5406R ZL2 - do they support IRF/Stacking?

[localhost123]

hi,

do 5406r zl2 support irf or stacking? I have two and i'd like to avoid going down the STP way if possible. In the product selection in the HP website if i filter by IRD/Stacking supported, the 5400 series comes up. But in the documentation i have found i cannot find anything related to that.

If you can also link configuration steps for irf/stacking (if supported).

thanks heaps.

Nick

[Vince-Whirlwind]

No, they don't stack.

I don't usually like non-stacking switches, but these switches are still very good. They are such good value-for-money that you can (almost) forgive the need for wasting interfaces on STP.

Even better, just use these as Access switches, you can have link-aggregated 10Gb uplinks spread across multiple modules and 2 processing modules and you have all the resilience you need for a wiring closet supporting almost 300 users in one chassis.

[localhost123]

thanks.

I managed to speak with someone at HP. 

It is a long story but i need to make them act like they are stacked.

For this reason i am gonna usa VRRP for layer 3 and DT-LACP for layer 2. This should make them work IRF-like.

cheers

[Richard Litchfield]

They do now - if you have the newest 5400R chassis populated with V3 modules only, you can now stack them using a technology called VSF (Virtual Switching Framework). It allows you to use 10Gb or 40Gb ports to stack a pair of 5400R - much the same as the ProCurve 3800/3810 switches, but without the stacking module.

Make sure you get the latest 16.01 code for them.

[JohnT007]

Hi Richard, thank you for the update. Regarding VSF, any idea how many members in a stack? Perhaps, where can we get more information/documents or best practices? cheers

JT

[Richard Litchfield]

• Max two switches, must be the same type (ie 5406+5406 or 5412+5412)
• 10Gb or 40Gb connections used to stack the switches only (not 1Gb)
• V3 modules only

Probably best to contact your local HPE Network Presales person (I know the ones in ANZ), or your local HPE Network partner.

http://h20566.www2.hpe.com/portal/site/hpsc/public/psi/home/?sp4ts.oid=7074783&ac.admitted=1456198480919.125225703.1938120508

From the 16.01 release notes:

VSF HPE Virtual Switching Framework (VSF) technology virtualizes two physical devices in the same layer into one Virtual Fabric that provides high availability and scalability. A Virtual Fabric is therefore two physical devices in the same layer that utilize VSF technology. VSF allows supported switches connected to each other through normal Ethernet connections (copper or fiber) to behave like a single switch. For more information, see the HPE ArubaOS-Switch Management and Configuration Guide for your switch. Platform: 5400R with v3 blades

[jodygreen]

I'm strugging to find any config how to's for this.  I've had it enabled on one of my switches with all the correct modules in. I have two management cards in each swtich and when I enable VSF my second sup card goes off line.

[Richard Litchfield]

Only one management module is supported in each switch - if there is a second, it is disabled (like you have seen).

vsf
enable domain 101
member 1
type "J9850A" mac-address 5cb901-4a0000
priority 200
link 1 1/A21
link 1 name "I-Link1_1"
exit
member 2
type "J9850A" mac-address 5cb901-4a1111
priority 100
link 1 2/B21
link 1 name "I-Link2_1"
exit

[JonathanBourke]

Hi there, 

I just logged a case for this very issue. Good to know that it's expected behaviour, though not documented very well as far as I can tell.

Since I would like to use the extra management module that I paid for, what would you recommend for a datacenter interconnect over 2 x dark fibre other than VSF?

LACP?

Mesh?

(using 10Gig SFP+ ER optics, approx 20km, 1ms latency)

Jonathan

[Richard Litchfield]

Trying to do VSF over 20km with just 2 or 4 cores probably isn't an ideal use case anyway. Reason: to maximise the benefit, you want devices to connect to both switches - not just one or the other. In the event of a failure you would probably end up with split brain (you need MAD over another link to avoid that), or the standby switch would shut down all ports.

If you own the actual dark fibre pair(s), a CWDM solution may be the best way to maximise the bandwidth. Multiple wavelengths can be passively split out to allow 2-64 channels of 1/10Gb ethernet, 2/4/8/16 FC in whatever combination you want.

Best option for connection is likely to be LACP. MESH isn't going to do anything useful for you in that instance.

[JonathanBourke]

Richard, 

thanks for the input; understand your rationale for the vsf/both switches comment.

dont own the fibre, so looks like LACP is the way to go. Goal is spanned vlans to allow remote / IPSec clients to access multiple physical devices on a single subnet accross two locations.

[FabianK]

Hi Nick,

IRF is a comware propriatary solution, so not available on provision switches.

You might want to consider to check the meshing function of these switches.

Meshing is not that common, but a good solution for L2 Redundancy and load balancing.

Greetz Fabian

[Stephen A Swain]

VSF (IRF like) available in version 16 and later for 5400Rzl2. also VXLAN is there too. However, it's still early days, I'd expect there would be some bugs to iron out yet.

I was looking at dt-lacp for a dci. I'd rather not do LFN (long flat networks), but whatever the replacement feature, it does need to be 100% reliable, and dt has had some serious bugs. However, now that VSF and VXLAN are coming, I've kind of given up on dt.

Anyone know of production deployments of either VSF or VXLAN on the code which is so young?

I was looking at v16.01.0004, but that's got some major bugs STP wise.