Mark Minasi's Reader Forum
Mark Minasi's Reader Forum
Home | Profile | Register | Active Topics | Active Polls | Members | Search | FAQ | Minasi Forum RSS Feed
 All Forums
 OTHER HALP! Linux, Hardware, and Anything Else
 All things Virtual!
 HA on Hyper-V... Cluster HOST or GUEST?

Note: You must be registered in order to post a reply.
To register, click here. Registration is FREE!

Format Mode:
Format: BoldItalicizedUnderlineStrikethrough Align LeftCenteredAlign Right Horizontal Rule Insert HyperlinkInsert EmailInsert Image Insert CodeInsert QuoteInsert List

* Forum Code is ON
Smile [:)] Big Smile [:D] Cool [8D] Blush [:I]
Tongue [:P] Evil [):] Wink [;)] Clown [:o)]
Black Eye [B)] Eight Ball [8] Frown [:(] Shy [8)]
Shocked [:0] Angry [:(!] Dead [xx(] Sleepy [|)]
Kisses [:X] Approve [^] Disapprove [V] Question [?]

Note: please do not cross-post.
Cross-postings will be deleted and ignored.
Thanks for helping to keep this forum junk-free!
Check here to subscribe to this topic.

T O P I C    R E V I E W
anthony Posted - 11/08/2011 : 4:58:34 PM
Which HA solution is better? Clustering the HOST Hyper-V Server? Or the Guest OS? Or is it even possible to cluster the HOST? I mean if the HOST is clustered, would that not allow you to pick which guests run which hardware.
25   L A T E S T    R E P L I E S    (Newest First)
Pesos Posted - 12/10/2011 : 11:18:54 PM
Aidan is correct, it is now supported to live migrate Exchange 2010 DAG members. We run our DAG boxes on the local storage of our hyper-v cluster boxes though, taking SAN failure out of the picture.
joe_elway Posted - 11/15/2011 : 2:40:59 PM
Originally posted by anthony

Then again... is MS going to show and ask to see how many hosts are running in each VM?

If you have volume licensing, then don't be surprised that they do. Globally, they have been recruiting and training specialist auditing partners. -Correct- virtualisation licensing is one of the targets ... lots of people using VMotion/DRS/HA for Windows Standard SKUs more often than once every 90 days, and lots of people abusing the 4 VM limit of Enterprise.

Here, we've found that once you hit 7 VMs on a 2 socket/CPU host, it's more economic to switch to Datacenter licensing.
anthony Posted - 11/14/2011 : 4:37:56 PM
I was just kidding... yep. That's what I was doing... kidding.
wobble_wobble Posted - 11/14/2011 : 4:31:00 PM
I cannot answer that question, that's down to what is going on in the environment, disgruntled employee's, sheer bad luck and maybe an MS LAR checking the Minasi forum....

anthony Posted - 11/14/2011 : 3:45:59 PM
Then again... is MS going to show and ask to see how many hosts are running in each VM?
wobble_wobble Posted - 11/14/2011 : 3:44:19 PM
Originally posted by anthony

OK. Licensing question...

Let's say I buy 3 copies of Enterprise R2 to do a 3-Node cluster. I can run 4 copies of Windows on top of that as part of the SKU. If those servers are in a cluster does that mean those copies are aggregated across the 3 node cluster? What if for load balancing reasons I need to run only two copies on one? Can I run 6 on another?


The Enterprise license is applied to a server, therefore you get 4 "free" Windows guests on that server. Two licenses, 8 guests. (A max of 4 per license per server. Not 12 on one host at a certain time. For that sort of thing you need Data Center edition)

How many servers you got in total?
anthony Posted - 11/14/2011 : 3:19:49 PM
OK. Licensing question...

Let's say I buy 3 copies of Enterprise R2 to do a 3-Node cluster. I can run 4 copies of Windows on top of that as part of the SKU. If those servers are in a cluster does that mean those copies are aggregated across the 3 node cluster? What if for load balancing reasons I need to run only two copies on one? Can I run 6 on another?
wobble_wobble Posted - 11/14/2011 : 1:31:01 PM

This is fairly recent on the Exchange setup.

Says you got to shut down servers.


