Best Way to Manage Failover & Redundancy

Associate
Joined
7 Oct 2012
Posts
11
Hi all,

After much advice from people here in previous threads, we have opted decided on a server solution.

We have chosen 2 x HP DL380 G7's.

These are to run 1 x instance of SBS for print & file share and 1 x exchange for our internal mail.

My idea was to run the File & Print Share on a VM on server "A", whilst running Exchange on a VM on Server "B". Then create a second VM on each machine to replicate the work of the other in the case of a failure.

Is this the best way to achieve load balancing? I don't really want to have a server sitting there waiting for a disaster, so am trying to find the best way to set them up so that if one were to fail, we could quickly have both services on the 1 server whilst the other server was being repaired.

Key questions I'm looking for input on:

  1. How many 10k SAS drives in each server.
  2. What level of RAID to use for boot and/or storage on each.
  3. How best to set up the servers so that both are working and ready to take over all services in the event of failure.

We have also chosen UPS's, 1 x Ultrium Tape Drive, 1 x 48 Port Switch, cabinet & Cable Tidy's.

We also plan to have an internal SAS drive for incremental back ups for ease of access should we need to retrieve files quicker than off the tape back ups.

Any help will be gratefully received as always.

Thanks in advance.
 
Quick post...

By SBS, I presume you mean 2011. In which case, you'd be paying out for an Exchange license to run it seperately. You'd need 3rd party software like DoubleTake or NeverFail to replicate the live server to the "backup".

I guess uptime is critical to your business, as that's a lot of money to be spending based around SBS. FWIW, HP would sell you a hardware Care Pack which is 6 hour call-to-repair contract (depending on your location).

My initial thought would be to look at a SAN for shared storage (be it iSCSI or SAS connected to the two DL380s). Then have vSphere Essentials Plus installed to provide High Availability & vMotion. You could achieve something similar using Server 2012 Hyper-V if you wanted an all MS solution.
 
Chris is completely right for the software solution, without full hardware to software layer clustering or virtual machines sitting on both boxes at the same time.

One thing you may want to watch out for is what is essentially a "dumb" cluster when you are looking for high availability / failover. Once you set anything in motion before you make it "live" try powering down one machine in tandem, and then power down the other to verify it is still available with either machine physically powered down. You don't want both boxes working together as one machine. What you want is two separate instances with replication of data between each box. Trying not to go overboard or overload you with terminology - what Chris says for vSphere Essentials Plus is probably the best way you are going to achieve this simply without tearing your hair out.

For your specific questions:

I'ld say 6 SAS drives in each box but that's just me. Depends what RAID configuration, the data you are holding, what sort of redundancy you want. For simple RAID 1 I'ld go for 3x300GB drives with one hot spare in each box.

Again RAID depends on how critical it is / what you are using it for / how much data.

Last one vSphere Essentials Plus as Chris says.

Recommend you do not use internal drive backups-
It may seem like a saviour of time / space / storage but remember these are internal to the server and internal to the building. If anything happens to the servers (are they in the same rack in the same location?) then your backups are kaput. If you can store to a network share - preferably outside the building than that is the best bet, if not tapes, if not then maybe internal backups.
 
[
QUOTE=#Chri5#;23107043]
By SBS, I presume you mean 2011. In which case, you'd be paying out for an Exchange license to run it seperately. You'd need 3rd party software like DoubleTake or NeverFail to replicate the live server to the "backup"[/quote]
I said SBS, but I should have said MS Server. We were toying with 2011 - or even 2008 R2, but at the moment, we are looking at options still.

What would you recommend for this kind of setup?

I guess uptime is critical to your business, as that's a lot of money to be spending based around SBS. FWIW, HP would sell you a hardware Care Pack which is 6 hour call-to-repair contract (depending on your location).
It's not so much uptime is critical, but more that in the longer term, we will be replacing other servers. The thinking is that if we get 2 of these units now and virtualise everything, when we move our primary business DB and business intelligence in a year or so, we can adapt what we have in place already by upping the RAM and adding a second chip.

My initial thought would be to look at a SAN for shared storage (be it iSCSI or SAS connected to the two DL380s). Then have vSphere Essentials Plus installed to provide High Availability & vMotion. You could achieve something similar using Server 2012 Hyper-V if you wanted an all MS solution.

That sounds perfect. I would be happy to use 2012, but what software/hardware requirements would there be. If you have time, please be as specific as possible (my knowledge is still basic). As with all my posts, I'm totally open to suggestions.
 
