• Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 388
  • Last Modified:

Server redundancy with identical hardware and disk swapping?

Theorectically, if I have 3 identical hardware platforms (motherboard, RAID controller, RAM, NIC etc.), is it possible to shut down the systems, swap out the hard disks (in this case 2 x RAID1), restart and the OS does not care (it will boot as normal)???

The reason I ask... I am thinking about a cheap hardware disaster recovery solution for an SBS 2003 and Windows Server 2003. I was thinking about ordering 3 identical hardware platforms, and installing on the 1st an SBS 2003 (Application server), on the 2nd a Windows 2003 Server (Fileserver) and the 3rd a Windows XP or another Windows 2003 Server (Backup server).

The only hardware difference between all would be that the Backup server would have more disk space (enough to backup the application & file server using disk to disk backup software).

In an event of a hardware failure on either the 1st or 2nd servers (motherboard, controller...), I could simply shut the relevant system and the Backup server, remove all disks from the Backup server, plug the disks from the relevant server into the Backup server, reboot and the OS would run as normal. I could then repair the 'broken' hardware in peace, my users are up and running quickly and the only services that don't run until the repair is over are those on the Backup server.

Does my theory have any holes? Is this a solution which would work and I could rely on? Are there cheap alternatives I have not thought of?

Thanks for your comments.
0
omb
Asked:
omb
  • 4
  • 3
  • 2
5 Solutions
 
Jeffrey Kane - TechSoEasyPrincipal ConsultantCommented:
Your theory seems a bit expensive and perhaps unworkable because it doesn't seem like the most efficient plan if your main goal is business continuity.  Because there are a number of steps you can take to make sure that workers will be able to keep working should your SBS have a hardware problem.

If what you are trying to protect from is a motherboard failure, why not just keep an extra motherboard on hand, without the processor... which would save you a ton of money and be much simpler to replace!  You can also keep any extra parts you are concerned about, but the parts that generally wear out are disks... don't get cheap ones to begin with, and if you're REALLY concerned... go with a RAID 5 or 6 instead of RAID 1.

You can just use the 2nd Windows Server as an additional Domain Controller which would ensure authentication for users.  

Enabling My Document folder redirection with Offline File Caching will allow users to keep working on any of their files should they lose connectivity to a server.

Outlook is automatically configured to run in Cached Exchange Mode which keeps all received email available to them until the Exchange server has been restored.

Finally instead of a separate Backup Server, use USB Hard drives for your backup.  Because it's much easier to rotate one or two off-site which is much more protective than an XP Machine holding the backup.  Becuase then you are also protected for complete recovery should the server be destroyed.  See http://sbsurl.com/backup for full overview of this.

You might also look at http://www.storagecraft.com 's new ExactState Backup which images your volumes.

Also see http://sbsurl.com/postinstall for a few other data protection ideas.

The important thing about a disaster recovery plan is to be clear about what events you are protecting from, and then you can determine what the likelihood of such an event would be and how much effort you should put into preparing for it.


Jeff
TechSoEasy

0
 
mattridingsCommented:
Does my theory have any holes?
No major holes that I can see.

Is this a solution which would work and I could rely on?

At one point in time it would have worked rather easily, I'm not positive offhand of the current algorithym that is used to calculate the hardware/software combination 'signature'  now for Microsoft on Windows Server 2003.  If it's the same as Windows XP then you will run into an issue (even with identical hardware) on bootup with the key not matching and have to have your license number reissued....which if you are going to switch back to the other machine may not be a great idea.

Are there cheap alternatives I have not thought of?

Well, yeah :)  Given the functions of your two other machines (File Server and Backup) , your desire to run SBS 2003, and your high level of paranoia regarding complete server failure then I'd probably recommend virtualization.  Create an absolute baseline installation of Windows (Server preferably, but could even be Linux, etc. if using VMware).  Then create a 'virtual' server installation of SBS....and take snapshots fairly regularly of the entire virtual server.  SBS will contain your file server, your backup, and everything else.  In the event of a 'failure' you don't have to worry any longer about matching up hardware at all, grab the last snapshot of the virtual server image and simply boot it up on any number of platforms that you have in the office.  You can use VMware's VMware Player (for free) to boot up that image anywhere you want.

See VMware or Microsoft Virtual Server:

http://www.vmware.com/
http://www.microsoft.com/windowsserversystem/virtualserver/default.mspx

Cheers,

Matt Ridings
MSR Consulting
0
 
Jeffrey Kane - TechSoEasyPrincipal ConsultantCommented:
Actually Virtual Server 2005 R2 is now freeware as well!

