4

I have a fully redundant HP 8/20q FC SAN that has been working well, but as I dig deeper into things looking for more performance, I'm quickly at the end of my knowledge when it comes to FC topologies.

The SAN is comprised of three dual-port hosts, two switches, and two dual-controller+port arrays (each array is a HP P2000 G3 MSA, has two controllers, and each controller has two FC ports). Each host is connected to both switches, and one port on each array controller is in each switch. The round-robin access strategy is active/active ULP.

As I understand it, the hosts should connect to the switches in F_PORT (p2p/fabric) mode, and not FL_PORT (loop) mode, but what about the arrays? What are the practical differences between the two in terms of redundancy and performance? Each switch has a single port from the MSA logged in as a loop device.

Edit:

I can change the modes easy enough, without replugging anything. What I'm looking for is a good reason to choose one over the other. I have been operating under the assumption that there really is no point to ever using loop mode, as that's what "everyone says", but I'm don't understand why -- are there no good reasons to use loop in situations that don't require it, period?

alzee
  • 427
  • 3
  • 14
  • 5
    Have a good read here: http://en.wikipedia.org/wiki/Fibre_Channel – MichelZ Apr 13 '14 at 19:23
  • Are all your devices connected in as F ports? Do you know how to check? If you give me some more detail about what you're really trying to get at here, I might be able to help more. – Basil Apr 13 '14 at 19:55
  • Have you had a look at the wiki MichelZ mentioned in the comments? I think the "Fibre Channel topologies" section answers your questions pretty thoroughly. Including the limitations and advantages of each. – hookenz Apr 13 '14 at 22:51

1 Answers1

6

If you have a single port on each switch that's logging in as a loop port, you need to address this. The first thing I'd do is try unplugging it and plugging it back in. If it still shows up as a loop port when it relogs back into the fabric, take a look at hard-coding the port to be an F port. I found instructions here.

Be prepared that if there's some underlying problem that is causing it to default to loop ports, this might prevent the port from coming online. If that happens, you'll need to open a support call and possibly replace hardware.

The reason you need to address it is mostly because it's a sign of something else possibly wrong. The part of the FC address space they use is now used by NPIV, however that's not something typically needed by an HP MSA storage device. If it's working well enough that you don't want to take risks, you can leave it. That said, it is best practice to not use FL ports unless you have no other choice.

Basil
  • 8,811
  • 3
  • 37
  • 73
  • Just to add one additional use case for loops, I've used them with modern equipment (Nexsan) for attaching a bunch of disk arrays to a backup server for disk staging pools. In this use, performance is not much of a concern, so I used a loop directly connected to the HBA instead of needing to get an FC switch or multiple HBAs involved. – EEAA Apr 13 '14 at 20:02
  • That doesn't involve connecting them as fabric loop ports, which is why I left it out. I didn't want to muddy the waters :) Although I guess that's kind of similar to a storage array using loops to connect to their backend disks, which I did mention. – Basil Apr 13 '14 at 20:03
  • Yes, that's true. – EEAA Apr 13 '14 at 20:04
  • The guides aren't aging well, but the option is definitely there in both of the arrays (HP P2000 G3 MSA) and both of the switches (HP 8/20q). All devices are connected as F_PORT right now except one port on each switch; for some reason, one of the controllers on one of the arrays selected FL_PORT to each switch. -- L.L, not loggedin somehow it seems. – alzee Apr 13 '14 at 21:55
  • Ah, I see now. I'll update my answer. – Basil Apr 13 '14 at 22:12
  • I've updated the question Basil to clarify -- it's not that I can't change it easy enough, I'm just looking for details on why to choose one over the other. The wiki article is really light on that subject. – alzee Apr 13 '14 at 22:26
  • Yeah, my answer is a little vague, but it boils down to the fact that there's no advantage to loop devices other than that they don't need a switch. In order to support NPIV, they took over the part of the address that used to be used for loop devices. – Basil Apr 13 '14 at 23:35
  • Fine by me, credited the answer. Is that saying or suggesting that NPIV won't work or may have problems coexisting on a fabric with some of the ports in loop mode? – alzee Apr 14 '14 at 02:43
  • I've never tried it to be certain, but obviously NPIV won't on the port negotiating as FL. I suspect it would be able to connect with another NPIV device, but since this is a storage port, you don't technically need to worry. I've seen EMC refuse support on fabrics containing loop ports, but not HP. – Basil Apr 14 '14 at 09:58