Exchange server virtual machines (including Exchange Mailbox virtual machines that are part of a DAG), may be combined with host-based failover clustering and migration technology, as long as the virtual machines are configured such that they will not save and restore state on disk when moved, or taken offline. All failover activity must result in a cold boot when the virtual machine is activated on the target node. All planned migration must either result in shutdown and cold boot, or an online migration that makes use of a technology like Hyper-V Live Migration. Hypervisor migration of virtual machines is supported by the hypervisor vendor; therefore, you must ensure that your hypervisor vendor has tested and supports migration of Exchange virtual machines. Microsoft supports Hyper-V Live Migration of these virtual machines.

joe_elway Posted - 11/14/2011 : 11:20:49 AM
Originally posted by NMDANGE

I think the issue is you can't do a live migration of a Microsoft Cluster member.

Yes you can. See previous post.
NMDANGE Posted - 11/13/2011 : 7:26:03 PM
I think the issue is you can't do a live migration of a Microsoft Cluster member.
joe_elway Posted - 11/13/2011 : 2:47:34 PM
Exchange 2010 with SP1 supports both layers of clustering, host and in-VM.
wobble_wobble Posted - 11/11/2011 : 7:39:26 PM
Gonna throw a big wobbly in here for information.

Exchange 2010 will support 1 (one) HA technology.So pick Clustered Exchange as a solution or Exchange as a guest on a HyperV cluster as a solution.
You cannot cluster and HA Exchange.

This may have changed recently, but I don't think so.

I cannot speak for SQL, but I think it supports Clustered Servers as one option OR HA as an option. You pick one or the other, not both.
anthony Posted - 11/11/2011 : 4:56:45 PM
Thanks Aiden.

Currently reading "Building a Hyper-V Cluster using the Microsoft iSCSI Software Target" and that is providing a lot of insight as well. Although I wont be using the MS target, but still helping me get my head around all the different parts.

Thanks for responding, I know a lot of this is already out there in different places... I'm trying to be mindful of that.
joe_elway Posted - 11/11/2011 : 4:23:16 PM

Live Migration and HA do not require System Center. They are functions of Windows Server Ent/DC and Hyper-V Server. Read the guide I linked earlier.

Clustered hosts should be all Intel or all-AMD. Can be different generations of CPU but you have to turn on compatibilty aka disable advanced VM CPU features for that. Strictly speaking, you want to be as identical as possible, esp if using iSCSI or fiber channel SAN, right down to the firmwares throughout the h/w. OS config, patching, and n/w configs should be identical.
anthony Posted - 11/11/2011 : 2:19:19 PM
OK... One LAST question...

How similar do the HYPER-V HOSTS need to be in order to cluster them? Identical? Semi-Identical? Etc.?
wobble_wobble Posted - 11/10/2011 : 11:57:42 AM

For 50 servers or less look at System Center Essentials Plus.

SCVMM, a chopped down, tuned version of SCOM and DPM all in one package.

You need the servers for the applications, but the license pricing is unbeatable.
anthony Posted - 11/10/2011 : 11:36:40 AM
OK, so a few last questions.

We certainly can stand 30+ seconds of downtime during a hardware failure. If that changes then we can go deeper.

So other than the obvious servers running Enterprise with the Hyper-V role, what other components are needed for the live migration?

System Center VMM I'm assuming. Is that in addition to SC Operations Manager? The MS website really makes it difficult to know what I need. They should have a questionaire I could answer that would push me in the right direction at least.

Also, would there need to be a separate AD domain that the HOSTS run on? Or do you just have a physical DC outside the VM infastructure that allows the hosts to authenticate? In the small installations I have done (that are 3 or 4 VMs on one server) we just left the HOST on WORKGROUP. But I imagine with the clustering scenario that won't fly...
joe_elway Posted - 11/09/2011 : 3:20:09 PM
Anthony, clustering the hosts means that VMs are mobile between hosts. That means Live Migration for maintenance, or HA for hardware/host failure give VM fault tolerance. Live migration (where you move stuff) has zero downtime. But if a host fails, the VM fails over and boots up. The service in it is offline for 30+ seconds.

Clustering in the guest means the service that is running in the guest can be fault tolerance. Imagine a SQL VM cluster. I can run 2 virtual machines as nodes. VM1 on host A. VM2 on host B. If host A fails -> VM1 powers off, VM2 notices and SQL service fails over, VM1 fails over & powers up (i Host A and Host B are clustered). Minimal downtime to SQL of a few seconds. Yup, we're talking extreme here.

