Comments on: Can I get high availability by installing Exchange on a second server and replicating?

By: Wrobinson

Wrobinson — Sat, 24 Nov 2007 17:18:30 +0000

Yes, there are some good offerings from Doube-Take and CA XOsoft that will allow you to replicate Exchange databases from a source server to a target server or one source to many target servers for disaster recovery and high-availability. The steps to fail-over can be manual or automated depending on the desired configuration.

The caveat here is that these solutions are host-based and as such bears some operational overhead on the Exchange servers. They also run in kernel space and have been known to cause fatal exceptions due to high utilization.

By: enterprisephil

enterprisephil — Mon, 11 Dec 2006 03:02:54 +0000

WRT #14….
as a h/w technician, handled plenty of clients who did just that and then called to say it dont work so smooth in the rebuild or that the redundancy is lost once a hdd in the mirror pair was removed.

The scsi controller and the backplane is involved here in the intelligent control of a mirror pair hdd losing its validity and integrity. In the case of SMART errors or h/w failures, the electrical signals are conveyed back to the controller, which initiates a shutdown process to the specific hdd. Once the shutdown occurs, the hdd can be appropriately replaced.
When hdd are unceremoniously removed, without the controller being informed correctly, there is a greater chance of the existing mirror failing totally. A single drive in a mirror pair may revert to a concatenous single hdd which is unable rebuild or mirror again.
This is an ineffective and an unresilient solution for any company worth its salt.

By: adanbluearc

adanbluearc — Fri, 08 Dec 2006 13:09:26 +0000

WRT to Exchange 2003: one of the big improvements of Exchange 2003 that is relevant in this context is Recovery Groups. So basically you can restore a mailstore into the same server (but a separate temporary area) while the same mailstore is online. This is good for restoring individual mailboxes without resorting to brick-level backups. Also, this is pretty crucial in a ‘dialtone restore’ situation.

With regards to mirroring the hard drives and removing them manually while the system is running, that be the equivalent of dirty shutdown on that drive. In addition the database and the logs could be in an inconsistent state so you could have problems if you have to bring them up in a recovery situation. If you have circular logging enabled or you have logs and store files in the same drive it could be better but neither of these are recommended for other reasons. The ghost technique is superior in this respect. I think a scheduled NTBackus are better and free.

You could also use separate storage groups and mailstores and spread out your mailboxes accross them, you have up to 19 stores in total (you have Exchange Enterprise and not standard I hope). This way if one of the stores goes down you have less recovery time to face.

If you use redundant components in your server (hard drives, power supplies), have a good hardware support contract, keep your system regularly patched, install antivirus software, and follow other best practices you could get pretty good uptime. At this point, the additional cost of doing cluster to get some additional 9’s to the right of the decimal point might not be worth it in your environment.

By: brandonbates

brandonbates — Fri, 08 Dec 2006 12:25:03 +0000

In reply to #14:

Actually before I did disk imaging, that’s what I did. Even now, that is one of the ways I image systems that are online or can’t afford much down time. By mirroring the disk to another HD, yanking the mirror, imaging it (so I don’t have to tie up a hard drive) then putting it back. I have some nervous customers that just leave the extra HD on a shelf for an emergency. That in conjunction with NT Backup allows for a decently quick restore compared to a reinstall. Also more recently I’ve messed with doing a mirror to an iSCSI volume so it’s a little more hardware friendly (hot-swapping and all)

I use Paragon Drive Backup 5.5 (Didn’t like 6 as much so I still use the old one.) More recently though I have been using the Linux System Rescue CD http://www.sysresccd.org/ since I can (on most machines) boot up on the cd and image that machine to a network share over the network. Whereas with DB55 I had to have a second FAT32 disk handy to back up to (DB6 supports backup to NFTS) I still use both though.

Just always make sure you have a full NTBackup fairly recently. If the hardware goes kak you might be in real trouble if you change to different/newer hardware.

