New P4300 G2 - Multiple Switches for failover

[Brad Endsley_1]

I have a new P4300 G2 SAN and want to connect both pieces of it to two different switches. There are two pieces that come with the total SAN and each piece has two NICs in them.



If I wanted to protect against failover and maximize throughput...what would be the optimal way to set these up? I've tried looking for a manual or something on the HP site that would explain the better way to do this...but have been unable to find something.



I have two HP 48 port gigabit switches that this SAN will connect to.



Option 1) One the first piece of the SAN, both NICs are cabled to the first switch and bonded on the switch. On the second piece of the SAN, both NICs are cabled to the second switch and bonded on the switch.



Option 2) On the first piece of the SAN, NIC1 is cabled to the first switch and NIC2 is cabled to the second switch. On the second piece of the SAN, NIC1 is cabled to the first switch and NIC2 is cabled to the second switch.



Any ideas or suggestions or something I can read about this?

[Fred Blum]

Hi Brad,

A good way to start is to browse through this forum.

You will find a lot of information and experience from others. The manual you are looking for: Networking Best Practises Guide:
http://h10032.www1.hp.com/ctg/Manual/c01750150.pdf

SAN: on both nodes create ALB NIC team. Connect Nic 1 to Switch 1, Nic 2 to switch 2.

This will require setting up a static LACP trunk (2 to 4 port connection) between the switches as you loose connectivity.
Intermediate step to test would be to connect everything on one switch. Read the advanced networking administration manual of your switch for setting up trunking and enabling RSTP (Rapid Spanning Tree).

Optional would be to use Jumbo frames and flow control on both server/SAN nics and enabled on switch ports.

[Evert Goor]

Depending on the model of the switch you can use distributed trunking feature from HP.

On the 3500 and higher model this option is availble. Place the ports of the p4300 in lacp mode and all is ready to go.

Flowcontrol is a Must allways activate it. Jumbo frame is an option but not so much profit from this.

[Brad Endsley_1]

The two switches I'm using are 2910al 48 port gigabit switches.

[Evert Goor]

Those 2 switches only have the standard options.
The best is then for performance and as the p4000 is already redundant to place cable from one unit to one switch. So you have in case of disturbance allways the full performance from on unit.

Do not forget to place all network ports on the switch in flowcontrol. And aftter that you activate it on the network cards of the p4000.

Jumbo frame is on the p4000 more work and disturbance then a profit.

Good luck with the implementation.

[Brad Endsley_1]

So with those two switches, you're saying to cable NIC1 and NIC2 to the first switch and same for the NIC1 and NIC2 to the second switch instead of crossing them?

Flow control needs to be turn on for every switch port or just the switchports for the SAN and the VM servers?

[Steve Bishop]

Brad, i'm in the same boat as you. Recently purchased 2 P4300 and 2 procurve 2910. i'm still trying to figure out if i have everything setup correctly. this is what i've done so far.

Create a bond on each P4300 with both nics (adaptive load balancing)

Nic1 from each p4300 is connected to Switch1
Nic2 from each p4300 is connected to Switch2

I've create a 4 port trunk between the 2 switches (Trk1 - LACP).

Can someone jump in tell me if i'm missing anything? Should flow control be enable?

Brad, i hope this will help you out!

Thanks in advance for any help
Sbishop

[Fred Blum]

"Recently purchased 2 P4300 and 2 procurve 2910."

Same here.

"Create a bond on each P4300 with both nics (adaptive load balancing)"

Good. No jumbo frames and no flow control? Enable flow control, jumbo frames are optional as per the Network Best Practices guide.

Set flow control on the SW2910al. If you want jumbo frames set those also.

To change the SAN NNCS you have to move one node at a time out of the management group to be able to change the TCP/IP settings of the NICS. Re-enter it and start with the next.

Don't forget your server NICs. Use the HP Network Configuration Utility.

"Nic1 from each p4300 is connected to Switch1
Nic2 from each p4300 is connected to Switch2"

Good cross connected for switch redundancy.

"I've create a 4 port trunk between the 2 switches (Trk1 - LACP)."

Good, enable Rapid Spanning Tree as per the HP Networking Best Practices and a must according to group members here, allthough I have read an MS paper on iSCSI SAN infrastructure and fail-over that did not recommend it as the algorithm takes a little bite of time.

I guess you are connecting two teamed server NIC's with Transmit load balancing (=ALB) cross connected to your switches and to the SAN for higher throughput and fail-over redundancy?

In that case you do not need to install MS MPIO and HP DSM MPIO as it doesnot support teamed NICS.

If everything is right with your network configuration you will be able to ping, connect with the CMC to the SAN and able to add the volumes/targets by configuring iSCSI Initiator in windows.

DSM MPIO 8.5 works with two seperatly configured NICS in iSCSI Iniatiator connections and remembered connections. It gives a Active/Passive fail-over setup in 8.5. I have read that DSM MPIO is planned to have Active/Active Round Robin by default in W2008 R2. When I tried to set it however from vendor specific (Active/Passive) W2008 R2 in devices I ran into the same behavior as in 8.5 that it would not let me.

I also believe by trail-and-error that it doesnot support the HP 382i's enable iSCSI device accelerator in multiple NIC MPIO setup.

On friday I have planned to continue testing with DSM MPIO 9.

[Brad Endsley_1]

I'm going to try and figure out the configuration on these switches. I'm a Cisco networking person, but we switched (no pun intended) to HP procurve to try and save some money on the purchase price and warranty.

Anyway, I'll look into doing what you guys have described.

Thanks for your input, all of you.