Most people are just doing host clustering. But for some businesses, 30 seconds of downtime to a service is unacceptable.

The added benefit of in-guest clustering is VM OS maintenance. Say I'm installing W2008 R2 SP1 in VM1. I can failover the SQL service to VM2. Now I can afford to let SP1 install, taking it's 60-90 minutes, without impacting the business for that long.
Playwell Posted - 11/09/2011 : 10:26:22 AM
It's just different levels really

Uptime scenario of none clustered guests would depend on clustering the hosts, so the guests can be automatically started again on another host.

When i read this sentence back i realised how confusing this can be :)

anthony Posted - 11/09/2011 : 09:54:05 AM
I think I'm understanding. So if an application supports clustering (Exchange, SQL, Etc.) it can be used in this way. But if the application does NOT support clustering (QuickBooks Enterprise, etc.) then it would NOT be able to be clustered so that uptime scenario would not be available.
Playwell Posted - 11/09/2011 : 09:51:18 AM
Uh not really Anthony,

Clustering the hosts gives you redundancy on OS level.
Clustering the guests gives you redundancy on application level

if it is interrupted or not depend on the guest clustering mechanisms
anthony Posted - 11/09/2011 : 09:45:18 AM
Thanks. Aiden I promise I DID read your book! But it's a lot of info to absorb. So, if I'm getting the jist of that. Clustering the HOSTS gives you a certain level of HA (kinda entry level HA) - but clustering the Guests is the HA solution that offers true uninterrupted service in the event of a hardware failure.

Thanks for the response. This will help me sell this to management.
Playwell Posted - 11/09/2011 : 05:43:37 AM
I'm with Aiden
joe_elway Posted - 11/09/2011 : 04:51:17 AM
Anthony, in a perfect world both options would give the best solution.

Most people fear hardware failure. That's why we cluster hosts. A VM then has two things it can take advantage of in W2008 R2 Hyper-V. It can Live Migrate between hosts with zero downtime. This is a proactive administrative decision to relocate the execution of a VM from host A to host B, done either by an admin or some automated admin system (e.g. VMM 2012, or VMM/OpsMgr PRO). This is a proactice response (host maintenance, load balancing, etc). The second feature is HA. This means that if host A does die, then the VM fails over to another host and starts back up again, automatically. This is a reactive response to a host failure (BSOD, power loss, hardware failure).

The mistake most admins/consultants and many s/w vendors are making is they are forgettin the most important thing: the service that is being provided to the business is not servers, switches, networks. It's the apps that sit on that stuff. Imagine a SharePoint farm: an IIS VM and a SQL VM. Put them on a VMware/Xen/Hyper-V cluster and they have HA and Live Migration. They are independent of the host h/w able to move about freely to get the best performance and highest availabilioty possible. What happens if a VM OS has a BSOD? What happens when you need to do maintenance on the VM? What happens if you want to patch or service pack a VM?

That's why we can -also- do clsutering at the VM level. W2008 R2 guests have an iSCSI initiator built in, and that can be used to create a cluster at the VM level, using physical storage in an iSCSI SAN/storage server as the shared storage. I guess vSphere might have a virtual HBA adapter, I know that's coming in Windows 8 Hyper-V, meaning we can use fibre channel SANs in this type of solution too. With a virtualised cluster we can give HA to the service. So, in that SharePoint farm, a cluster of SQL VMs could be created. You could NLB (hardware preferred) the frontend IIS VMs. Now you can take down 50% of the SharePoint service and still leave it running.

Best of both worlds:
1) Host clustering (and usually the 1st thing done) abstracts the VMs from the host h/w. Good for performance and HA at a h/w level.
2) Guest clustering abstracts the service from the VMs. Good for higher levels of service delivery, no matter what virtualisation you use.

Step-by-step Hyper-V clustering guide for iSCSI here:
wobble_wobble Posted - 11/08/2011 : 6:07:19 PM

You cluster the hosts to offer HA to the guests.
This way if a host has to be rebooted/ patched or has a hardware failure, the guest has another host to go to.

Mark Minasi's Reader Forum © 2002-2011 Mark Minasi Go To Top Of Page
This page was generated in 0.06 seconds. Snitz Forums 2000