Follow up:
I have discovered 'my' problem with the "Signaled to Death". It turned out to be
the kernel. I prefer to update a new install with the latest 'stable' version of the
kernel, however, it appears that the later kernels have problems with the virutal
memory. I use redhat so I reverted to the latest Redhat kernel (2.4.9-21) and
everything is working like a champ. Here is a breakdown of how I configured.\
1. Install redhat with all available/needed rpms.
2. Upgrade the kernel to 2.4.9-21 (if using redhat)
2. Create an additional link /usr/lib/libdb-3.so -> /usr/lib/libdb-3.2.so
3. Get the src rpm for cyrus-sasl and force it to use db-3 in the
configure.in and in the spec file.
4. Rebuild sasl and re-install (either rpm -e all cyrus-sasl* or rpm -U)
5. Rebuild cyrus-imap and force it to use db3
6. Install cyrus-imap
You should be on your way...I had 6 server with the same problem and once the
above steps were done, all have been running with no errors for over 2 weeks now ( I
was seeing 4 to 5 deaths per day).
As for the kernel, I just get the kernel-source-2.4.9-21.i386.rpm which
installs the kernel source tree in /usr/src/linux-2.4 and re-compile. This way I can
remove all junk I don't need. From what I understand, the redhat kernels have the ac
patches which revert back to the old virutal memory code.
If anyone would like my rpm-spec files for sasl and imap, I will be glad to
post them or send them directly.
Vernon Fort
-----Original Message-----
From: Vernon A.. Fort
Sent: Thursday, January 31, 2002 3:49 PM
To: [EMAIL PROTECTED]
Subject: RE: Signaled to Death by 11
I had another thought.
1. I use postfix and cyrus-imap
2. Both running processes are called master
3. Since I recompled using db-3, the death messages much less frequent.
4. using the pidof command reports the postfix master when cyrus is down
even when you specify the full path pidof /usr/cyrus/bin/master when postfix is
running, it reports the postfix command.
Someone on the postfix users group stated he renamed the cyrus master form master to
cyrusmaster and some of his problems went away. I noticed the past few morning that
the cyrus/master was reporting the lmtpd signal on 3 different servers and all between
4 and 6am. I wonder if some system shell scripts running out of cron (the
/etc/cron.daily, hourly, etc) are trying to send a -HUP signal to the postfix master
and since I always stary cyrus first, its getting the cyrus master instead. Most cron
scripts use the /etc/rc.d/init.d/functions which function uses the pidof command
throughout.
On one server I have renamed the cyrus/master to /usr/cyrus/bin/cyrusmaster and
everything still seems to be functional from the email perspective. Mail still gets
delivered and users can still pop and or imap. Each one of these server reports the
death each morning. I will see if its stops tommorrow morning on this server and
continues on the other 2.
Anyone else have any thoughts?
Vernon Fort
-----Original Message-----
From: Michael Fair [mailto:[EMAIL PROTECTED]]
Sent: Thursday, January 31, 2002 12:55 PM
To: Vernon A.. Fort; [EMAIL PROTECTED]
Subject: Re: Signaled to Death by 11
Just a shot in the dark, but is Postfix using
Berkeley DB for anything?
LMTP is shared among both applications and if
either needs to access the same DB then one or
the other will surely choke....
If master is dying then perhaps it is accessing
a DB from Postfix and Postfix is using an older
Berkely version?
-- Michael --
----- Original Message -----
From: "Vernon A.. Fort" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Friday, January 25, 2002 10:14 AM
Subject: Signaled to Death by 11
> Hello all,
> I have a problem for which has consumed me. The master process is
reporting the following:
>
> Jan 25 09:58:33 mail master[8412]: process 18150 exited, signaled to death
by 11
>
> What I have done:
>
> 1. Re-compiled both cyrus-sasl and cyrus-imap using db3
>
> [root@mail bin]# ldd master
> libssl.so.2 => /lib/libssl.so.2 (0x40033000)
> libcrypto.so.2 => /lib/libcrypto.so.2 (0x40060000)
> libdb-3.2.so => /lib/libdb-3.2.so (0x40123000)
> libresolv.so.2 => /lib/libresolv.so.2 (0x401ca000)
> libcom_err.so.2 => /lib/libcom_err.so.2 (0x401dc000)
> libc.so.6 => /lib/i686/libc.so.6 (0x401df000)
> libdl.so.2 => /lib/libdl.so.2 (0x4031a000)
> /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000)
> [root@mail bin]#
>
> [root@mail lib]# ldd libsasl.so.7.1.8
> libdb-3.2.so => /lib/libdb-3.2.so (0x40027000)
> libdl.so.2 => /lib/libdl.so.2 (0x400ce000)
> libcrypt.so.1 => /lib/libcrypt.so.1 (0x400d2000)
> libpam.so.0 => /lib/libpam.so.0 (0x400ff000)
> libc.so.6 => /lib/i686/libc.so.6 (0x40107000)
> /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x80000000)
> [root@mail lib]#
>
> From what I can tell, both the master and libsasl are using db3.2 All
other imap related programs report the same from ldd.
>
> 2. Increased the menory from 128 to 384meg. The problem still occures,
however, only 3-6 times per day verses once or twice per hour.
>
> After monitoring, the process only dies when a message is sent to mutiple
receipents. Once I restart cyrus, the mail queues flush correctly and all
mail is delivered. It appears to be a memory leak issues or the master is
referencing an invalid memory assignment, however, I am not a programmer.
Cyrus runs very well for a period of time and then just stops. The last
message in the postfix logs is always to mutiple receipents.
>
> System configuration:
>
> 1. Redhat 7.2 with all redhat updates available using SGI xfs file
system.
> 2. cyrus-sasl-1.5.24 and cyrus-imapd-2.0.16
> 3. Pentium III 866mgz with 384meg of memory 20gig drive.
> 4. The Postfix MTA using lmtp for the delivery Unix-domain sockets, not
TCP
> 5. Imap using PAM authentication.
>
>
> Any help would be greatly appreciated!!!!!
>
> --------------------------------------------------
> Vernon A. Fort
> Jobsoft Design & Development Inc.
> http://www.jobsoft.com
> (615) 905-9559 ext 19
>
>
