cyrus-master dies/exits on startup

[Wolfgang Breyha]

Hi!

I had this on 2.3.x already and today a brand new 2.4.11 cyrus-master died  as
well....

On my murder frontends sometimes (~1 out of 10) cyrus-master dies/exits(?!)
right after forking the initial set of processes defined in cyrus.conf.

This initial set of mupdated/idled/lmtpd/imapd/pop3d processes are left in the
system, but cyrus-master is gone afterwards.

After killing the leftovers and restarting cyrus-master he usually survives.
And he never died/exited if he survived the initial execs.

The 2.3.x frontends are RHEL5/i386 and the 2.4.11 are RHEL6/x86_64.

I thought it is the heavy initial load on our frontends causing this, but the
2.4.11 frontend has no load or connection at all currently.

cyrus-master is started with "-d -x /var/agentx/master -P 30"

The log of the master says nothing unusual:
> Sep 19 14:20:32 jarvis master[1738]: process started
> Sep 19 14:20:32 jarvis master[1747]: about to exec /usr/lib/cyrus-imapd/ctl_cyrusdb
> Sep 19 14:20:49 jarvis master[1786]: about to exec /usr/lib/cyrus-imapd/idled
> Sep 19 14:20:49 jarvis master[1738]: ready for work
> Sep 19 14:20:49 jarvis master[1789]: about to exec /usr/lib/cyrus-imapd/imapd
> Sep 19 14:20:49 jarvis master[1790]: about to exec /usr/lib/cyrus-imapd/imapd
> Sep 19 14:20:49 jarvis master[1791]: about to exec /usr/lib/cyrus-imapd/pop3d
.... and some more exec, but no note about exiting.

My cyrus.conf:
------------------------
START {
  recover	cmd="ctl_cyrusdb -r"
  idled		cmd="idled"
}

SERVICES {
  mupdate       cmd="mupdate" listen=3905 prefork=1

  Fimap		cmd="imapd" listen="imap" prefork=5
  Fimaps	cmd="imapd -s" listen="imaps" prefork=2
  Fpop3		cmd="pop3d" listen="pop3" prefork=3
  Fpop3s	cmd="pop3d -s" listen="pop3s" prefork=1
  Fsieveold	cmd="timsieved" listen="sieve-old" prefork=0
  Fsieve	cmd="timsieved" listen="sieve" prefork=0

  Flmtp		cmd="lmtpd" listen="lmtp" prefork=1 maxchild=30
}
....
------------------------

My 2.3.x backends run a similar config (except mupdate) and never died on
startup yet.

Has somebody else seen this behavior already or should I try to get more
details? As a first step I allowed coredumps.

Greetings, Wolfgang
-- 
Wolfgang Breyha <wbreyha at gmx.net> | http://www.blafasel.at/
Vienna University Computer Center | Austria