Hyper-V and real fail over

Hi,

I’m building High Available solution using Hyper-V and Microsoft Fail over cluster. My target is to double all hardware to eliminate single point of failure. I know how to provide iSCSI network redundancy, heart beat network redundancy. I also know how to set up firewall redundancy.

Unfortunately weak point in my plan is public network redundancy. I tried to find any solution for eliminating single switch and NIC point of failure (as far as I know I can use teaming due not compatibility with Hyper-V issue). In my opinion it is quite probable that one NIC or switch will be down (especially that I met Cisco switch completely stopped to work). Only advice I’ve got is to provide second, stand bay network and when I will find NIC or switch be down I can manually change virtual switch to another adapter (it is not solution I’m looking for. It has to be automatic process).

However I have still hope someone can give me answer for this question.

Best Regards

Find more posts tagged with

Comments

dtlokee

Not sure if this will do exactly what you want but it may be a start

http://www.cisco.com/en/US/docs/switches/lan/catalyst3560/software/release/12.2_35_se/configuration/guide/swethchl.html#wp1376466

You can use a 3750 stack to eliminate the single switch issue.

astorrs

NIC teaming isn't supported by Microsoft but it works for most implementations. There is a bug with the HP teaming utility when used under Hyper-V, but there is a workaround available, I will try to scrounge up a URL for you if that's what you were planning on using. What servers were you using for the Hyper-V hosts?

PiotrIr

Astorrs, many thanks for this reply. It is giving me a hope.
I’m going to use DELL servers, (probably PowerEdge 2950 III). If you could give me some details I would rely appreciate.

Dtlokee, this post will help me to fix next problem and also many thanks for this.

astorrs

PiotrIr wrote:

Astorrs, many thanks for this reply. It is giving me a hope.
I’m going to use DELL servers, (probably PowerEdge 2950 III). If you could give me some details I would rely appreciate.

If you're using Dell 9th gen servers then the Broadcom teaming utility will do it (Broadcom Advanced Control Suite or whatever they call it now).

Microsoft's Robb Mapp wrote:

NIC Teaming is a capability provided by our hardware partners such Intel and Broadcom. Microsoft supports our partners who provide this capability. This is true whether the customer is running Windows, Exchange, SQL Hyper-V, etc. We'll have a detailed KB article about this coming out soon

http://virtualizationreview.com/blogs/weblog.aspx?blog=2296.

The key thing here is "support", just like with all Microsoft Clustering, support is provided by the 3rd party hardware vendor (whoever made the driver): http://support.microsoft.com/kb/254101/

Also have a look through HeroPsycho and I's back and forth over here: http://www.techexams.net/forums/viewtopic.php?t=37859

Essentially you'll be fine.

PiotrIr

I found this

Hyper-V Virtual Networking not functional on Broadcom Network Adapters with
BACS-based teaming and VLANs
You can use the Broadcom Advanced Control Suite (BACS) software to team
multiple adapters to offer fault-tolerance and load-balancing capabilities.
Hyper-V virtual networks that are bound to network adapters configured with
BACS-based teaming or VLANs may not be functional and may disrupt
network traffic from virtual machines to the external network.
There is currently no workaround to support Hyper-V virtual networking on
BACS-based teams or VLANs. A resolution may be available in a future
release. Dell currently does not support BACS-based Teams or VLANs on
network adapters bound to Hyper-V virtual networks.

I know you will tell me that it is only support. But if something is supported it will work in 100% if not it may work or not. If it would be test environment it is not a problem but it will be production environment and if I will have problems with network connections they will kill me.

Are you sure it will work? Any configuration examples? Any settings should be set up? Any advices? Did you or someone you know tested it in production environment? What should I do to be sure?

astorrs

I'm not making any guarantees. You would need to heavily test anything you were planing on deploying - supported or not.

This is what I've seen that worked for others:

We had a similar issue in our environment as well. We're running a dell m1000e blade chassis with m600 dual quad cores, 16gb ram, mirrored sas drives, windows 2008 x64 host with multiple windows 2003 32bit guests. The host has broadcom BCM5708S netxtreme II GigE adapters. BACS 3 version 11.0.20.0, NIC driver version 4.1.3.0.

The symptoms in our case were that the guests had no network connectivity off of the host machine. That is to say that the host could ping the guests, the guests could ping the host, but the guests could not ping any other computers on the network and no computers on the network could ping the guests - although other computers could ping the host. If we broke our team and bound the IP to one of the NIC's all would work fine, going back to teaming would break network communications once again.

After much playing we found that the following resolved our issue:

1. Force speed/duplex, disable IPV4 checksum offload and receive side scaling
2. Break the team
3. In hyperv manager remove the virtual network
4. Recreate the team using the "create team" wizard
1. team type is smart load balancing
2. set one adapter to be a secondary adapter
3. enable auto-fallback disable mode
4. no vlans in use
5. In our case the 2 physical Broadcom adapters in Network Connections have only QOS Packet Scheduler and Broadcom Advanced Server Program Driver checked
6. The team in Network Connections has only Broadcom Advanced Server Program Driver and Microsoft Virtual Network Switch Protocol checked
7. Go back into hyper-v manager and create a new external virtual network adapter and bind it to the BASP Virtual adapter.
8. Assign your static ip to the hyper-v external virtual network adapter created above. In properties of the external virtual network adapter everything except Microsoft virtual network switch protocol is checked.

After this was done we found that we now had full network connectivity to the guests and vice versa. Failover also worked beautifully as well in this configuration. Disabling a switch port to the primary nic in the team caused the secondary nic to take over without even dropping a packet. One thing to note - not setting one of the nic's as secondary appears to work fine until you shutdown both switch ports and try to bring them back up - network connectivity is not restored. Configuring the team with one adapter as secondary overcomes this issue.

Hope this helps someone out there...we fooled around with it forever to get it working correctly.

dynamik

PiotrIr wrote:

But if something is supported it will work in 100%...

You're funny!

PiotrIr

Thanks for this. It helped me a lot. I’m starting to prepare project and praying it will work.
Best Regards

PiotrIr

dynamik wrote:

PiotrIr wrote:

But if something is supported it will work in 100%...

You're funny!

Ok, so it will work in 99.999 % but if it is not supported 50%. Any way I have never met something supported for reasonable company like MS, DELL or HP and not working. Any examples?

msright1981

Hi,

My friend work for a bank who run IBM blades HS21 with Broadcom network cards. They are running Hyper-V & network teaming for the past 2 months. It seems to run fine for them, but still not supported by MS. I told him to ask MS to certify the setup, or shift to VMware VI3. I am not sure what he did after that.