Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I am about to implement an iSCSI solution on our network. The thing that is bothering me though is that we are only using 1 switch from the SAN to node connections. Since each node has 2 NICs, is anyone able to give me some tips on how to set up a second switch and use the second NIC in such a way that if the first switch were to die, then the second switch would take over?

Maybe in any case, even if there was a failover solution, would data loss still occur due to the "switch over" time required?

O/S is Centos 5.5

share|improve this question
    
Question is lacking detail, for one the server operating system will effect answers regarding the 2nd NIC configuration, can you please provide this information?. –  Oneiroi Apr 25 '11 at 20:00

3 Answers 3

up vote 4 down vote accepted

set up a second switch and use the second NIC in such a way that if the first switch were to die, then the second switch would take over

Simply. You do not.

You keep both lines separate. One dies, who ares.

The trick is to use MPIO (Multi Path IO) which is part of the ISCSI implementation / Standard. Your server and the SAN agree to use two paths for every LUN. Then the server can, when one path dies, just not use that one.

Also does load balancing while both paths work without funny playing around with bonding etc.

share|improve this answer
    
Thanks for this. Is there any documentation on how to get this going with a CentOS 5.5 target and initiator? I've tried Google, however I'm lost with all the different iSCSI implementations.. –  jtnire Apr 25 '11 at 20:13
    
No, sorry. I dont use Linux. –  TomTom Apr 25 '11 at 20:14
    
I'm accepting this answer as it gave me some very helpful pointers on what to search for on Google. I found this guide which at least helps me with the initiator end of the puzzle: how2centos.com/… –  jtnire Apr 25 '11 at 20:25
    
Great ;) The really good news is that MPIO makes a LOT of stuff a LOT easier, like "real" redundancy with syncing backend SANS ;) –  TomTom Apr 25 '11 at 20:27
    
I think this is what you are looking for. docs.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/5/html/… –  Aaron Copley Apr 25 '11 at 21:22

Also, it's typically preferable to keep iSCSI traffic separate from the rest of your network traffic; ideally you use separate NICs and switches from your regular network. If you have two single port NICs in the machines, you may want to think about adding another NIC, so that you can have >=2 separate paths for iSCSI, plus a port or two for regular network traffic.

Before you go down the rabbit's hole of redundant iSCSI connections, you should stop to think about the rest of your architecture. Yea it's great to have multi-path iSCSI, but if you have only one power circuit, and it goes down, well that extra iSCSI connection isn't doing much for you. Or if you have only one SAN, same problem.

TomTom's answer is spot on --I'd vote if I could, but I'm not registered.

share|improve this answer

Yes you should have a second network switch, with the switches cross connected to either other. Each NIC on your client machine should have it's own IP address.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.