1

We have an HP Layer 3 Switch that is doing IP routing between vlans. The general set up is that the switch has an IP address on each VLAN and IP routing is enabled. On our servers VLAN, we have a firewall that has a connection to the outside world.

To set a IP route on the HP router, we use IOS command

ip route 0.0.0.0 0.0.0.0 192.168.2.1

where 192.168.2.1 is the address of our firewall, and the zeros essentially mean to route all traffic that the switch doesn't know what to do with out the firewall as a gateway.

We're in the middle of an ISP and firewall change. I set up the new firewall and ran the IOS command

ip route 0.0.0.0 0.0.0.0 192.168.2.254

(the address of the new firewall). Things started working nicely. When I reviewed the configuration of the switch though, I noticed that it did not replace the previous ip route command, but just added another route.

Now, I know how to remove the old firewall route (no ip route 0.0.0.0 0.0.0.0 192.168.2.1), but what is the effect of having these two 0.0.0.0 routes? Is it switch implosion? Will a server just respond back over the route it receives the request from?

I've read elsewhere that having two default gateways is an impossibility by definition, but I'm curious about this situation that our switch allowed.

Thanks!

1
  • And of course, I'd be happy to provide additional information as needed!
    – SteadH
    Commented Jun 26, 2013 at 8:32

2 Answers 2

0

in the case you have 2 default routes the switch will send one packet to the first gateway the second to the second gateway the next to the first and so on. This can be a problem if a host is just reachable via one gateway or all hosts are reachable via both gateways but one gateway fails. in both cases you will have a performance impact because you will lost every second packet.

it is possible to configure more default routes but it's not a good solution without configuring IPSLAs: https://supportforums.cisco.com/docs/DOC-6078. I'm not sure if a HP Switch supports IPSLAs. Maybe it's related to cisco.

The best solution for redundancy would be using Routing Protocols.

Will a server just respond back over the route it receives the request from?

No. the packest will go the way they are routed back.

0
0

There are two basic kinds of traffic routing:

  1. ECMP - traffic is routed to all equal cost paths simultaneously
  2. redundant - route(s) with the lowest distance/metric are used.

If you want to multipath, and, assuming your platform supports it, adding two default routes without specifying a metric value should do it - it may route traffic on a per-packet basis, but this is not common and it is typically persistent either by just destination, or optimally, both source and destination. You should be able to verify this by having the router show you its routing table - if you see two nexthops for default, it's multipathing.

if on the other hand you want a fallback route, add the additional route with a higher metric and, providing your device has built in unreachability detection to determine that a nexthop is unreachable, it will provide redundancy.

I would, however, recommend using a proper routing protocol instead, OSPF would be a good choice, RIPv2 would also suffice.

1
  • Thanks Olipro! I've removed the second route now, as that ISP connection and firewall will be going away. Perhaps later on we'll have a chance to experiment with multipath.
    – SteadH
    Commented Jun 26, 2013 at 16:52

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .