Core Switch failover

[HMCTech]

We have one core switch in our main server room that we need some sort of redundancy on in case of a hardware failure on the core switch. Never had to implement this before and not very knowledgeable in this area. Does anyone have any ideas on how to have an identical switch to the core running as a backup if the primary should fail? We are a solely HP ProCurve network.

[plexer]

what model is it? a chassis switch? if so redundant psu's which they support and maybe some spare cards on hand rather than a full blown replacement on hand?

 

Ben

[HMCTech]

The switch is a chassis 5412. They already have redundant power supplies, everything is being duplicated in some way over the next twelve months including all fibre running in different areas of the grounds so a break in a cable wont close down a building.

[CyberNerd]

We have 4 core switches (no chassis, but you would use extra modules in a chassis environment).

Each link has at least 2 cores which are aggregated links. These will connect to different switches in the core (or modules in your case) so if a switch fails there is redundancy, or a link fails there is.

Further redundancy is achieved using OSPF between the near edge switches and hte core.

 

 

Edited March 14, 2012 by CyberNerd

[Soulfish]

You'll be wanting to look at VRRP - Virtual Router Redundancy Protocol (Virtual Router Redundancy Protocol - Wikipedia, the free encyclopedia), and for the HP 5400zl series specifically (http://cdn.procurve.com/training/Manuals/3500-5400-6200-6600-8200-MRG-June2009-59923062.pdf). Can't remember if you need to buy extra licenses for the 5400 to use VRRP though.

 

Essentially you connect all edge switch(s) to both cores using LACP trunks, with a LACP trunk between both cores. This thread at the HP forums explains more about someone with a similar setup - HP Communities - STP for LACP Trunks at Core1, Core2 (VRRP enabled)... - Enterprise Business Community

 

Had it working (albeit with 2 x Juniper EX8200 chassis switches) at my last school and it worked really well. Killed one of the core switches and we didn't loose a single ping

 

Edit:

One thing I forgot to add was that if you do it right your able to split all traffic between both cores, and by having all devices connected to both cores you can increase available bandwidth, number of network ports available etc.

Edited March 14, 2012 by Soulfish