[squid-users] 2.5-stable1: mbuf clusters on fBSD

From: Dave Raven <dave@dont-contact.us>
Date: Wed, 26 Mar 2003 18:17:05 +0200

Hi all,
    Having some serious troubles with a clients squid box,
    its running on FreeBSD-4.7-RELEASE-p9; and I've
    just installed the squid through ports, -STABLE1 +
    all the patches in ports Makefile.

26229/26624/133120 mbufs in use (current/peak/max):
        23620 mbufs allocated to data
        2609 mbufs allocated to packet headers
23367/23478/33280 mbuf clusters in use (current/peak/max)
53612 Kbytes allocated to network (11% of mb_map in use)
0 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines

    Thats the output I am currently getting from netstat -mb; as you
    can see the mbuf cluster usage is EXTREMELY high, considering
    this box is on a 256k link with +/- 50 users. Not high load at all.

    When I reboot the box the mbuf clusters value is obviously reset
    to 0; but from there is steadily rises as the cache is used, and
continues
    to rise until reaching 33280 (an enormous value for mbuf clusters).
I've
    pushed this value up throughout the day, and am certain that
something
    is wrong. What could be causing squid to chew through these mbuf
    clusters at a constant rate until its completely used? The value
does drop
    a little now and then, like from 23367 to 23365, then continues to
rise.

    What can I do to find out whats causing the usage? Or how can I halt
it,
    I've checked most of the mailing lists and people are just having
problems
    because they dont have enough mbuf clusters, I have far too many, a
    usual box like this uses less than 5000 in my experience.

    Any help will be much appreciated

Thanks
Dave
Received on Wed Mar 26 2003 - 09:15:29 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:14:21 MST