2

I have two trunked L3 switches forming a network backbone. One is an old HP2848 and the other is a new Juniper EX3300. Currently only the EX3300 connects to our WAN Router and is trunked via 4 LACP links to the HP2848. Multiple access switches carrying 6 VLANs connect to both switches (roughly equal traffic loads).

Currently the EX3300 handles all interVLAN routing for all VLANs. My question is:

What is the best way to handle InterVLAN routing? Should I enable SVIs on the HP2848 but keep the physical configuration as is? Should I setup routing on the HP2848 for 1 or 2 of the VLANs to account for its smaller CPU and use one of the trunk link cables to connect the HP2848 directly to the router? This would remove the single point of failure of the EX3300-Router link and I could maybe configure VRRP at this point?

There is, for most of the time, very little local traffic with the majority of all communication being towards the WAN router and on to the internet. The EX3300 and the Router are in one cupboard, the HP2848 is in another. I am not in a position to run extra cables. Traffic loads are not at a level where any CPUs are overloaded currently but traffic will increase for a period of time in the future.

Your advice is appreciated

  • 1
    In terms of the redundancy of the two "core" switches, how is the network set up to handle the failure of one "core" switch? Is it STP that's preventing loops and STP convergence that would restore connectivity in the event of a switch failure? – Todd Wilcox Jun 8 '16 at 13:08
  • Currently there is NO redundancy outside of the 4 bundled links between the two core switches. STP plays no real role in the network other than protecting against stupidity. If the EX3300 fails, the whole network goes down. If the HP2848 fails, half the network goes down. The only real physical change I can see making is taking a link from the bundled trunk and configuring it as a link between HP2848 and the router. – eto.ttnflt Jun 8 '16 at 18:41
3

Actually if your access switches connected to the so called "core switches" have just one uplink, either to the HP or to the Juniper, full redundancy is not an option given your constraints of no additional cabling.

If also an additional cabling from the WAN router to the HP switch is not an option, even resiliency to the Juniper switch failure is not achievable as far as I can see.

Back to interVLAN routing, any FHRP protocol (VRRP etc.) can effectively work only assuming that a client is able to reach the backup gateway if the active one fails, but in your case if the Juniper switch goes down there is no way for the clients (connected via the switches with a unique uplink to this Juniper) to reach the HP switch. Furthermore, even in the case you have VRRP in the example just above, the clients connected via the HP would only be able to establish interVLAN communications only with clients connected via the HP (no WAN or anything else).

Hope it helps.

  • To be honest, I need to replace the router so I can link it to the HP2848 taking on of the currently aggregated links as a backup. Believe me I would love to add some redundancy between the access and core switches but I'm working for a group that informed me that adding a new switch and 10 VLANs to a network with no documentation and loops throughout would be a one day job... – eto.ttnflt Jun 9 '16 at 9:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.