Failover router ?

Hi all,

This may be OT but I am open to the idea of building a custom linux
router if need be ... We have an application that during 8am-5pm it
cannot fail. We want to use redundancy (not a cluster) should the main
machine fail, and on failure route everything to a backup perhaps on
another subnet. The big question is would we manually have to change
the routing rules? Or can a router detect the failure and re-route
automagically? I'm googling alot for routers and what I see is for
detecting internet failure, not so much just one machine on a subnet
failure.

Any ideas?
Robert

robert wrote:
> Hi all,
>
> This may be OT but I am open to the idea of building a custom linux
> router if need be ... We have an application that during 8am-5pm it
> cannot fail. We want to use redundancy (not a cluster) should the main
> machine fail, and on failure route everything to a backup perhaps on
> another subnet. The big question is would we manually have to change
> the routing rules? Or can a router detect the failure and re-route
> automagically? I'm googling alot for routers and what I see is for
> detecting internet failure, not so much just one machine on a subnet
> failure.

To reroute everything to a different box upon failure would require
that every client monitor the router and know how to change. What you
need to do is have the backup router monitor the main. If the main
stops responding the backup needs to take over the master's IP etc.

Hi,


> We have an application that during 8am-5pm it
> cannot fail. We want to use redundancy (not a cluster) should the main
> machine fail

you are looking for an IP failover solution.

> and on failure route everything to a backup perhaps on
> another subnet. The big question is would we manually have to change
> the routing rules? Or can a router detect the failure and re-route
> automagically?

you don't need changing routes as soon as the (virtual) gw IP will
still be seen up when one of your two routers will die...

Look at VRRP for more info.
carp/ucarp are some free userspace implementations of the latest.

--
Mathieu

On 29 Aug 2006 05:54:37 -0700
"robert" <(E-Mail Removed)> wrote:

> This may be OT but I am open to the idea of building a custom linux
> router if need be ... We have an application that during 8am-5pm it
> cannot fail. We want to use redundancy (not a cluster) should the main
> machine fail, and on failure route everything to a backup perhaps on
> another subnet. The big question is would we manually have to change
> the routing rules? Or can a router detect the failure and re-route
> automagically? I'm googling alot for routers and what I see is for
> detecting internet failure, not so much just one machine on a subnet
> failure.

1. make two routers that are identical routing/firewall tables
2. set the internal/external ips are IPVS (IP virtual server)

If I remember the IPVS (kernel module) should be a balancing protocol
that handles failover. If not, use Linux-HA to detect the failure and
shoot the other node in the head (STONITH).

Personally I use OpenBSD/carp to do the above, it just works easier for
me, but there's no reason why one cannot do the above with the above
methods, or similar.

--
Regards, Ed :: http://www.linuxwarez.co.uk
proud unix hacker
Upon Realizing how many lives it would take to defeat the Japanese in
WWII, Gerneral McArthur sent Mr.T and Chuck Noris plane tickets to a
Anime convention in Hiroshima. I think we all know what happened
next.