One thing you may want to watch out for is what is essentially a "dumb" cluster when you are looking for high availability / failover. Once you set anything in motion before you make it "live" try powering down one machine in tandem, and then power down the other to verify it is still available with either machine physically powered down. You don't want both boxes working together as one machine. What you want is two separate instances with replication of data between each box. Trying not to go overboard or overload you with terminology - what Chris says for vSphere Essentials Plus is probably the best way you are going to achieve this simply without tearing your hair out.
Hi Jake,

I see what you are saying.I would be more than happy to have a setup that doesn't necessarily take over automatically, but requires a quick transfer of the VM from the backup of the failed machine. The main things are that we're getting the throughput from both units, and that we're one time fail, the other could take on the role of the failed unit in a relatively short space of time.

I want to keep things as simple as possible really.
For your specific questions:

I'ld say 6 SAS drives in each box but that's just me. Depends what RAID configuration, the data you are holding, what sort of redundancy you want. For simple RAID 1 I'ld go for 3x300GB drives with one hot spare in each box.

Again RAID depends on how critical it is / what you are using it for / how much data.

Last one vSphere Essentials Plus as Chris says.

Recommend you do not use internal drive backups-
It may seem like a saviour of time / space / storage but remember these are internal to the server and internal to the building. If anything happens to the servers (are they in the same rack in the same location?) then your backups are kaput. If you can store to a network share - preferably outside the building than that is the best bet, if not tapes, if not then maybe internal backups.
The internal backups would be in addition to tape backup. The tape backup would be the off- site. We're just thinking fir file restores and recovery, having an incramental internal data backup would save time in that we wouldn't have to go to the tapes in the first instance.

Going back to the RAID setup, would a RAID 1 on the bootable drive with a 3 drive RAID 5 setup on data be a sensible route?

Data isn't business critical, but loss would be incinvenient.

Ideas welcome as always.
 
How much money do you have to spend, and how much are licenses for things like windows server 2008 std / enterprise for your organisation?

Also, how much hassle do you want in administering it?

You will probbaly want to put in a HP MSA shelf for shared storage and either go down the microsoft route of virtualising or vmware route. You would need to get costs for both from your vendors.
 
Are both servers in the same location? Guessing so if you've only got one switch.

I'd budget for another, and another connection into the network if possible. All this redundancy and the like is all well and good, but if you've got a single point of failure in the network switch then it's all a bit moot.

Some great advice in this thread btw. :)
 
How much money do you have to spend, and how much are licenses for things like windows server 2008 std / enterprise for your organisation?

W don't have a fortune to spend, but if it were circa £10k (give or take...) inc software and licences, that would be acceptable.
Also, how much hassle do you want in administering it?

As little as possible really. We're hoping once the initial setup is done, most admin tasks will be automated. As such, we're looking for the simplest effective solution.


You will probbaly want to put in a HP MSA shelf for shared storage and either go down the microsoft route of virtualising or vmware route. You would need to get costs for both from your vendors.

They're quite expensive, but would consider. Am I right in thinking we would then have 2 x SAS drives in each Server to boot from and run the VM's - with all storage being held in the MSA?

If so, would a RAID 5 3 x 320GB setup be sufficient to go in the MSA? If not, please expand a little further if you would...

Thanks in advance....
 
I am no expert in pricing up this type of stuff, i work for a large enterprise and stuff just turns up when we say what we need!
But from looking on google you could buy the msa for £2k, and two windows 2008r2 enterprise licenses for £3-4k.
Your servers would have a raid 1 and hold only their C:\. The MSA would have the VM's and both servers would have access to them.
You then need disks for the MSA, no idea how many you need and what capacity. You need to think about what your current needs are and the needs in 2-3 years and plan for it now.

There are loads of people on here with lots of this type of experience and i'm sure they can suggest someting much better / cheaper.
 
Thanks for that Showboat, I really like that suggestions. I think a RAID 5 @ 3 x 320GB for storage would be adequate for both exchange and file storage - and the various plans above should be sufficient to protect that data. One more question:

If we did the above, then migrated other software to new VM's on our new servers, could I just add more SAS drives to the MSA and create a secondary 'cluster'?

We have 25 users all requiring Server Access and Exchange Access. Anyone have any ideas of the cheapest way to licence everything? Or any alternatives to the above that would be more cost effective?

