Unsolved
This post is more than 5 years old
5 Posts
0
15757
April 12th, 2009 02:00
Best practice ip-addresses Hyper-V + failover cluster
Hi,
I'm looking for a best practice in assigning ip-addresses in a Hyper-V environment in a fail-over cluster.
3x PowerEdge 1950 with 3 dual-port networkcards (onboard broadcom + 2x intel pro)
Windows Server 2008 Enterprise x64
PowerVault MD3000i (dual controller)
2x PowerConnect 6248
1x PowerEdge 2950 -> domain controller + backup server
I try to follow the Dell solution guides for Hyper-V and fail-over clustering, but there are no recommendations regarding ip-addresses in these guides.
The networkcards are connected like this:
port 1: switch 1 -> private cluster network
port 2: switch 2 -> public cluster network, parent partition
port 3: switch 1 -> iSCSI 1
port 4: switch 1 -> Virtual Machine network 1
port 5: switch 2 -> iSCSI 2
port 6: switch 2 -> Virtual Machine network 2
I need LAN & WAN ip-adresses for both Virtual Machine networks and the parent partitions. I'm not sure if I can assign an ip-address in the same subnet to multiple networks (due to cluster configuration) and if it's possible (and recommended) to assign multiple subnets to a single network to avoid single point of failure at the switches for the virtual machine networks.
Is there a recommended ip-address configuration available?
Dev Mgr
6 Operator
•
9.3K Posts
0
April 13th, 2009 07:00
As each company uses it's own IP address ranges, it's impossible to provide any recommendations that would 'work for everyone'.
However, there are a few requirements in regards to subnets;
You should end up with 4 different subnets:
- your regular network subnet + VM networks
- the private cluster network (if it's just 2 servers a patch/crossover cable is a good choice, but if you're going through a switch you need to make sure it's not the same switch as where the physical host is connected (can't afford both links to go down just just 1 switch went down)
- iSCSI 1 subnet (MD3000i uses a default of 192.168.130.x)
- iSCSI 2 subnet (MD3000i uses a default of 192.168.131.x)
The iSCSI subnets should be on separate VLANs and switches (or if you're fully dedicating 2 switches to just the iSCSI traffic (no links to the rest of your network), you could skip on the VLAN part). The reason for this is to be able to withstand a switch failure and still keep your iSCSI SAN reachable by your server(s).
Schepnet
5 Posts
0
April 13th, 2009 13:00
I understand the ISCSI part and my configuration is exactly as you described above (subnets 192.168.130.x and 192.168.131.x on 2 different switches with VLANs).
Since I have a 3 server configuration the private cluster network (subnet 10.1.1.x) also uses one of the switches using port 1 of the on-board network adapter.
The hosts are also connected to the second switch using port 2 of the on-board network adapter. I'm not sure what subnet to use for these connections. The hosts need to be connected to the domain controller, so this should be the same subnet as the domain controllers network, isn't it? Let's say 192.168.70.x
Is it recommended to use both a LAN subnet (192.168.70.x) and public subnet (213.206.???.x) for the network adapters dedicated to VM network traffic? I need the VM's to communicate with each other (i.e. WEB with SQL), to connect to the domain controller and some to be reachable by the world (WEB & MAIL)
I would like to assign 2 virtual networks to each VM. One for LAN and one for WAN, but there I'll create a single point of failure, due to the fact every single VM network is connected to 1 switch only.
So, I think I need 5 subnets: iSCSI 1, iSCSI 2, private cluster network, LAN, WAN. Is that possible in a fail-over cluster configuration?
Thanks again!