RE: CustomLog to a pipe question

tnx, but isn't there a way to overcome this? I have more of these error
messages than I have logs :). Doesn't it loose that log when it fails
connecting the first time. It's kind of strange to ignore such serious error
messages. How will I know if it really fails?
Janis Putrams

-----Original Message-----
From: Jose R. Ilabaca [mailto:[email protected]]
Sent: Saturday, November 22, 2003 12:07 AM
To: [email protected]
Cc: 'Janis Putrams'
Subject: RE: [mls-users] CustomLog to a pipe question


Every a new httpd process is started the error log receives that error,
but this is not a problem.


-----Mensaje original-----
De: [email protected]
[mailto:[email protected]] En nombre de Janis Putrams
Enviado el: Viernes, 21 de Noviembre de 2003 17:30
Para: George Schlossnagle; Leonardo Herrera
CC: [email protected]
Asunto: RE: [mls-users] CustomLog to a pipe question


Thank you very much for advice. I did as y suggested but still have some
problems. I have three web servers and one of them is writing log. On
each server I have /etc/spread.conf

> Spread_Segment  192.168.123.95:4803 {
>         srva 192.168.123.69
>         srvb 192.168.123.71
>         srvc 192.168.123.70
> }
> EventLogFile = testlog.out
> DaemonUser = spread
> DaemonGroup = spread

on each server I run: spread &
and on each server I have in httpd.conf
        <IfModule mod_log_spread.c>
         SpreadDaemon 4803
         CustomLog $www virtualhosts
        </IfModule>
But still on the weakest server from time to time I get
[Fri Nov 21 21:58:54 2003] [error] (2)No such file or directory: Could
not connect to spread  with private_name ap27959. Error -6 [Fri Nov 21
21:58:54 2003] [error] (2)No such file or directory: Could not connect
to spread  with private_name ap27958. Error -6 When I run spuser and
join www I see some logs coming in. When doing netstat -a I see that
it's using unix socket to connect to
spread:
unix  3      [ ]         STREAM     CONNECTED     143538335 /tmp/4803
and I have average 35 httpd processes running now and I have same count
of
those:

# ps aux|grep "/usr/sbin/httpd"|wc -l
     35
# netstat -a|grep "/tmp/4803"|wc -l
     33
# uptime
 10:05pm  up 139 days,  5:34,  4 users,  load average: 0.27, 0.21, 0.34
# cat /proc/cpuinfo |grep MHz
cpu MHz         : 549.849
# ss|grep  requests/sec
   1.3 requests/sec - 29.6 kB/second - 22.7 kB/request

Sometimes httpd process count jumps to 100 and I would like it to handle
that as well. I would really appreciate if somebody could help me.

janis


-----Original Message-----
From: [email protected]
[mailto:[email protected]]On Behalf Of George
Schlossnagle
Sent: Thursday, November 06, 2003 6:18 PM
To: Leonardo Herrera
Cc: [email protected]
Subject: Re: [mls-users] CustomLog to a pipe question


On Nov 6, 2003, at 11:12 AM, Leonardo Herrera wrote:

> George Schlossnagle wrote:
>> On Nov 6, 2003, at 10:41 AM, Janis Putrams wrote:
>>> Hi!
>>> Is mod_log_spread working fine for y? Because I installed it
>>> recently and got connection failures on heigh load. Posted to
>>> mailing list but had no
>>> answer so far. Sad but true
>> Didn't see your mail.  Can you define 'high load'?  Specifically, how

>> many children do you have trying to send messages at peak? George
>
> I think he meant this:
> http://lists.backhand.org/pipermail/mls-users/2003-October/000210.html

Sounds like he is overconnecting to a single daemon (as is described in
README.spread in m_l_s sources).  It's not advisable to run a single
daemon and have everyone connect to it.  Much better to have every host
run their own daemon, all belonging to the same spread ring.

George


_______________________________________________
mls-users mailing list
[email protected]
http://lists.backhand.org/mailman/listinfo/mls-users



_______________________________________________
mls-users mailing list
[email protected]
http://lists.backhand.org/mailman/listinfo/mls-users