[OpenSIPS-Users] OpenSIPS core dumps
thrillerbee
thrillerbee at gmail.com
Sat Nov 13 23:20:54 CET 2010
Bogdan,
Since I made those changes, it doesn't core dump anymore - it just runs out
of memory and stops processing packets. It happened on both of my proxies -
the mem dump can be downloaded here:
http://www.starviewconnect.com/tmp/core1_mem_dump_20101112.gz
http://www.starviewconnect.com/tmp/core2_mem_dump_20101112.gz
Thanks.
On Thu, Nov 11, 2010 at 12:43 PM, thrillerbee <thrillerbee at gmail.com> wrote:
> Bogdan,
>
> I had already increased the PKG_MEM_POOL_SIZE by *10. I'm making the
> changes suggested in the link you provided to try to narrow down the memory
> issue. I'll send over as soon as I have another crash with mem info.
>
> Thanks.
>
>
> On Thu, Nov 11, 2010 at 5:10 AM, Bogdan-Andrei Iancu <
> bogdan at voice-system.ro> wrote:
>
>> This last crash had the same bt as the one from previous email ? the prev
>> fix I made reports this:
>>
>> ERROR:db_flatstore:flat_db_insert: uninitialized connection
>> and does not crash, so this time the crash may be in a different place.
>>
>> Now, it seams after all that the root of your problem is the mem
>> exhaustion. To see what causes this (if a mem leak or simply not enough
>> mem), there is a doc - http://www.opensips.org/Resources/DocsTsMem (it is
>> for shm, but the same for pkg).
>>
>> Maybe, first you should simply try to increase the pkg mem ( in config.h
>> you have PKG_MEM_POOL_SIZE that you can increase) - you NEED to recompile
>> and reinstall after that.
>>
>> If more pkg mem does not solve the problem, I can help you with using the
>> memory debugger to see what is going on there.
>>
>> Regards,
>> Bogdan
>>
>>
>> thrillerbee wrote:
>>
>>> My other proxy crashed as well with these ERRORs in the syslog:
>>>
>>> Nov 10 22:01:02 core2 /usr/local/sbin/opensips[22959]:
>>> ERROR:db_flatstore:get_name: pkg memory allocation failure
>>> Nov 10 22:01:02 core2 /usr/local/sbin/opensips[22959]:
>>> ERROR:db_flatstore:flat_reopen_connection: failed to get_name
>>> Nov 10 22:01:02 core2 /usr/local/sbin/opensips[22959]:
>>> ERROR:db_flatstore:flat_db_insert: uninitialized connection
>>> Nov 10 22:01:02 core2 /usr/local/sbin/opensips[22959]:
>>> ERROR:db_flatstore:flat_db_insert: uninitialized connection
>>> ...
>>> Nov 10 22:01:21 core2 /usr/local/sbin/opensips[22959]:
>>> ERROR:db_flatstore:flat_db_insert: uninitialized connection
>>> Nov 10 22:01:22 core2 /usr/local/sbin/opensips[22959]:
>>> ERROR:db_flatstore:flat_db_insert: uninitialized connection
>>> Nov 10 22:01:22 core2 /usr/local/sbin/opensips[22959]:
>>> ERROR:db_flatstore:new_flat_id: no pkg memory left
>>> Nov 10 22:01:22 core2 kernel: [4297088.404734] opensips[22959]: segfault
>>> at 10 ip 7f3db577e21f sp 7fffa260d640 error 4 in
>>> db_flatstore.so[7f3db577b000+5000]
>>>
>>> On Wed, Nov 10, 2010 at 10:19 AM, thrillerbee <thrillerbee at gmail.com<mailto:
>>> thrillerbee at gmail.com>> wrote:
>>>
>>> Bogdan,
>>>
>>> Well, I spoke too soon - it's not just an issue with the
>>> opensipsctl fifo calls - looks more like a memory leak. It
>>> crashed again today, but I did get some errors in the syslog this
>>> time right before the crash:
>>> Nov 10 15:42:32 core1 /usr/local/sbin/opensips[27044]:
>>> ERROR:db_flatstore:new_flat_id: no pkg memory left
>>> Nov 10 15:42:32 core1 kernel: [5508366.582447] opensips[27044]:
>>> segfault at 10 ip 7fa7ff74c21f sp 7fffdc101700 error 4 in
>>> db_flatstore.so[7fa7ff749000+5000]
>>> To be thorough, I've attached the backtrace & output from print
>>> commands (although they're the same as before).
>>>
>>> To answer your question, yes - I do use the flat_rotate MI command.
>>>
>>> Thanks again.
>>>
>>> On Wed, Nov 10, 2010 at 4:04 AM, Bogdan-Andrei Iancu
>>> <bogdan at voice-system.ro <mailto:bogdan at voice-system.ro>> wrote:
>>>
>>> Hi,
>>>
>>> opensipsctl takes care that each command takes a separate fifo
>>> reply, so here it should be no problem. But the problem may be
>>> when comes with sending multiple commands (via FIFO) in the
>>> same time - this translates into parallel writes to the same
>>> file and depends on the atomicity of the write op.
>>>
>>> But in the worst case, a mixture at the FIFO level may lead to
>>> bogus command and not in any kind of crash....Do you use the
>>> "flat_rotate" MI command ?
>>>
>>> Regards,
>>> Bogdan
>>>
>>> thrillerbee wrote:
>>>
>>> Bogdan,
>>>
>>> It seems the issue is with 'opensipsctl fifo' - it's very
>>> sensitive to simultaneous calls. Basically, I've combined
>>> all my scripts to prevent 'opensipsctl fifo' from being
>>> called too frequently and that seems (so far) to have
>>> mitigated the issue. Is there anything one should know
>>> about how (not) to use /opensipsctl/?
>>>
>>> Thanks.
>>>
>>> On Mon, Nov 8, 2010 at 6:07 AM, Bogdan-Andrei Iancu
>>> <bogdan at voice-system.ro <mailto:bogdan at voice-system.ro>
>>> <mailto:bogdan at voice-system.ro
>>> <mailto:bogdan at voice-system.ro>>> wrote:
>>>
>>> Hi,
>>>
>>> strange if you do not have any errors :(....
>>>
>>> I just made a fix on both trunk and 1.6 to extend some
>>> checks in
>>> flatstore and prevent crashing (even if the DB op will
>>> not be
>>> executed).
>>>
>>> Could you update from SVN and see if stops crashing ?
>>>
>>> Regards,
>>> Bogdan
>>>
>>>
>>> ------------------------------------------------------------------------
>>>
>>> _______________________________________________
>>> Users mailing list
>>> Users at lists.opensips.org
>>> http://lists.opensips.org/cgi-bin/mailman/listinfo/users
>>>
>>>
>>
>>
>> --
>> Bogdan-Andrei Iancu
>> OpenSIPS Bootcamp
>> 15 - 19 November 2010, Edison, New Jersey, USA
>> www.voice-system.ro
>>
>>
>> _______________________________________________
>> Users mailing list
>> Users at lists.opensips.org
>> http://lists.opensips.org/cgi-bin/mailman/listinfo/users
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opensips.org/pipermail/users/attachments/20101113/20a108cb/attachment-0001.htm>
More information about the Users
mailing list