Re: [squid-users] comm_accept

From: Chris Conn <cconn@dont-contact.us>
Date: Wed, 21 Mar 2001 17:05:37 -0500

Henrik Nordstrom wrote:
>
> These might be seen if the client aborts the connection during early TCP
> setup (i.e. before the connection is accepted by Squid). Generally
> harmless, but might be a sign of a run-away client/robot.
>
> --

Hello,

I do agree that it may be harmless, however I didn't get this message at
all with kernel 2.2.16 (if I revert back to 2.2.16 it stops). I have
only seen this since upgrading to 2.2.18 (as did the person who wrote a
similar email in the list a few months back).

In addition, I belive this error only occurs once I go past the 1024
filedescriptor point; here is what I have done to increase the FD limit
under 2.2.18:

ulimit -HSn 4096 -n 4096 (in rc.local as well as my RunCache script)
echo 4096 >/proc/sys/fs/file-max (in rc.local)
edited /usr/include/bits/types.h to change __FD_SETSIZE to 4096
compiled squid.

Interestingly enough, with 2.2.16 I don't have to edit
/usr/include/bits/types.h, I leave the default 1024 value and it will
still see 4096 file descriptors with 2.3STABLE4 (2.4PREXXX does only
compile with 1024 though). However, if I do not edit types.h before
compiling or using squid on 2.2.18, it will only allow me to use 1024
FDs no matter what I cat to /proc/sys/fs/file-max.

Is there a link?

Thanks again,

Chris

> Henrik Nordstrom
> Suqid hacker
>
> Chris Conn wrote:
> >
> > Hello,
> >
> > I am searching for a problem with messages in cache_log and have come
> > across this:
> >
> > http://www.squid-cache.org/mail-archive/squid-users/200101/0303.html
> >
> > ========
> > I have recently upgraded from Linux 2.2.16 to 2.2.18 and have begun
> > suddenly getting a TON (about 20 per minute) of these:
> >
> > 2001/03/20 17:04:10| comm_accept: FD 4: (104) Connection reset by peer
> > 2001/03/20 17:04:10| httpAccept: FD 4: accept failure: (104) Connection
> > reset by
> > peer
> > 2001/03/20 17:04:10| comm_accept: FD 4: (104) Connection reset by peer
> > 2001/03/20 17:04:10| httpAccept: FD 4: accept failure: (104) Connection
> > reset by
> > peer
> > 2001/03/20 17:04:12| comm_accept: FD 4: (104) Connection reset by peer
> > 2001/03/20 17:04:12| httpAccept: FD 4: accept failure: (104) Connection
> > reset by
> > peer
> >
> > I am running S2.3STABLE4 and had never seen until today these messages.
> > I have checked my FDs and I have 4096 useable and about 3000 available.
> >
> > Any ideas?
> >
> > Thanks,
> >
> > Chris
Received on Wed Mar 21 2001 - 15:04:46 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:58:46 MST