[mapserver-users] Mapcache segmentation faults under high load

[Pieter Callewaert]

Hi,

I’ve been testing further, on other servers to eliminate hardware failures, but I keep getting consistent this problem.
I made now consistent the same core dumps, I think my first core dump was not a correct one.

http://pastebin.com/L0Lhru4m

You can see at start:
1. Program terminated with signal SIGSEGV, Segmentation fault.
2. #0  process_socket (my_thread_num=42, my_child_num=<optimized out>, cs=0x7fee0d1042b8, sock=<optimized out>, p=<optimized out>, thd=<optimized out>)
3.     at event.c:1064

If you scroll down, the last thread is this thread, but I'm not sure yet what the problem exactly is. Any pointers?

Kind regards,
Pieter Callewaert

From: thomas bonfort [mailto:thomas.bonfort at gmail.com] 
Sent: Tuesday, 7 March, 2017 17:56
To: Pieter Callewaert <pieter.callewaert at be-mobile.com>; mapserver-users at lists.osgeo.org
Subject: Re: [mapserver-users] Mapcache segmentation faults under high load

this doesn't seem to be a segfault, but the server that received a kill signal... could you be getting out of memory errors ?

--
thomas 

On Tue, Mar 7, 2017 at 8:18 AM Pieter Callewaert <pieter.callewaert at be-mobile.com> wrote:
Thanks Andreas, you brought me on a lead, indeed the ulimit for core size was on 0. Also it seems ubuntu running apport does not generate coredumps, had to change the core pattern.

I've a core dump now, checked with GDB, but not sure what I'm searching at...
http://pastebin.com/fAXdEg15

Kind regards,
Pieter C

-----Original Message-----
From: Eichner, Andreas - SID [mailto:Andreas.Eichner at sid.sachsen.de]
Sent: Monday, 6 March, 2017 15:40
To: Pieter Callewaert <pieter.callewaert at be-mobile.com>; mapserver-users at lists.osgeo.org
Subject: AW: [mapserver-users] Mapcache segmentation faults under high load

Hi Pieter,

have you checked the ulimit for the user the Apache runs under? See man core(5) for possible reasons why no core is dumped.

HTH

> -----Ursprüngliche Nachricht-----
> Von: mapserver-users [mailto:mapserver-users-bounces at lists.osgeo.org]
> Im Auftrag von Pieter Callewaert
> Gesendet: Montag, 6. März 2017 11:01
> An: mapserver-users at lists.osgeo.org
> Betreff: Re: [mapserver-users] Mapcache segmentation faults under high
> load
>
> Hi Thomas,
>
>
>
> The documentation for debug compiling is not complete
> (http://mapserver.org/mapcache/install.html
> <http://mapserver.org/mapcache/install.html> ).
>
> If I add -DCMAKE_BUILD_TYPE="Debug" to the cmake, it this enough?
>
> Also added CoreDumpDirectory /tmp/apache2-gdb-dump to apache2.conf,
> and now I see this in the error log:
>
>
>
> [Thu Mar 02 08:46:42.588518 2017] [core:notice] [pid 13:tid
> 140052444931968] AH00051: child pid 14 exit signal Segmentation fault
> (11), possible coredump in /tmp/apache2-gdb-dump
>
> [Thu Mar 02 08:46:55.606507 2017] [core:notice] [pid 13:tid
> 140052444931968] AH00051: child pid 2097 exit signal Segmentation
> fault (11), possible coredump in /tmp/apache2-gdb-dump
>
>
>
> Unfortunately, no dumps yet, so I’m afraid I did something wrong with
> the debug compiling.
>
>
>
> Kind regards,
>
> Pieter C
>
>
>
> From: thomas bonfort [mailto:thomas.bonfort at gmail.com]
> Sent: Wednesday, 1 March, 2017 09:33
> To: Pieter Callewaert <pieter.callewaert at be-mobile.com>; mapserver-
> users at lists.osgeo.org
> Subject: Re: [mapserver-users] Mapcache segmentation faults under high
> load
>
>
>
> Try to get a backtrace from the crashes by compiling mapcache in debug
> mode, and configuring apache to allow coredumps. The "stale lock"
> messages you are seeing are side effects of the crashes so that's not
> where I'd investigate at first.
>
>
>
> regards,
>
> thomas
>
>
>
>

_______________________________________________
mapserver-users mailing list
mapserver-users at lists.osgeo.org
https://lists.osgeo.org/mailman/listinfo/mapserver-users