[Okta Webinar] Learn how to a build a cloud-first strategyRegister Now

x
  • Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 863
  • Last Modified:

Multipath for iSCSI storage using one or two vlans and switches

I'm setting up a HP MSA P2000 SAN that has two controllers in it.  I understand why you'd want each controller to plug into a different physcial switch for redundancy, but do you really need to have two vlans/subnets, one for each controller to connect to?  Or can you just have one and do it all through a single vlan/subnet?
0
jpletcher1
Asked:
jpletcher1
  • 4
  • 3
1 Solution
 
FizicistCommented:
As you wrote, the idea of two controllers is so you can connect each one to a different switch.  I'm not sure exactly by what you mean by "Or can you just have one".  Do you mean one vlan or one switch.  If you only want to have a single switch, that's fine.  If one controller fails, the other will continue.  Each controller is usually in it's own broadcast domain, so yes I'd split the switch into two vlans, one for each controller.  Just keep in mind that if the switch fails, everything will come to a crashing halt.  You can start with this setup and in the future just add the 2nd switch with no need to bring the setup down.  In the setups I manage, I have single SAN's to single switches and multiple SAN's to multiple switches.
0
 
jpletcher1Author Commented:
In my case, I have two physical switches, and each controller connects to a different one for redundancy.  But we have one vlan for all storage traffic.  My question is, is there really a need to have two vlans?  Or is there a reason you wouldn't want both controllers and sans all working off a single vlan network?
0
 
FizicistCommented:
In your case you don't need two vlans.  Treat each controller as it's own network.  On the ESX hosts, you should setup iSCSI for both controllers.  Just put the controllers on separate subnets from each other.  After your done, your iSCSI setup should have both targets.
Check out...
http://www.vmware.com/pdf/vsphere4/r40/vsp_40_iscsi_san_cfg.pdf
and see if it helps.
0
VIDEO: THE CONCERTO CLOUD FOR HEALTHCARE

Modern healthcare requires a modern cloud. View this brief video to understand how the Concerto Cloud for Healthcare can help your organization.

 
giltjrCommented:
-->In your case you don't need two vlans.
-->Just put the controllers on separate subnets from each other

From a networking standpoint typically each IP subnet is in its own VLAN.
0
 
jpletcher1Author Commented:
Yes, I was just about to ask what giltjr mentioned.  You say I can have one vlan, but then to put my controllers on different IP subnets.  Typically a vlan is correlated to a subnet.

What is the issue with having both controllers on the same subnet and same vlan as long as they are plugged into different physical switches?
0
 
jpletcher1Author Commented:
I'm just going to assume there are no issues with having both controllers use the same vlan and same IP subnet.  The info above is kind of contradicting in saying I don't need two vlans but I need to have separate subnets.  

Can anyone clear that up before I close this?
0
 
FizicistCommented:
You really don't need separate subnets.  At one customer I have an Equallogic SAN which has 2 controllers in a Active/Passive layout.  So I only have to configure the IP's for one controller and they get mirrored to the 2nd when it goes active (This setup, two controllers, one subnet).  I also have one customer with an EMC san, which has 2 controllers in an Active/Active layout.  ESX multipathing keeps track of the different paths for the systems.

Look at
http://bizsupport2.austin.hp.com/bc/docs/support/SupportManual/c02523110/c02523110.pdf
Page 4, two switches two servers as an example.  Not all possible setups are outlined in the PDF, but it should give you an idea.

Separate subnets may help in diagnosing a failure.  I don't have an HP San to review to check if they run on an Active/Passive setup.
0
 
jpletcher1Author Commented:
I guess in my mind it wouldn't matter which vendor you have or if it is active/active or active/passive would it?  It seems like it would be a general principal applied to storage networking.  

I guess having separate subnets might be good in the case of if you have two subnets that each run on completely different hardware.  

In any event, thanks for the info.  I'll go ahead and close this out.
0

Featured Post

Independent Software Vendors: We Want Your Opinion

We value your feedback.

Take our survey and automatically be enter to win anyone of the following:
Yeti Cooler, Amazon eGift Card, and Movie eGift Card!

  • 4
  • 3
Tackle projects and never again get stuck behind a technical roadblock.
Join Now