I've also made these changes. The only solution was to move heavy traffic
sites to another server. I also think that I/O to the disk is a bottleneck.
I have tried replacing Qpopper with Dovecot and got the same results. On my
next server I plan on putting some enterprise Western Digital drives in to
see if they perform any better.
----- Original Message -----
From: "Rodrigo Ordonez Licona" <rodrigo (at mark) xnet.com.mx>
To: <coba-e (at mark) bluequartz.org>
Sent: Thursday, May 04, 2006 9:48 AM
Subject: [coba-e:05015] Re: The ongoing QPOPPER Debate
> To all
> =====================================
> We made that Change to xinetd,
> server_args = in.qpopper -s -R
>
> And also we did the PAM Correction:
> #%PAM-1.0
> #auth requisite /lib/security/pam_nologin.so
> #auth requisite /lib/security/pam_shells.so
> #auth required /lib/security/pam_pwdb.so shadow nullok
> #account required /lib/security/pam_pwdb.so
> auth required /lib/security/pam_stack.so service=system-auth
> account required /lib/security/pam_stack.so service=system-auth
>
> ===================================
>
> The pop authentication seemed to run smoother ,
>
> HOWEVER when 90 users tried to access pop server within one minute all
> turned into Hell.
>
> We think PAM under stress fails-slows not pop, Also at failure FTP shows
> setcred errors on the LOG(probably unrelated),
>
> Also during this period all login services that use pam, get slow or start
> timing out (including openwebmail, ftp, admserv)
>
> This suggestion was given in the past. Probably worked for a Few, but we
> think it is stress related(probably TCP), We removed the site whose users
> connected within the same minute and the problem went away, But a RAQ 550
> could handle the same 90 connections without problem.
> =========================================
> MICHAEL STAUBER --> any news about your IRC Server and max TCP Connections
> problem, in our case it seemed to be Related.
>
> Regards
>
> Rodrigo O
> Xnet / Pwworks
>
>
> -----Original Message-----
> From: Brian N. Smith [mailto:brian (at mark) nuonce.net]
> Sent: Jueves, 04 de Mayo de 2006 06:22 a.m.
> To: coba-e (at mark) bluequartz.org
> Subject: [coba-e:05011] The ongoing QPOPPER Debate
>
> To all,
>
> I have had this problem to. Yesterday I made a very small change, and I
> "think" it seems to have cleared it up some. KNOCK on wood, but I haven't
> received any Active Monitor emails about CPU Utilization, or any
> complaints
> at all.
>
> I know eventually it will be switched to Dovecot, but if anyone wants to
> try
> my "magical" fix.
>
> Edit: /etc/xinetd.d/pop3
>
> Change: server_args = in.qpopper -s
> To: server_args = in.qpopper -s -R
>
> The -R says NOT to do a reverse lookup. Like I said, remarkably it seems
> to
> have fixed it. I haven't seen 10-15 POP3 processes like I have been
> seeing.
> It seems to have made the box very happy. If someone who is having the
> same
> problem can try it, and let me know if I am smoking crack, or does this
> actually make a big difference?
>
> The only thing that I noticed to, is that PAM is just sometimes really
> slow.
> Any good how-tos on speeding it up?
>
> Thanks,
> Brian
>
>
>
>