Native SAN for Hyper-V in 2012 Beta coming soon?

Public beta (bugs, reports, suggestions, features and requests)

Moderators: anton (staff), art (staff), Max (staff), Anatoly (staff)

Post Reply
sammybendover
Posts: 25
Joined: Sun Jul 15, 2012 10:05 pm

Thu Sep 13, 2012 9:23 pm

So now that Server 2012 RTM is out, what is the word for this? I am wondering if I can do this between 2 or 3 2012 rtm hosts, each with the following:

-Intel i5 2500k
-32gb mem each

-6 intel 1gbe ports teamed running trunk links with lacp port channel, and virtual adapters for each vlan as well as QoS for the replication iSCSI traffic having majority of bandwidth. All are connected to Cisco 3750G-24T switch. 6 GbE links per host should provide pretty nice bandwith for whatever. Jumbo Frames are enabled. Check here for more info http://www.aidanfinn.com/?p=12588


about 4 disks each in raid 0 ( i may change that later, but for know that is what i have) I would like to run raid 10, but that gets expensive pretty fast for a home lab! Also, what about storage spaces to run the iscsi disks? good idea, not a good idea?

Also, what is the "offical" work for netsh configuration of nics in hyper-v for 2012 rtm?

Also, in anticipation of SSD cache that you are hopefully working hard on, between each admin, what is the preference for setting up the disks so when that comes out in beta i can test right away?

And what would you like me to test if anything? I am going to rebuild this weekend, so best tips and tricks for this would be awsome!
sammybendover
Posts: 25
Joined: Sun Jul 15, 2012 10:05 pm

Fri Sep 14, 2012 2:30 am

also, if the local disk array fails, with the vm's on that node continue to run vm's via iscsi on another node or not? Reading the documentation i am still trying to wrap my head around how it works
User avatar
Bohdan (staff)
Staff
Posts: 435
Joined: Wed May 23, 2007 12:58 pm

Fri Sep 14, 2012 2:50 pm

Hi sammybendover,

we are actively testing StarWind with Windows Server 2012 and we'll be able to give you exact answers a little later.
You definitely should perform testing with storage spaces.

As for netsh, StarWind shows good results on default settings, but sometimes some tuning may be required (http://www.starwindsoftware.com/forums/ ... t2293.html)

L2 SDD cache support will be implemented in future StarWind Native SAN releases, but not in the upcoming one.

I would ask you to test 3 and 2 node HA configurations.

"if the local disk array fails, with the vm's on that node continue to run vm's via iscsi on another node or not? " could you please clarify what do you mean?
sammybendover
Posts: 25
Joined: Sun Jul 15, 2012 10:05 pm

Sat Sep 15, 2012 12:23 am

What i mean is, let's say i have 3 hyper-v hosts, all with das storage only. Using native san for hyper-v, if one of the disk arrays fail on one of the hosts, will the guest vm's that were running on that node with the failed storage continue to run using storage from another hyper-v host, or will the vm's on the host with the failed array eventually blue screen, and come back on one of the other hosts, but in a crash consistent state by just using failover clustering in windows?

Thanks!
User avatar
anton (staff)
Site Admin
Posts: 4008
Joined: Fri Jun 18, 2004 12:03 am
Location: British Virgin Islands
Contact:

Sat Sep 15, 2012 8:47 am

You're mixing everything into one bowl :) OK, let's take the things apart:

StarWind has nothing to do with hypervisor VM management and handling. So if...

...physical host with running VMs will go down - VMs will go down as well. StarWind is not in use here at all.

...physical host with running VMs will go down but VMs have been configured to be highly available - VMs will reboot on another
physical host with very little downtime presented to user services. StarWind provides shared storage for this scenario.
Without StarWind whole thing has no way to work.

...physical host with running VMs will go down but VMs are configured to have guest VM clustering (failover) or VMs exploit own built-in
clustering (for example log mirroring with SQL Server) - VMs will transparently failover (no user downtime) to another physical host.
StarWind does feed shared storage to the whole cluster and it does provide resilient iSCSI services to guest VM cluster thing (required).

That's the way it works.
sammybendover wrote:What i mean is, let's say i have 3 hyper-v hosts, all with das storage only. Using native san for hyper-v, if one of the disk arrays fail on one of the hosts, will the guest vm's that were running on that node with the failed storage continue to run using storage from another hyper-v host, or will the vm's on the host with the failed array eventually blue screen, and come back on one of the other hosts, but in a crash consistent state by just using failover clustering in windows?

Thanks!
Regards,
Anton Kolomyeytsev

Chief Technology Officer & Chief Architect, StarWind Software

Image
sammybendover
Posts: 25
Joined: Sun Jul 15, 2012 10:05 pm

Sat Sep 15, 2012 4:33 pm

anton (staff) wrote:
StarWind does feed shared storage to the whole cluster and it does provide resilient iSCSI services to guest VM cluster thing (required).
That is what I am wondering about. I am trying to understand what this product does when things go south.

When i have 2 hyper-v hosts, both configured with DAS array only and using this product, when an array fails on one of the hyper-v hosts, not the hyper-v host itself mind you, just the local storage. When a local storage array fails, what happens to the vm's? If they go down in a crash consistent state?

