MD3000i Hyper-V cluster and other nodes

Question

Hi Everyone

I would like to hear the comments regarding which is the correct method of assigning IP's to my SAN and my hosts.

I have 2 identical Dell servers which us gonna be in a cluster configuration for Hyper V. 2 Dell powerconnect switches for redundancy and 1 dual controller MD3000i

Currently the IP Config are like this

MD3000 -

0,0 - 192.168.130.101

0,1 - 192.168.131.101

1,0 - 192.168.130.102

1,1 - 192.168.131.102

Contr 0 managent 192.168.1.100

Contr 1 managent 192.168.1.200

Cluster Host 1

LAN - 192.168.1.10

iSCSI 1 - 192.168.130.104

iSCSI 2 - 192.168.131.104

Heartbeat - 192.168.100.1

Cluster Host 2

LAN - 192.168.1.20

iSCSI 1 - 192.168.130.103

iSCSI 2 - 192.168.131.103

Heartbeat - 192.168.100.2

Stand Alone Host

LAN - 192.168.1.30

iSCSI 1 - 192.168.130.105

iSCSI 2 - 192.168.131.105

The 2 switches are connected to each other.

What I have done with the cabling is to plug 1 of the iSCSI connections from each machine into the separate switches and the same goes for the standalone host and MD300i.

To make the mental picture easier - the top switch has all the 192.168.130.x connections plug into it and the bottom switch has all the 192.168.131.x connections plugged in. Remember the 2 switches are connected to each other.

Ok my point or question is this, is my setup suitable for HyperV clustering and Cluster Shared Volumes ? If one switch fails then all the traffic of the one subnet is lost.

Must I change it to all be on the same subnet ?

I have read contradicting configurations and I am starting to question my method and if its needs to be on one subnet.

Hope this is enough info. Please do not hesitate to ask if anything is unclear.

Looking forward to all the replies :emotion-2:

Dev Mgr · Answer

First; disconnect the switches from eachother. There is no reason for the 2 to be connected and it only increases your risk (in the rare situation of a network storm or so, having them disconnected prevents it from taking out your whole SAN).

Assuming you're using Windows 2008 R2 (core or full) with the hyper-v role, you want to install at least the host software for the MD3000i so it installs the MPIO driver. iSCSI and LAN are 2 different protocols in regards to how you make them resilient/highly-available; for LAN traffic you team NICs, but in iSCSI the MPIO driver splits the IO packets between the available iSCSI paths, and if 1 or more paths go down (but at least 1 is still working), the IO will be redirected by the MPIO driver. If you team NICs for iSCSI or try to use a single subnet, you're reducing your resilience.

mrokkam · Answer

In general, even if you lose a switch, you still have the 2nd subnet connecting all components (controllers, clustered hosts and infividual host). So, your config appears correct. I don't see any place where it asks you to keep everything on the same subnet... in fact, MD3000i documentation recommends using multiple subnets for maximum redundancy.

I agree with Dev Mgr :) S/he rocks!

-Mohan