ASSP 2.1.1 12090 crashes

[myrolm]

Hi guys,

after update to 2.1.1 12090 ASSP random crashes with this error in log 1 time a day

*** glibc detected *** /usr/bin/perl assp.pl MainLoop - next: Wed Apr 4 04:34:35 2012: malloc(): memory corruption (fast): 0x00007fe0d52c41c0 ***

Previous stable was OK for >1 month.

Any ideas how to solve this?

[thockar]

the two most possibly reasons are:

- too less RAM - assp memory should be never paged out
- your perl is not compatible to the installed glibc (was not compiled with the same compiler and header files like glibc)

Thomas

[myrolm]

Hi,

my server has 6GB RAM, ASSP takes 1,7 G of it.

perl version is 5.10.1 on debian squeeze.

is there some problem with this configuration?

Miro.

[thockar]

>is there some problem with this configuration?

possibly yes - see the conversation with Anders below.

>perl version is 5.10.1 on debian squeeze.

upgrade your perl. Build 12090 uses some unicode perl core modules, which are not used in early versions. The unicode implementation in perl 5.10.1 is not the best.

Thomas


---------------------------------------
Hi Thomas
Thanks for your help and suggestions, I will look into this later when I
got the time.

and I know the golden rule "never touch....." as you said we always ignore
it.

------------------

Anders,

do you has upgraded only the kernel ?

Possibly the glibc.so was also upgraded. So it could be possible that your

perl is now somehow incompatibe with this new glibc. Have a look at the
linux distro, if there is also an upgrade for perl (related to this
kernel) available. Or even better, build perl from the source with the
current system. Oh, and if you want to do that, don't forget to recompile
openssl, imagemagic, tesseract, clamav, mysql and berkeleydb (what ever is

used of them) prior to that!

>I changed to the "2.6.32-5-686-bigmem #1 SMP" kernel

I can remember an old IT rule - " never touch a running...." - we all know

it and ignore it constantly

Thomas



Von: Anders Westin
An: ASSP development mailing list <assp-test@lists.sourceforge.net>
Datum: 03.04.2012 11:45
Betreff: Re: [Assp-test] Antwort: ASSP version 2.1.2(12087)
segment fault



Hi Thomas
Thanks for the answere!

and that make sense with the OS malloc(), beacuse the problem started
after I changed to the "2.6.32-5-686-bigmem #1 SMP" kernel in Debian
Squeeze .


hmm..... must think what todo next .




From: Thomas Eckardt
To: ASSP development mailing list <assp-test@lists.sourceforge.net>
Date: 2012-04-03 09:46
Subject: [Assp-test] Antwort: ASSP version 2.1.2(12087) segment
fault



Anders,

>malloc()

malloc() - there are two versions used in perl. The internal Perl malloc()



or the OS malloc(). Which version is used, is set at compile time of Perl.



The default is the OS malloc().
The difference is, the OS malloc() makes Perl able to give freedup memory
back to the OS - if this works or not depends on the used OS.

malloc() - requests at least one new memory page from the OS. An segfault
could be for example happen, if Perl requests a new memory page and the OS



is still swapping and it takes the OS too long to satisfy Perls malloc()
request.

An segfault could be also forced, if the OS allocates the new memory page
to perl, but the RAM is defect at this location.

Thomas




Von: Anders Westin
An: ASSP development mailing list <assp-test@lists.sourceforge.net>
Datum: 02.04.2012 18:10
Betreff: [Assp-test] ASSP version 2.1.2(12087) segment fault



Hi Thomas

ASSP version 2.1.2(12087)

I have problem with Segment fault in ASSP(perl) so I installed Enbugger
and added this to ASSP.PL use Enbugger::OnError;
and below is output from the screen when assp crashed, do you get
something out of this

Apr-02-12 15:49:04 [Worker_2] Disconnected: 10.33.65.125 - processing time




2 seconds
Apr-02-12 15:49:04 [Worker_1] 201.8.138.82
<KarsynHuckabay@howarddentalgroup.com> info: start damping (1 s)
Apr-02-12 15:49:04 [Worker_4] Connected: 84.38.68.35:53917 > x.x145.200:25




> x.x.145.198:125
Apr-02-12 15:49:04 [Worker_4] LDAP - found juergen.runkel@domain.com in
LDAPlist
Apr-02-12 15:49:04 [Worker_4] 84.38.68.35 <g.haid@dreher.de> renewing
tuplet: (84.38.68.0,dreher.de) age: 1s
*** glibc detected *** /usr/local/bin/perl assp.pl MainLoop - next: Mon
Apr 2 15:49:10 2012: malloc(): smallbin double linked list corrupted:
0x0fd1def0 ***
Segmentation fault
root@smtp01:/assp2#

[myrolm]

OK, so it looks like that upgrading to v2 was a mistake

anyway, I set number of workers from 5 to 2 aby it looks like it is running fine for now

Miro.

[thockar]

So I think you have large main hashes and lists with V2, without using a DB for them.
V2 writes warnings about this in to the startup log.

Thomas

[hbruun]

I'm continuing to see random crashes after upgrading ASSP to build 12090 and making sure all perl modules for 5.14.2 are up to date. My crashes started a couple of builds ago and I'm running on a Windows 2008 64 bit server using 32 bit Perl. It crashes 1-2 a day.

My earlier posting is at viewtopic.php?f=12&t=1586

[myrolm]

Good news,

In my case lowering number of workers helped, no crashes since than

On windows 32bit with 32bit perl 5.12.4 is everything OK with 5 workers.