Just s note if you do the mirror thing, DONT BREAK THE MIRROR, just yank the drive (Ideally shut the system down first, otherwise at least qwell the system as much as you can, or perhaps pause/stop exchange store). If you break the mirror it may give your system drive a different drive letter (or the mirror copy) Pain in the neck to change back (a little scary too wondering if it will still boot (Voice of experience)

Yes, watch out for the PSTs on restore, the duplicate data does add up. For me to the tune of about 20%, but better than nothing and a heck of a lot easier way to forklift users & fix mailboxes)

By: AaronCutshall

AaronCutshall — Fri, 08 Dec 2006 10:04:38 +0000

One thing to keep in mind with exporting Exchange to PST files is that you loose any single storage advantages. In other words, if multiple people receive the same attached document, it is stored only once and each mailbox message merely points to it. When you export to PST files, the document will be replicated for each PST. When imported back into Exchange, each document is still stored within each individual mailbox. Hence, your storage requirements may increase after restoring from PST files.

By: JohnTWT

JohnTWT — Fri, 08 Dec 2006 08:08:37 +0000

Where’s the downside in this theory please?

I have a mirrored pair of disks which hold the system and Exchange. (Databases and data held on a different pair)

I offline one of these and stick it in the safe.

I put another disk in and rebuild the mirror.

If everything goes bang, I offline both disks and put the original disk back in. I bring that online and if all is well, add another disk and rebuild the mirror.

Just seems a bit quicker than imaging.

Anyone think this sucks and care to tell me why?

By: JohnTWT

JohnTWT — Fri, 08 Dec 2006 07:20:28 +0000

Good comments on Disk imaging and using NTBackup - belt and braces.

As it happens, I have been waging a gentle war on users to prune their mailboxes and have used ExMerge to create just the PST’s you mention. Belt and two sets of braces!

May I ask what imaging solution you use? I found Ghost 2003 on the system when I arrived so have been using that…

By: brandonbates

brandonbates — Thu, 07 Dec 2006 12:17:32 +0000

Disk imaging is a good solution (I do that), however I have been burned very badly by disk imaging. Disk images do not always restore to DIFFERENT hardware. The physical system must be based on the same storage controller for the boot drive and in some cases the same chipset, etc. The “microsoft supported method” for forklifting an os (Including applications) is NTBackup. So now I do both, I image my system occasionally (like once a year or so, doesn’t really matter). Then I run NTBackup nightly. That way I can just reimage then restore the backup, or if the hardware dies, reinstall and restore the backup. That’s why I’m moving to VMware is the hardware independence. Heck I could run the server on my desktop if I had do

P.S. It’s a good idea to do use EXMerge to backup your mailboxes to PSTs. As mentioned in the article just referenced, microsoft’s supported exchange recovery is really delicate. PSTs are much more portable (I.E. install a new exchange server, different name, domain, whatever and PSTs always restore fine. Just make sure to do a directory/AD backup as well if need be). Gives me some peace of mind. Once an exchange store goes really bad, its a long time to get it back if you ever do.

By: JohnTWT

JohnTWT — Thu, 07 Dec 2006 08:57:01 +0000

Great articles there - many thanks.

It would appear that disk imaging is a cheap and dirty solution. The more elegant solution may be to add Exchange 2007 - and the 64bit box it sits on.

Ghosting our main production server sounds scary given that it has failed on two XP Pro clients recently…

By: msfttim

msfttim — Thu, 07 Dec 2006 08:41:53 +0000

Remember that up to 90% of reported system outages are not hardware related and a large majority of Exchange Server outages are storage failure related. So that the most important item is backup. NTBackup can perform an online or offline backup to tape or disk.

I see that you have 2 primary requirements. First is reduncancy via such methods as replication. Second is little is no cost.

The Microsoft support method in Exchange 2000/2003 is clustering, which at minimun requires two servers in Active/Active or Active/Passive. (note that Active/Active configuration are not recommended), which requires external shared storage. Extra hardware cost/no software cost.

Other option is to use a second server for dial-tone recovery. This is where you mount the databases from backup (you can use NTBackup) to second Exchange Server. Time consuming/No extra hardware/software cost. For Refer to http://www.microsoft.com/technet/prodtec… and http://www.msexchange.org/tutorials/Exch…

The last option is to upgrade to Exchange Server 2007 which support Local Continuous Replication (LCR) also known as log shipping/replication.