Help understanding VRRP on Procurve Switching

[shadow175]

Hi All.

I will explain the equipment we have to hand and the desired failover we are looking for. What I want to do is work with you through setting this up from scratch. I must note that I attempted this myself and have obviously fallen fowl to some mis configuration but couldnt figure out what.

Here goes.

2x VM Ware Hosts with a SAN Datastore. Connected over copper Gb and using iSCSI for data. Each box has 6 NIC/Ports

This is incidental but help clarify the Servers.

The PCs are connected to Procurve 1800's, of which we have three. They are, in-turn connected to a Procurve 5120.

What we want to do is introduce a second 5120 in standby or passive to "Step In" should something happen to the primary/live 5120.

Hopefully this is enough detail  but here is a lovelly picture too. (attached)

If each of the switches is configured  the same with port groups and VLANs I have created a lovely manual failover that relies on only one switch being powered on at a time. If both switches are live, I would surely get Network loop-o-rama.

If VRRP works as I beleive buy electing active and passive links across the two switches, then how would we go about setting this up?

What am I telling the switches to do?

Does network traffic flows between the 5120's and therfore dictate how many ports I allocate to VRRP ?

or am I missing the point completely.

[paulgear]

Hi shadow175,

VRRP is a first hop redundancy protocol. This means that it is designed to provide redundancy for the default gateway of your devices. If you aren't presently using the 5120 as your default gateway, VRRP on it won't achieve anything.

If your default gateway is elsewhere and you merely want active/passive failover of individual switch-to-switch links, then you should look at Spanning Tree (STP/RSTP/MSTP) for Layer 2 redundancy.

[Richard Brodie_1]

What model of 5120 have you got? Setting them up as an IRF stack might be another option.

[shadow175]

The exact model is HP 5120-24-SI (JE074A)

I will also go read up on IRF Stack.

All input is appreciated

[shadow175]

Thanks Paul.

If we can achieve what we require using L2, have I over engineered the solution by purchasing L3 switching. Or is it possible to put these switches to better use in the given environment?

All input is appreciated

[Richard Brodie_1]

OK, these are Comware switches (from the 3com acquisition), and they support IRF.  When you have that set up (which is a little fiddly) the switches work like a cluster, and it acts like one big switch.

Having done that, you can then use link aggregation or whatever on your uplinks to it. The downside is that you need to maintain the stack as a unit, and you get some new common failure modes, so it's worth having a DR plan that includes having the whole stack fail at once.

[paulgear]

IRF is definitely the way i would do it. Then you could set up LACP between the edge switches and the 5120 stack, which means that your links will be utilised more efficiently.

I certainly don't think you've overengineered by buying 5120s. If anything, you've underengineered by buying 1800s. :-)

[shadow175]

Thanks to you both for the advice. Also thanks to the community as a whole.

I will concentrate on creating the stack as discussed.