Network architecture for the M/C service guard

[vaibhav_1]

I am implementing cluster setup for 2 HP L class server . The server has tree lan cards
one as primary , standby and heartbeat . Primary cards for the both the servers is attaced to a switch and standby cards to another switch . The switch has back to back connectivity so they should act as a bridge in case of lancard failure .
This setup seems to be not working .
The linkloop command gives FAILED status when I try to check connectivity from primary to standby network interface . Am I missing somthing ?

[Carsten Krege]

Your setup sounds good, you probably need to check your switch configuration. linkloop between primary and standby lans must be possible for local lan failover to function (i.e. primary and standby lan must be on the same bridged network).

You might want to refer to the somewhat outdated paper

http://docs.hp.com/hpux/onlinedocs/ha/wpnetwork.pdf

for reference and you will find your setup on page 5 (figure 6).

Carsten

[melvyn burnard]

I would hazard a guess that the switch-2-switch connection may not be correct.
You need to do as Carsten says, and check all the configutrations on the switches.
It may be a worthwhile test to first check all the lan connections by connecting them all to just one switch to prove at least the cables are ok.

[Mark van Hassel]

When the linkloop command fails it can be two things:

1) physical problem: cable, NIC switch port etc.
2) logical problem: when the primary and standby lans are not in the same Vlan (subnet) on the switch linkloop will fail.