[tpop3d-discuss] signal 11?

Chris Lightfoot chris at ex-parrot.com
Mon, 23 Jul 2001 20:01:49 +0100


On Mon, Jul 23, 2001 at 11:28:25AM -0700, Nancy Pettigrew (work) wrote:
> >>Errm. Can you verify that it was the main server process
> >>which crashed, and not a child process serving a client?
> 
> It was both, apparently a couple different problems.  The parent process
> finally gave me a core file - backtraced it, and yes indeed, the
> uninitialized y variable was a problem.  Changed the code to init all that
> stuff at the beginning of auth_mysql.c and put it back out there.  However,
> the logs kept showing those pesky signal 11s, but the parent process never
> dies.  After a handful of signal 11s are reported in the logs, the pop
> server stops accepting connections for about 5 minutes (I'm hoping this will
> happen again today, so I can see if it's actually rejecting connections, or
> accepting and timing out).  Then, it starts handling connections again, and
> things continue on their merry way.  This only happens during peak load
> time, possibly related to MySQL connections (just a hunch at this point, saw
> a few Lost Connections, but can't verify causality yet).  Still
> investigating...

OK, that does sound like the child processes are dying
somewhere. I wouldn't expect it to be related to MySQL
connections, since the MySQL connection is only used in
the parent. I can't explain the failing to accept
connections bits, unless it is somehow related to handling
of child processes in the parent. Does the log mention
`connection rejected owing to high load' at all?

-- 
Chris Lightfoot -- www.ex-parrot.com/~chris/
 ``The practical scientist is trying to solve tomorrow's problem with
   today's computer; the computer scientist... often has it the other way
   around.'' (from `Numerical Recipes in C')