I think we're almost there!
 
I would suggest:

OS Disks: 2x 146GB 10K or 15K in RAID1
Data Disks: 4x 300GB 10K in RAID10 (fast, ~600GB useable space, maximum of 2 dead disks if they are in separate pairs) Beware if you intend to use all 8 drive bays on the DL360, I'm not sure this is possible using the standard RAID card

OS: Hyper-V 2012
£0

Failover/DR: Hyper-V Replica with dedicated NIC port going directly to the other DL360 (switch not necessary)
£0 - included with Hyper-V 2012
 
Just to add to that, you would need a Windows 8 host to manage the Hyper-v servers properly until the Windows 7 Hyper-v console gets an update. Also the failover would be manual (start vms) unless you script it.
 
Thanks for that SoundsGood, but I am a little confused.

In your suggestion, would server B be doing anything, or would server A be doing all the work with the second server merely being there for failover?

Also, with the win8, do you mean any computer with Win8 would be required to use the Hyper V console. As in, so long as the workstation that's accessing it has Win8, it's fine.

Apologies for the lack of understanding. I've learned a lot here, but still get confused by aspects. If you could detail the setup of both servers and their respective roles,I'd sincerely appreciate it.
 
In the example no, server B would be simply replicating server A (every 5 minutes) then if server A dies you simply bring up the VMs on server B. Note this won't really help you if director x deletes half his files and wants them back, you will need a full backup solution for that.

I'm not sure if it's possible to split the load between them and still use failover, never tried it. However if you only have 25 users a single DL360 won't break a sweat.

For managing the server you can use the Hyper-v console in Windows 7 but the new features will not be visible, including the replica feature. Apparently there is an update due out but for now you would need Windows 8 or a full fat Server 2012 install.
 
I'm with you now- thanks for the clarification.

Following on from it, would it be viable to do the same thing, but instead of having Server B redundant, use an MSA and have the VM's replicating to that.One of the reasons I'm reluctant to have Server But idle is we plan to migrate other servers accross to the 2 units over the next 12-18 months. And while 1 server will handle exvhange, print & file sharing for 25 users coomfortably, I think it would struggle with the rest on its own.

Would the above make sense?

Also, any tips for advice on efficient licensing? 2008, 2011, or 2012? Cheapest way to go about it, or a company who can advise/sell?

Sorry to be a pain!
 
What OS will your VM's require?

If it's windows, then you can't use the free version of hyper-v.

Aidan Finn -- a Microsoft Valuable Professional (MVP) with an expertise in Virtual Machine who works for MicroWarehouse Ltd. -- alerted me to the fact Hyper-V 2012 is generally available. He noted in a September 4 blog post that there are "no free virtualisation rights to install Windows Server in guest OSes" with this product. "This means it is good for labs, VDI (virtual desktop infrastructure), Linux hosting, and upgrading older hosts without SA (Software Assurance." In other words, it is pretty stripped-down, but may be the right choice for particular workloads.

Taken from :
http://www.zdnet.com/windows-server-2012-spinoffs-what-else-is-microsoft-shipping-7000003715/


Looks like you could buy the standard 2012 which includes clustering for £500, but you would be limited to 2 VM's so if you will ever grow it may not suit!

http://searchservervirtualization.t...ver-2012-removes-licensing-option-for-Hyper-V
 
If you go for vmware essentials plus you can use the vsphere storage appliance as you shared storage so no need for msa.
 
I'm with you now- thanks for the clarification.

Following on from it, would it be viable to do the same thing, but instead of having Server B redundant, use an MSA and have the VM's replicating to that.One of the reasons I'm reluctant to have Server But idle is we plan to migrate other servers accross to the 2 units over the next 12-18 months. And while 1 server will handle exvhange, print & file sharing for 25 users coomfortably, I think it would struggle with the rest on its own.

Would the above make sense?

Also, any tips for advice on efficient licensing? 2008, 2011, or 2012? Cheapest way to go about it, or a company who can advise/sell?

Sorry to be a pain!

If you have an MSA then I don't think you need to replicate the data as it's already accessable by both devices, this also means if the MSA dies you have no VMs.

I had a quick play with this Replica feature today, you can specify the replicate destination for each VM individually so you could have exchange on one and file&print on the other like you wanted and still have copies on both.

Email my trust if you want a good licensing/reseller recommendation (competitor rules)
 
Back
Top Bottom