Hi,
The message you refer to was mine. Turned out there was a bug in the
handeling of icmp (port unreachable) packets from clients. It was fixed in
2.6.12-rc5 (iirc). I don't know if the problem and fix applies to 2.4.
Did the memory usage grow slowly, or explode overnight? (Until you run out
of memory you prolly wouldn't notice unless you monitor memory usage.)
Rgds,
Mark.
On Mon, 6 Mar 2006, Alex Joseph wrote:
> HI,all:
I'm running a LVS over 5 caches, the director is redhat-9 with ipvsadm
v1.21 2004/02/23 (compiled with popt and IPVS v1.0.12) and the self-compiled
kernel is 2.4.30. in the passed 200+ days, it works fine, but this morning
the service has low response rate,
the mem usage are listed below,
/proc/slabinfo:
ip_vs_conn 198 3942360 128 196 131412 1 : 252 126
here is a similar topic:
http://archive.linuxvirtualserver.org/html/lvs-users/2005-05/msg00009.html,
but I havent got a clear answer, Is there anybody help me? thanks for any
relply!
Regards,
Mark
|