And that's not a bad idea.

Jeff
TechSoEasy
0
Concerto's Cloud Advisory Services

Want to avoid the missteps to gaining all the benefits of the cloud? Learn more about the different assessment options from our Cloud Advisory team.

 
ombAuthor Commented:
Many thanks for your comments.

TechSoEasy... yes, we had also thought about keeping the parts inhouse (motherboard etc..). Is a RAID6 a RAID5 with hotspare? Using the 2nd server as another DC is also nice. Cache mode for files and exchange... lots of users always synchronising the same files (network share of projects) - can this not give problems with versions etc.? Using USB hard disks as opposed to a 3rd PC... OK. Do you know how the image/DR solutions from StorageCraft compare with Acronis? Many thanks for the informative links.

mattridings... I also thought that there could be problems with swapping hard disks in identical hardware setups and that Windows might 'complain' for whatever reason. Hence, I thought I post this question with the experts.

You idea with using a virtual machine is very, very interesting! I would like to take this concept a little further, using VMware...

1. Are there peformance issues here? My thoughts would be to use Linux as a basis system (saves another expensive Windows license cost!) and run the VMware machine here. The SBS needs to be fast...

2. I assume that the 'snapshots' are done via VMware itself? Can these be setup automatically? Can the destination of these be a normal network share? How often can they take place? Is there a role-backup function, ie. can multiple versions of snapshots be kept? Are there peformance issues when the snapshot takes place? Can VMware instantly synchronise itself?

3. I assume that the SBS would still have it's own backup strategy for data? I am concerned about data (including Exchange). If the VMware takes snapshots every hour, what happens if there is a hardware failure 59 mins after the last snapshot - how is the data for the last 59 mins restored?

4. Can the whole VMware solution really be setup for free?

Sorry for all my questions. Many thanks for your comments.
0
 
mattridingsCommented:
1.  Yes, with all virtualization there will be overhead.  We tend to plan for around 25% overhead and compensate accordingly with hardware.  Unix variants tend to have less overhead as the host than Windows, just make sure you're as competent in unix as you are windows since you still have to secure the unix box and be prepared to make modifications if necessary.

2.  Yes, snapshots are handled by vmware.  As to the rest of those questions I need to refer you to their website as they have a lot of variants of vmware, some free, some not, and the abilitites when it comes to handling *multiple* snapshots varies considerably.  As to 'Synchronizing itself' , I'm not sure exactly what you mean.

3.  The SBS Server is a full normal installation so yes, it can do its own backup strategies, etc.  What we do is to run pretty typical SBS backup strategies at night, with a virtual image snapshot every morning.  In regards to hardware failures and amount of data recovered (i.e. your example of 59 minutes, if the hard drives were completely destroyed the data for the last 59 minutes would *not* be restored....if you really need a level of absolutely zero data loss in all situations you are way, way, out of SBS' league).  

4.  Nothing in life is free as they say, you pay for the hardware to compensate for performance degradation.  You pay for the time to initially set it up in addition to SBS.  But yes, if you can do what you need with their free versions of the product (Or Microsoft's) all of the virtualization software can be free and in the case of vmware run on a free OS as well.

Cheers,

Matt Ridings
MSR Cosnulting
0
 
ombAuthor Commented:
2. I am unfamiliar with VMware... Can a VMware machine snapshot itself while online (up and running) and how often? By synchronising itself, I was hoping that VMware could somehow duplicate all changes within itself, while up and running, to another destination image... just the blocks that changed would get written to the 2nd image. If the host of the source image had a hardware problem, I would just have the start the 2nd destination image on a PC and I'm up and running.

I have a final concern / setup consideration. Where would you place all the SBS data (Exchange, file data [word, excel...]), within the image, local to the 'host' or on a destination share (NAS)? Would the virtual machine simply be a SBS setup/config with the data elsewhere.

Many, many thanks for your comments on this very interesting idea!
0
 
mattridingsCommented:
Yes, the snapshot can be taken while up and running.  And it sounds like your description of synchronization is what a snapshot does (although you'd capture the snapshot locally and then move it rather than capturing it to a remote location....unless you've got a really, really, fast connection to the remote location.  If you are familiar with the windows server snapshots you can consider it basically the same thing, except in a portable, independently executable format.  I'll mention again though, just because you brought it up earlier.  One of the big differences between the commercial versions and the free versions is the frequency, etc. of snapshots....so be sure you read up on those differences and determine if your needs will be met with the free version.  I think they will once you're done studying it, but I don't want you thinking that you can just such generate a snapshot as often as you want with the free version because there are limitations.

