|
NMDANGE
Honorable But Hopeless Addict
USA
2054 Posts
Status: offline |
Posted - 05/15/2012 : 4:43:15 PM
|
| I've done Exchange restores in DPM with 200GB+ databases and they don't take nearly that long to restore. |
Michael D'Angelo
(former)MVP-MIIS, Pace University Senior Systems Administrator (Windows)
(MS)NMDANGE
PhoeniX WorX Systems Administrator. If you play Total Annihilation, please join us. http://www.phoenixworx.org |
|
|
| |
Topic |
|
Automatic Database Distribution
