Today, one of the server was uncontactable after a network switch flood incident. All other servers are working fine except this one. Log into this server through the console, reveals :
- ping to gateway is fine
- ping to a host outside this network failed
- ssh connection to same segment is fine but not beyond gateway
/var/log/messages has a dozens of dozens of these :
Apr 9 00:30:01 hostname kernel: arpresolve: can't allocate llinfo for 10.0.0.1
Apr 9 00:31:00 hostname kernel: arpresolve: can't allocate llinfo for 10.0.0.1
Apr 9 00:32:00 hostname last message repeated 15 times
It looks like connection from this server to machines within it's own segment is fine but other then that, all are unavailable. Another thing, machines connected before initiated from this server to it, is fine. Connections from other server within own segment to this server (if NOT initiated by this server before) still failed.
After some googling, it seems like a patch is needed to fix this error :
http://lists.freebsd.org/pipermail/freebsd-net/2011-August/029687.html
Since the arp cache might be leaking and causes the table to be corrupted, a reboot is needed for the network stack to work. After reboot, the usual patching applies.
The moral of the story,
updating FreeBSD is as important upgrading ports. :p
Ciao !!!