|
Re: 2.6.18-028test007.1 crush [message #9176 is a reply to message #9175] |
Thu, 21 December 2006 01:42 |
dagr
Messages: 83 Registered: February 2006
|
Member |
|
|
BTW - used to got another error messages before.
Server was under high number connections load.
After starting server under openvz kernel we start get messages in /var/log/messages like this
TCP: time wait bucket table overflow
printk: 116 messages suppressed.
.......
............
Looks like some VPS 0 restrictions, although its unlimited
sudo cat /proc/user_beancounters (this is under 30% of average load)
uid resource held maxheld barrier limit failcnt
0: kmemsize 65457707 67576465 2147483647 2147483647 0
lockedpages 1019 1019 2147483647 2147483647 0
privvmpages 393767 409333 2147483647 2147483647 0
shmpages 4615 5271 2147483647 2147483647 0
dummy 0 0 2147483647 2147483647 0
numproc 928 965 2147483647 2147483647 0
physpages 80131 84819 2147483647 2147483647 0
vmguarpages 0 0 2147483647 2147483647 0
oomguarpages 80131 84819 2147483647 2147483647 0
numtcpsock 4225 4282 2147483647 2147483647 0
numflock 3 7 2147483647 2147483647 0
numpty 5 6 2147483647 2147483647 0
numsiginfo 0 4 2147483647 2147483647 0
tcpsndbuf 38173792 38625464 2147483647 2147483647 0
tcprcvbuf 70973068 71957612 2147483647 2147483647 0
othersockbuf 1156384 1237324 2147483647 2147483647 0
dgramrcvbuf 0 9800 2147483647 2147483647 0
numothersock 549 586 2147483647 2147483647 0
dcachesize 0 0 2147483647 2147483647 0
numfile 23462 23895 2147483647 2147483647 0
dummy 0 0 2147483647 2147483647 0
dummy 0 0 2147483647 2147483647 0
dummy 0 0 2147483647 2147483647 0
numiptent 113 113 2147483647 2147483647 0
[Updated on: Thu, 21 December 2006 03:53] Report message to a moderator
|
|
|
Re: 2.6.18-028test007.1 crush [message #9177 is a reply to message #9176] |
Thu, 21 December 2006 02:50 |
dagr
Messages: 83 Registered: February 2006
|
Member |
|
|
restarted with kernel 2.6.9 023stab037.3
got just same error in message log
Checked sysctl
sudo sysctl -a | grep -i tcp_max
net.ipv4.netfilter.ip_conntrack_tcp_max_retrans = 3
net.ipv4.tcp_max_syn_backlog = 1024
net.ipv4.tcp_max_tw_buckets_ve = 2048
net.ipv4.tcp_max_tw_kmem_fraction = 128
net.ipv4.tcp_max_tw_buckets = 180000
net.ipv4.tcp_max_orphans = 131072
set net.ipv4.tcp_max_syn_backlog = 2048
No more erorr messages so far.
First question - why net.ipv4.tcp_max_tw_buckets_ve and net.ipv4.tcp_max_tw_buckets - differ so much ? It looks like this parameter affects HN as well , ie VPS 0 .
How far is it safe to increase this parameter in case i have no running VPSs at all, is it safe to set it 180000 ?
Still wondering if that kernel panic was caused by this thing , it used to work ok for 2 day with that error messages.
|
|
|