Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 5 Apr 2002 11:59:27 -0500 (EST)
From:      Marius <marius@mail.communityconnect.com>
To:        <freebsd-stable@freebsd.org>
Subject:   UDP problems in 4.5-STABLE
Message-ID:  <Pine.BSF.4.33.0204051113450.58305-100000@utterlux.hq.communitconnect.com>

next in thread | raw e-mail | index | archive | help

I recently upgraded a number of my machines to from 4.2-STABLE to
4.5-STABLE.  Now I
think I am having trouble with sending/receiving large amounts of UDP
traffic.  Each system has two interfaces, one for web traffic on the front end,
and one for backend traffic.  The backend traffic is mainly UDP, NFS and a
distributed logging system called spread (http://www.spread.org/).  Spread
is multicasting out our internal access logs via mod_log Spread in apache.
(http://www.lethargy.org/mod_log_spread/index.html)

Both NFS and spread logging are now both crawling along.  The apache error
logs are spewing out spread errors many times a second and my
daily reports are full of 'NFS
server not responding/NFS server alive again messages.'  Even with the
dumb-timer enabled (-d switch to the nfs mount)  I am still getting pages
and pages of NFS errors.

I have searched in for a number of days for a significant change  between
STABLE 4.2 and 4.5, but have been unable to come up with a difference that would
cause this.  NFS is still working, and spread logs are still casting out
but with the vast amounts of errors...I am certainly this is hurting
performance significantly.   I can not imagine it is a card, as all four
machines were effected right after the upgrade.  No packet filters in
place, the backend interfaces are not connected to the outside.  I am
quite running out of ideas.

Any one have a guess about what is going wrong?


So background info:

uname -a
FreeBSD cc50 4.5-STABLE FreeBSD 4.5-STABLE #1: Wed Mar 27 00:03:01 EST
2002     root@cc50:/usr/obj/usr/src/sys/CCI_KERNEL  i386

sysctl -a |grep udp
net.inet.udp.checksum: 1
net.inet.udp.maxdgram: 9216
net.inet.udp.recvspace: 41600
net.inet.udp.log_in_vain: 0
net.inet.udp.blackhole: 0

# netstat -m
1567/11856/65536 mbufs in use (current/peak/max):
        1546 mbufs allocated to data
        21 mbufs allocated to packet headers
1358/11250/16384 mbuf clusters in use (current/peak/max)
25464 Kbytes allocated to network (51% of mb_map in use)
0 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines

swapinfo
Device          1K-blocks     Used    Avail Capacity  Type
/dev/da0s1b       1048448     1040  1047408     0%    Interleaved
/dev/rda1s1b      3140548      852  3139696     0%    Interleaved
Total             4188996     1892  4187104     0%

Interesting kernel parameters (IMHO):
maxusers        0
options         NMBCLUSTERS=16384
options         SMP                     # 2 500Mhz cpus



Hmmm... could maxuser at '0' be biting me?  I will bet I set it to that
after upgrading to make use of the dynamic sizing.   Plenty of swap
though, so I did not think of that earlier.  Not sure how that would
factor in.


-Marius M. Rex







To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.33.0204051113450.58305-100000>