switch switches between SAN and ESX 4 host server

jorlando66 - the configs are the same.

gpiero74 - I have 2 switches on different subnets running from the esx hosts to the SAN. I have 4 NICs for the SAN (2 on the .50 subnet and 2 on .100 subnet) on each host.

As it seems possible to unplug/plug between the switches, my next concern is how that will affect my servers running from the SAN. if I unplug/plug I am concerned about data loss, especially on my exchange server and file server so I guess I may have to come in very early one day to switch the cables.

Thank you both for your input and quick responses.

If you have iSCSI multipathing configured and working you should be able to move one link at a time. If you don't it would be better to find a time when you can take things down to make the change.

Good Luck

bgoering - I am using the MS iSCSI initiator and MPIO as well as powerpath (EMC software). However, my concern is having both switches running and connected while having the same IP address. Should that be a concern? If not, I am considering pulling one and seeing what happens. Thanks

Why do the switches need the same IP address? Isn't it only used for switch management? I wouldn't expect there to be any issue with the management IP address of the switch as your iSCSI traffic doesn't need to talk to that IP address anyway.

If it is of concern couldn't you give the new switch a different IP address while cutting over, then when you retire the original switch you could change the IP address on the new switch back to the original.

bgoering - the SAN requires each of the 4 iSCSI ports to list a default gateway (this should be the IP of the switch between the SAN and the ESX host, according to the manual). So if I use a different IP, I'll have to change that on the SAN. I was looking for a way to do it without reconfiguring anything on the SAN and/or ESX hosts and/or Windows Server VMs and without having to come in early or shut down any servers.

Is your SAN storage on the same IP subnet as the servers? Or is routing and a gateway required to reach the storage. I know that for software iSCSI with vSphere ESX or ESXi the VMkernel port must be in the same subnet as the storage. In other words no routing to storage is allowed. This restriction may not hold for independent iSCSI HBA's installed in the host. I don't believe that restriction holds for the MS iSCSI initiator you mentioned.

In any event, if the SAN is on the same subnet as the Initiator the gateway won't be used, and you should still be OK even if it temporarily has a different IP address.

Are you using the iSCSI SAN to present LUNs used for VMFS Datastores?

Documentation: from iSCSI SAN Configuration Guide for vSphere 4.1:

"CAUTION If the NIC you use with your iSCSI adapter, either software or dependent hardware, is not in
the same subnet as your iSCSI target, your host is not able to establish sessions from this network adapter to the target"

bgoering - the SAN is on the same subnet. VMkernel is on same subnet as well. I am using the iSCSI SAN to present LUNs. I originally gave the unconnected switch the IP of 192.168.100.2 (the connected switch is 192.168.100.1), connected it to an open NIC port on one of the ESX hosts, created a vSwitch, configured an adapter in a windows VM to that subnet, then assigned it to the newly created vSwitch. However, I am unable to ping the 100.2 switch from that windows VM. Can ping 100.1 switch (it has been in production for about 5 months now).

I know you said that you haven't configured any VLANs, but if the switch is VLAN capable (sounds like it probably is because you can give it an IP) it will have a default VLAN. Typically (at least on Cisco equipment) that is VLAN 1. For things to work properly you will need to, at least temporarily, link the two switches together.

What kind of switches are you using?

the connected switch is a Cisco Small Business (remarketed Linksys) SLM 2024. The unconnected switch is an HP 2910al. I don't see why I can't ping the HP though. It has an IP on the same subnet as the adapter in the windows VM (.100) and the physical cable is connected to the host server and the switch. I can ping everything else on the .100 subnet (connected switch, VMkernel port IP, and the SAN ports). vCenter shows that the windows VM is on the newly created switch on the network config page.

I just noticed in vCenter that my newly created vSwitch does not have an IP range assigned as the rest do. The network adapters observed IP range shows none. How do I configure that? I don't remember manually adding that for the other vSwitches and I have 2 for the SAN subnets and 2 for the LAN and they all show the correct observed IP ranges in vCenter.

You have to create a VMkernel port group on the new switch in order to assign an IP address. No observed ranges are showing because there is no traffic on the new switch yet.

I did create a VMkernel and gave it a .100 IP (192.168.100.7). I then added a virtual machine group and added the windows VM I have been testing with. Still can't ping. Observed IP still shows none. Thanks for your help with this.

Did you ever link your switches together?

bgoering - had to leave the office for a meeting. I will do what I can remotely but will definitely link the 2 switches in the morning when i get in. Will that allow me to, eventually, remove the cisco and use only the HP? Thanks again for your help.

OK, I am home for the day also but will look in tomorrow. Linking the switches shouldn't impact your plans. Once everything is moved over just unlink them then change the IP address on the HP to what you want going forward.

bgoering - linking worked. I am able to ping the HP switch. So now what exactly? Thanks again.

bgoering - excellent. Thanks again for your help.

It would likely have worked to just pull the cables and plug them into the new switch but linking the switches ensured I wouldn't lose connectivity as I have multipathing correctly configured. Worked easily one at a time. Watched links go from active to inactive to active in about 10  - 15 seconds.