|
Hi,
i've a strange problem with my director box which runs now almost 2 years
very well. i've seen today over 67000 ! connection items in the connection
table and almost all of them shows strange timer values like:
IP 5960:54 NONE 80.38.149.nn:0 0.0.0.5:0 10.0.1.nn:0
IP 5948:53 NONE 83.129.30.nn:0 0.0.0.5:0 10.0.1.nn:0
All ip addresses occures only ones (uniqe) within the connection table.
I've a persistent node running configured with 3600 sec. using fwmark,
it's a Debian Woddy, patched 2.4.23 Kernel with ipvs-1.0.9 or ipvs-1.0.10
running. I had to restart the server a couple days ago due to electricity
reasons, since then i notice this problem. ....
I also did a test from home and didn't saw this behaviour from my home
address so i can't reconstruct.
After a restart of the machine it looks better but i still see a couple
entires > 3600 sec.
I grabbed this after the reboot from dmesg:
May 8 18:51:11 lvs kernel: ip_tables: (C) 2000-2002 Netfilter core team
May 8 18:51:11 lvs kernel: ip_conntrack version 2.1 (4094 buckets, 32752
max) - 292 bytes per conntrack
May 8 18:51:11 lvs kernel: IPVS: Connection hash table configured
(size=4096, memory=32Kbytes)
May 8 18:51:11 lvs kernel: IPVS: ipvs loaded.
May 8 18:51:11 lvs kernel: IPVS: [wlc] scheduler registered.
May 8 18:51:12 lvs kernel: IPVS: [wrr] scheduler registered.
lsmod returns:
ip_vs_wrr 1472 8 (autoclean)
ip_vs_wlc 1024 7 (autoclean)
ipt_LOG 3328 3 (autoclean)
ipt_MARK 800 92 (autoclean)
ip_vs 50200 17 (autoclean) [ip_vs_wrr ip_vs_wlc]
iptable_mangle 2176 1 (autoclean)
iptable_nat 14708 0 (autoclean) (unused)
ip_conntrack 17620 1 (autoclean) [iptable_nat]
iptable_filter 1728 1 (autoclean)
ip_tables 11008 7 [ipt_LOG ipt_MARK iptable_mangle
iptable_nat iptable_filter]
ipvsadm output of the persistent node is:
FWM 5 wlc persistent 3600
-> srv1.intern:0 Route 100 35 8
-> srv2.intern:0 Route 200 41 51
and another:
FWM 6 wlc persistent 3600
-> srv1.intern:0 Route 100 68 48
Does anybody have a idea how to solve this ?
Many thx
Matthias
|
