gateway failover.

Mailman List Mirror (Read Only)
1

i'm looking for pointers to any resources for doing failover on
solaris boxes. i have an environment in which i'd like a pool of
servers to use one of two solaris boxes as their gateway, the other
box being in hot standby until a failure of the primary box occurs,
at which time it would failover.

something like HSRP for unix would work nicely, though i've not found
anything similar, except for IRDP (rfc1285), however it doesnt seem to
be supported by anyone, in so far as i've been able to tell.

i'm aware of commercial solutions, but i'm hoping to find something
in the public domain as most of them are overkill and/or expensive.

i'd like to avoid doing something with gated/ospf (or another suitably
featured routing daemon/protocol), as the environment may include NT,
95, and other boxes that dont easily listen to any sufficiently useful
routing protocol, as far as i know.

i'm open to just about any suggestion, including an L2 solution,
except for something that would depend on clientside software,
because of the OS limitations.

respond privately, and i'll summarize to the list if necessary.

-foo

2

foo@eek.org (foo) writes:

i'm looking for pointers to any resources for doing failover on
solaris boxes. i have an environment in which i'd like a pool of
servers to use one of two solaris boxes as their gateway, the other
box being in hot standby until a failure of the primary box occurs,
at which time it would failover.

f.root-servers.net (192.5.5.241) is a virtual host on two different
physical hosts. these hosts speak gated and inject a /32 route into
ospf, which makes the /32 look dual homed. the exit gateways pick
the first one they saw (or sometimes i adjust the costs manually).
if the host that was picked as the path to the vhost dies, the exit
gateways flip the route over to the surviving host.

alteon, foundry, and radware make "layer 4 switch" boxes that do the
same thing but add local load balancing so that both hosts can be up
and serving at the same time. in this case the virtual host is in the
L4-switch box which does its own remapping to physical hosts.