IP router redundancy

[Wim Van den Wyngaert]

I have a node with 1 default gateway A.
Now suppose I have a 2nd gateway B going thru the same interfaces.

I have setup the routing table in such a way that B comes for A. B is non-existing.

I do a ping of C, which is behind the routers A (and B but B is not there). Result : device timeout after about 15 seconds.

I redo the ping : working correctly.

I redo the ping : timeout.

So, there is no failover or is my IP bad configured ? Does anyone has pointers to a description of the algorithme ?

Wim

[Antoniov.]

Wim,
may be interesting use traceroute command.

Antonio Vigliotti

[Wim Van den Wyngaert]

Antonio,

Since the node B doesn't exist, it will not respond to ARP and thus not to tracerout.

It concerns TCP 5.3 and I found in the release notes that it is normal behaviour (round robin without retry, thus 1 out of 2 usages fail).

But why ???

[Ian Miller.]

do the routes have the same cost?

[Wim Van den Wyngaert]

Ian,

Cost ? In IP ?
So, yes ...

[Wim Van den Wyngaert]

I refrase.

There is an RFC816 that specifies how to handle failed routes. E.g. by doing retransmit on the next available route.

Why is this not done on VMS ?

[Wim Van den Wyngaert]

I tested it on 4.2 eco 1 too.

There no round robin is done. The first route is taken and that's it.

Wim

[Gib Copeland]

Configure your routers to advertise a default route. Enable RIP on the router and on VMS (gated). Delete your static default gateway route. Gated will choose a default route, and if the router goes away, it will time-out and switch to the other router.