Re: [squid-users] Some errors

From: Hegedus, Ervin <airween@dont-contact.us>
Date: Mon, 9 Sep 2002 13:05:05 +0200

Hello,

first, thank you for quick and fast answer :)

> > - 2002/09/09 12:13:45| fqdncacheParse: No PTR record
> reverse DNS problems?
for which beam? the proxy is in a DMZ. what can't to resolv? the
client or Inet?

> > - 2002/09/09 12:14:16| httpAccept: FD 19: accept failure: (53) Software caused connection abort
> maybe the reason is in "No buffer space available"?
may be - then i hope, i will resolv that... :)

> > - 2002/09/09 12:10:31| comm_open: socket failure: (55) No buffer space available
> Wow!
> I have the same on FreeBSD 4.2,4.4, 4.6,4.6.2
> with all squid 2.4 and 2.5PRE versions! But it is very rare, and I can't find the reason :-((
> Let's help together.
>
> Please check after "No buffer space available" appears:
> 1. used clusters percentage from "netstat -m"

root@proxy:# netstat -m
1261/4064/32768 mbufs in use (current/peak/max):
        1261 mbufs allocated to data
1168/3748/8192 mbuf clusters in use (current/peak/max)
8512 Kbytes allocated to network (34% of mb_map in use)
0 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines

> 2. sysctl -a|grep files
root@proxy:# sysctl -a | grep files
kern.maxfiles: 1064
kern.maxfilesperproc: 957
kern.openfiles: 563
p1003_1b.mapped_files: 0

> to see if kern.openfiles is near maxfiles of maxfilesperproc
> 3. sysctl vm.zone
> line "socket" , maybe the number of sockets is near limit?
root@proxy:# sysctl vm.zone
vm.zone:
ITEM SIZE LIMIT USED FREE REQUESTS
.
.
socket: 192, 1064, 978, 88, 1361898
.
.

> > try to resolv these, i increase these sysctl's:
> > - kern.ipc.nmbclusters: 8192
> > - kern.ipc.nmbufs: 32768
> AFAIK it can be done only in /boot/loader.conf,
> and takes affect only after reboot.
yes, i know it, thanks :)

> I have this settings in my /boot/loader.conf
> kern.ipc.nmbclusters="16384" # Set the number of mbuf clusters
> kern.ipc.nmbufs="65536" # Set the number of mbufs = 4 * nmbclusters
okay, the previous state is also calculated by this way..

> > - net.inet.tcp.sendspace: 65535
> > - net.inet.tcp.recvspace: 65535
> Wel, I work with sendspace= 1024Kbytes, so it doesn't relate to this problem.
hmm, may be i try.. :)

Thank you, i look at your answer for this. :)

A.
Received on Mon Sep 09 2002 - 05:05:07 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:10:09 MST