Question regardin Sans and Hyper V clustering

[Rozzer]

Hello All,

 

I am in the process of doing a hyper V project. I intend to run two high spec servers attached to a san and setup clustering for live migration. I understand how clusters works and hyper v what i do need to get my head around is a possible san configuration.

 

I have been looking around and i have seen two options. Get a san which connects via SAS or get a iSCSI network attached storage. I prefer the look of the SAS option but can i do live migration and perform all i need with the cluster? Or would i be better with iSCSI?

 

Thanks in advance

 

Ross

[steve]

For the cluster you need clustered shared volumes. For this we've used iSCSI.

 

If both servers can access the SAS connected storage at the same time, then it should work

[Richie1972]

If you're looking for a Windows SAN solution there are two options:

1. Windows Storage Server (can only be purchased with certain hardware)

2. Windows Server (2003/2008) with Starwind Server installed (http://www.starwindsoftware.com I think) - which hosts iSCSI targets

 

Option 2 is what I used with our 2008 (SAN)/2008 R2 (Cluster) setup.

You will need to setup CSV (cluster shared volume) to enable live migration (on a quick test a live server dropped to ping responses during migration)

Easy to set up the iSCSI, easy to insall guests and easy to migrate

 

Richie

[m25man]

Any iSCSI storage that supports MPIO and CSV's will work.

You can spend anything between £0 - £50k to achieve the requirements.

We worked with Netgear using the ReadyNas 3200/4200 devices which have proved to be 100% reliable over the 18 months they have been running so far.

[strawberry]

If you're looking for a Windows SAN solution there are two options:

1. Windows Storage Server (can only be purchased with certain hardware)

 

No longer the case thankfully, you can download the Iscsi target software on its own now!. I've not tested it fully but it does have the advantage over starwind in that the free version can only have 2 connections.

[Achandler]

We have looked intot his, and will be buying a SAN in the summer.

 

You can do it all on both, SAS or iSCSI, we will be buying a SAS SAN and be connecting all hosts to the SAS SAN to allow the migration.

[netadmin]

We just bought two new servers and an iSCSI SAN to use for Hyper-V clustering and live migration. We debated about whether to get SAS or iSCSI, but ended up getting the iSCSI SAN because it provided more future expandability. We will be using Cluster Shared Volumes for the live migration/failover.

 

A quick note: the built-in Windows Server backup and many server backup programs will not work with Cluster Shared Volumes. If you go this route, you might want to look into something like Microsoft DPM Server 2010, which does support CSVs.

[Gibson335]

Question for those who have a single SAN solution: what kind of redundancy have you gone for and what kind of costs are we talking about above and beyond the SAN. I'm not talking about the disks - I mean more about the SAN itself failing after a firmware update or something similar.

 

Cheers.

[netadmin]

We will have a single SAN once the hardware arrives. We equipped the SAN with dual controllers which should improve reliability quite a bit as I believe each controller has separate firmware, etc (this second controller added about $3k to the SAN price). Ideally we would have gotten two SANs for redundancy, but SMT did not want to spend the extra funds for this.

[m25man]

The problem with this thread is that it is danger of going the same way as so many others...

 

There are so many ex corporate NM's on this site that think every school should have a data centre the size of Google and we rapidly go from "I intend to run two high spec servers attached to a san and setup clustering for live migration"

to "Ideally we would have gotten two SANs for redundancy"

 

In the UK where I look after several sites with "SANS" and they can barely afford an extra switch so full scale redundnacy is simply not an option.

 

We use 3 nodes, 1 SAN, NAS Backup with a tested DR Plan to restore any failed hardware and VM from backups within 48 Hrs (because this is a realistic timescale to obtain and setup replacement hardware).

 

Yes, most SMT's will sigh at the thought of 48Hrs outage, but we have powercuts lasting all day sometimes, that doesnt mean they will install site wide Genny's to keep things running.

 

You have to be realistic and look at the economies of scale, last year we almost purchased a £30k Equallogic but at the last minute we switched to a £10k solution as it became clear that finances would be strained in 2011

18 Months later, disaster hasn't struck, everything is running fine the Hyper-V 3Node runs perfectly on its 10Gbe Connected iSCSI SAN and the school had £20k left over to equip and entire lab they never thought they would be able to do....

 

Im not saying that everyone should be using budget SAN solutions BUT there are plenty of lower cost options out there that will suit a lot of scenarios without Dual Arrays, Dual Controllers, Dual Switchgear and Site Replication....

[Oops_my_bad]

Can CSV's be used with FC also?

 

Reason I ask is I am looking to do a similar thing as the OP but cannot stand the thought of having one server sitting there doing nothing incase the other node or component fails.

 

I was thinking of setting up a failover cluster, and zoning say 2 LUNS from the SAN to each host. This way we can spread VM's between 2 nodes and if a node fails the other can continue to provide service.

 

Is this possible or should I only do pure failover?

 

Do I need a witness disk if using CSV?

 

Thanks..

[strawberry]

The problem with this thread is that it is danger of going the same way as so many others...

 

There are so many ex corporate NM's on this site that think every school should have a data centre the size of Google and we rapidly go from "I intend to run two high spec servers attached to a san and setup clustering for live migration"

to "Ideally we would have gotten two SANs for redundancy"

 

....

 

Spot on, Our cluster uses £500 servers from John @ ict direct, Our standalone hosts cost around £1000 and our sans are either servers shoved full of disks or an MSA 20 off ebay for £500 all inc. The prospect of spending 10k + on a setup horrifies me. Disaster recovery is about having your eggs in enough baskets and having options to replace when it does all go wonky.