Hi all
I'm trying to build redundancy into two 3.8 boxes with trunk and carp.
Both boxes have 2 Nics each (fxp and rl), connected to two unmanaged dumb
switches without .1q tagging or other fancies.
On both boxes I have:
/etc/hostname.rl0
up
/etc/hostname.fxp0
up
and the corresponding versions of
/etc/hostname.trunk0
inet 192.168.1.{x,y} 255.255.255.0 192.168.1.255 trunkport fxp0
trunkport rl0
/etc/hostname.carp0
inet 192.168.1.254 255.255.255.0 192.168.1.255 vhid 1 advskew {1,10}
carpdev trunk0
When booting this gives me a working trunk, but carp0 remains in INIT state.
There is an error in the boot messages like "SIOAIFADDR (Cannot assign
requested address)" (not sure about the SIO..., sorry, I'm not at the
machine at this moment, but the error message is correct).
When running # sh /etc/netstart manually after login, I get an error
"SIO... No buffer space available" (or similar). When running it a second
time right afterwards, no error message appears, and the carp interface
goes up, replies to one ping (of a constantly running remote ping) and
then goes into BACKUP state (both physical interfaces are up and
connected, trunk failover kindof works (problems being related, i suppose,
to the fact that there is no "failover" trunkproto in 3.8)). The same
happens on the second machine, so that, although I have 4 physical
interfaces up, both carp interfaces are in BACKUP and the carp address is
unavailable to the network.
Now my question is: Is carp supposed to work on trunk interfaces? If yes:
what trunkproto would be best applied to such a setup? Should it work in
3.8 already or do I need to grab 3.9 or even -current?
thx /markus