For your second question, local to the 'host' witin the image.  You want to be able to bring up the *entire* server within the image....not depend upon other hardware being available to bring it up.  You could do it differently of course, but for your objectives you'd be sacrificing some of the reason you are wanting to use it in the first place (quick disaster recovery in case of hardware failure).  Otherwise wouldn't we also have to put in place some other form of synchronized redundancy on the data store? :)

Matt Ridings
MSR Consulting
0
 
ombAuthor Commented:
Regarding data within the VMware image, I still have slight apprehensions regarding taking snap shots of an image which could be over 300GB (we have large amounts of data!). I just thought VMware for DR of a installed and configured live OS and data through another solution (synchronisation software from 1 NAS to another???).
0
 
mattridingsCommented:
I'd have apprehensions too :)

The question is what you feel is necessary in regards to risk of data loss.  Personally I think your original statements like "what happens if there is a hardware failure 59 mins after the last snapshot - how is the data for the last 59 mins restored?" are way too restrictive for most businesses as insuring absolutely no risk of data loss would cost more than the business is worth.  It's always worth it to ask the question of the business owner, how much is the *data* loss worth (per day let's say) and how much is the *downtime* worth.

For example, one of my clients has a large call center where 100 customer service reps sit at their computers and take orders all day.  Those orders contain customer info, credit card info, etc. so you'd think protecting that *data* would be the most important....but it's not.  The client would lose much more money (24.7k per hour to be exact....we calculated) if they could no longer 'take' orders, than if they lost the last few hours of orders they had already taken.  So having the systems up to be able to take new callers money was much more important to them than getting back information on people they had already taken money from in the last 8 hours :)

In other words you need both a 'data recovery' plan, as well as a 'disaster recovery' plan.  Most people lump them together, and granted in the event of a disaster recovery you'll obviously be doing some data recovery as well. But I find that if you separate them you can much more quickly come to conclusions about how to focus the proper resources in the right areas.

That said if you want my *personal* recommendation then consider the following.  Use a combination of Windows SBS built in data recovery features for users (i.e. VSS, Exchange features, etc.) which in effect are 'Undelete' tools for shared folders and email data.  Determine the duration that users can recover their own files, email, etc. and state that to them and publish it in a major way.  That gets their comfort level up about their own personal needs and allows executives in a small business to separate the 'personal' data from that of what's needed to run and recover a business.  Something that they invariably tend to combine together and let cloud the issue.

From there focus on the best way to mitigate (not eliminate) the item with the highest cost to the business, whether that is speed of recovery (i.e. downtime expense) or amount of data loss.  When you have a plan, state clearly to the business owner, your boss, whoever what that plan *will* do and the timeframe it will do it in.  Basically like you're building a SLA.

Smaller snapshots, etc. will give you much faster speed of recovery and less impact to daily operations....but increase the amount of risk to data loss.  The more frequent, and the larger the size, of the snapshot the longer the recovery process and the greater the risk to daily operations and snapshot integrity.  Without knowing how frequently and in what volume your business critical data is changing I can't give you an exact approach.  But I can say that it's doubtful that anywhere near the full 300gig worth of data in a small/medium business is changing every day.  There's nothing that says you can't have different plans/mechanisms in place for different sets of data.  I certainly don't backup my 'clientapps' folder every day for example , but you can be sure that my Quickbooks datafile gets backed up :)

With unlimited money you could always setup a windows cluster, fiber point to point connections for near line offsite backup, mirrored raid5 clusters with snapshot splitoffs, etc.  But I doubt that's the situation.  Address the business needs with what's reasonable, then also generate a estimate for the cost of some of the items I listed above and I can guarantee you what you or the business owners would pick when given that choice.

It's a plan of choices, and they are some of the toughest choices made.  But they shouldn't be IT choices...they should be business choices that IT provides a mechanism to support.  Just don't let yourself get put into a corner of "we want no risk to data and immediate recovery....and you have a budget of 10k".  Business can only dictate one of those 2 variables, not both.


Matt Ridings
MSR Consulting

0

Featured Post

Independent Software Vendors: We Want Your Opinion

We value your feedback.

Take our survey and automatically be enter to win anyone of the following:
Yeti Cooler, Amazon eGift Card, and Movie eGift Card!

  • 4
  • 3
  • 2
Tackle projects and never again get stuck behind a technical roadblock.
Join Now