Again, this is not a whole host failure, just storage on one of the hosts.

Sorry for all the stupid questions, I am trying to understand how this works, and I can't seem to wrap my head around what it does besides replicate data from one das store to another. I see alot of scenarios where a host fails, and the data or vm's won't get replicated to the other hyper-v host storage resulting in corruption or loss of data. I want to make sure my data is secure and there are some kind of provisions to prevent this built into this product. If the vm's need to reboot on another host via HA, then so be it, as long as there is no data loss. Thanks!
User avatar
anton (staff)
Site Admin
Posts: 4008
Joined: Fri Jun 18, 2004 12:03 am
Location: British Virgin Islands
Contact:

Sat Sep 15, 2012 6:39 pm

StarWind does storage virtualization. So it takes a bunch of non-HA DAS LUNs and creates HA SAN (and optionally a failover NAS on top of it) from them. Result is represented to Windows / Hyper-V. So if DAS LUN will go AWOL StarWind will hide this fact from Windows / Hyper-V using data from mirror (first or second - depends on did you configure 2x or 3x synchronous replication between HA nodes) and everything will continue as it was before - no interruption of any service. With 3x replication even performance degradation because of decrased number of MPIO paths is going to provide minimal impact on resulting performance. In a nutshell: your VMs should have no problems at all.

Hope this helped :)
Regards,
Anton Kolomyeytsev

Chief Technology Officer & Chief Architect, StarWind Software

Image
sammybendover
Posts: 25
Joined: Sun Jul 15, 2012 10:05 pm

Sat Sep 15, 2012 10:00 pm

Sounds good. I think I will try running a 3 node setup. I may need a license for Native SAN if it is different from the Starwind iscsi 6 server.

For my lab, i am going to try and run 2 disks and raid0 (it's a lab) and I will be backing up the VM's as well.

How well does Intel SRT on Z68 chipset work for caching as well?
User avatar
anton (staff)
Site Admin
Posts: 4008
Joined: Fri Jun 18, 2004 12:03 am
Location: British Virgin Islands
Contact:

Sat Sep 15, 2012 10:05 pm

1) Dedicated license is required for Native SAN.

2) RAID0 with 3-way replication is what we recommend.

3) Work just fine.
sammybendover wrote:Sounds good. I think I will try running a 3 node setup. I may need a license for Native SAN if it is different from the Starwind iscsi 6 server.

For my lab, i am going to try and run 2 disks and raid0 (it's a lab) and I will be backing up the VM's as well.

How well does Intel SRT on Z68 chipset work for caching as well?
Regards,
Anton Kolomyeytsev

Chief Technology Officer & Chief Architect, StarWind Software

Image
sammybendover
Posts: 25
Joined: Sun Jul 15, 2012 10:05 pm

Sun Sep 16, 2012 4:58 pm

I attempted to download the trial version of it, and it appears my account is locked out again. I did ask sales to unlock my account so i can further test products.

I will try a 2 node setup until a new beta is released.
User avatar
anton (staff)
Site Admin
Posts: 4008
Joined: Fri Jun 18, 2004 12:03 am
Location: British Virgin Islands
Contact:

Sun Sep 16, 2012 8:16 pm

It's up to sales what to do here. Maybe they did not like your free e-mail and a fake phone number. If you'd contacted sales@starwindsoftware.com telling about your issue they should respond pretty soon.
sammybendover wrote:I attempted to download the trial version of it, and it appears my account is locked out again. I did ask sales to unlock my account so i can further test products.

I will try a 2 node setup until a new beta is released.
Regards,
Anton Kolomyeytsev

Chief Technology Officer & Chief Architect, StarWind Software

Image
sammybendover
Posts: 25
Joined: Sun Jul 15, 2012 10:05 pm

Mon Sep 17, 2012 7:10 pm

as we discussed last month here:

http://www.starwindsoftware.com/forums/ ... 53-45.html

(midway through the page) I cannot use my work email.

Since we are a large Netapp and EMC partner, I cannot use my work email.

I am an IT professional, and I am interested in using these products to test to possible recommend to some of our customers where it may be adventageous to them.

If you want to talk about my situation, I would be happy to discuss.

Thanks!
User avatar
anton (staff)
Site Admin
Posts: 4008
Joined: Fri Jun 18, 2004 12:03 am
Location: British Virgin Islands
Contact:

Mon Sep 17, 2012 8:33 pm

I know. That's fine. Contact sales and they will unban you. I don't have access to customer portal (neither do anybody from techies here) so cannot do it myself. I've sent them an alert already.
sammybendover wrote:as we discussed last month here:

http://www.starwindsoftware.com/forums/ ... 53-45.html

(midway through the page) I cannot use my work email.

Since we are a large Netapp and EMC partner, I cannot use my work email.

I am an IT professional, and I am interested in using these products to test to possible recommend to some of our customers where it may be adventageous to them.

If you want to talk about my situation, I would be happy to discuss.

Thanks!
Regards,
Anton Kolomyeytsev

Chief Technology Officer & Chief Architect, StarWind Software

Image
Post Reply