Setup redundant iSCSI network with 2 network switches

Question

Hey there,

I'm moving forward with a high availability project. I have the following infrastructure:

- 1 Dell PowerVault MD3000i SAN.

- 2 Cisco switches (supported by Dell).

- 2 Dell PowerEdge 710 servers.

The iSCSI ports on the SAN are configured as shown in the following table:

C0 I0 = 10.10.21.1 (connected to switch 1)

C0 I1 = 10.10.22.1 (connected to switch 2)

C1 I0 = 10.10.21.2 (connected to switch 1)

C1 I1 = 10.10.22.2 (connected to switch 2)

Each switch will be configured as a non-stacked switched because I plan to use different IP addresses for each ethernet port. PowerEdge servers has 4 ethernet ports each one, but we only use 2 and 2.

PowerEdge Server 1 - Port 1: 10.10.21.11 (connected to switch 1)

PowerEdge Server 1 - Port 2: 10.10.22.11 (connected to switch 2)

PowerEdge Server 2 - Port 1: 10.10.21.12 (connected to switch 1)

PowerEdge Server 2 - Port2: 10.10.22.12 (connected to switch 2)

I want to provide redundancy at the server port/switch level and RAID controller level. What's the best method to configure Microsoft iSCSI initiatior in order to achieve the high availability if one switch or RAID controller has failed?

Option A: One session with four connections, in a failover or round robin fashion.

Option B: Two sessions, each one with two connections to the same RAID controller through different switch, in a round robin fashion.

Option C: Four sessions, each one with one connection.

I'm not sure what will happenn if a RAID controller of the MD3000i fails. All IOPS will be routing trought the order controller?

Thanks in advance,

Pablo.

Dev Mgr · Answer

Use the MDSM MPIO option. You can either use the MDSU to let it do the iSCSI connections, or do them by hand (you have to do 4 iSCSI connections where each time you specify the source and destination IP to get the 4 possible combinations).

The MD3k doesn't support (Windows) native MPIO. This means that it may or may not work (use at your own risk in other words).

Also; the MD3000i is only validated up to 2008 R2 if memory serves (haven't looked at that support matrix in quite a while).

pabloafain · Answer

Just to make sure I understand... Are you talking about four iSCSI connections or four sessions in one iSCSI connection? Thanks for your advise! Pablo.

Dev Mgr · Answer

You should see 1 target (on the target list), but you have to use the "Connect" button 4 times.

Each time you use a different combination (other than staying with the iSCSI software initiator).

I usually go down the line:

- 10.10.21.11 -> 10.10.21.1

- 10.10.21.11 -> 10.10.21.2

- 10.10.22.11 -> 10.10.22.1

- 10.10.22.11 -> 10.10.22.2

After the first connection it'll say it's connected, but you just click connect again and make the next connection. The target tab will not look any different, but the properties and then on one of the tabs you can see how many connections are there. Depending on the OS (and/or iSCSI initiator) version, it may show you which connection goes from which IP (on the server) to which IP (on the array).

pabloafain · Answer

Ok, thanks for your help. I've made these changes but:

1. When disconnected the cable for the 10.10.21.0 network, all traffic was routed through the 10.10.22.0 network. This is OK.

2. When disconnected the owner RAID controller (I pull out the controller), the sessiones connected to it (10.10.21.3 and 10.10.22.3) change their status to "Reconnecting..." and I lost the volume.

Any idea?

PowerVault

Setup redundant iSCSI network with 2 network switches

Was this post helpful?