|
Francois JEANMOUGIN wrote (at Wed, Mar 30, 2005 at 09:51:36AM +0200):
> Casey Zacek :
> > TCP 00:20 NONE CLIENT-IP:0 VIP:0 RIP1:0
> > TCP 00:42 NONE CLIENT-IP:0 VIP:0 RIP2:0
>
> This only should happen if one realserver goes down and up again during the
> persitence interval. Otherwise, there is a bug somewhere. What versions of
> IPVS/Linux kernel are you using so that people who are using the same could
> check?
I checked for realserver failures, and they didn't happen. We have
the healthcheck timeouts for these servers set ridiculously high
specifically to avoid that situation.
# cat /etc/debian_version
3.1
# ipvsadm --version
ipvsadm v1.24 2003/06/07 (compiled with popt and IPVS v1.2.0)
# uname -r
2.6.11-fai-p3
(stock 2.6.11 (compiled w/SMP) on a single P3/750Mhz)
In case it's of any use, the system appears to be pretty low on
LowMemory:
# grep Low /proc/meminfo
LowTotal: 254204 kB
LowFree: 3684 kB
Keepalived (which isn't responsible for connection tracking) is
1.1.11, but patched with the patch I sent to this list a few weeks
ago.
--
Casey Zacek
Senior Engineer
NeoSpire, Inc.
|
