[asterisk-users] Asterisk CLI unresponsive

[Jonas Kellens]

On 02/03/2012 04:32 PM, Paul Belanger wrote:
> On 12-02-03 09:53 AM, Jonas Kellens wrote:
>> On 02/03/2012 03:48 PM, Paul Belanger wrote:
>>> On 12-02-03 09:05 AM, Jonas Kellens wrote:
>>>> On 02/03/2012 02:52 PM, Paul Belanger wrote:
>>>>> On 12-02-03 07:55 AM, Jonas Kellens wrote:
>>>>>> This is a production server. Will it affect theserver ?I already 
>>>>>> have
>>>>>> "dont_optimize" checked in the debug options.
>>>>>>
>>>>> Yes, reproduce the issue on your test infrastructure. then generate a
>>>>> backtrace [1].
>>>>>
>>>>> [1] https://wiki.asterisk.org/wiki/display/AST/Getting+a+Backtrace
>>>>>
>>>>
>>>> How can I reproduce when I don't know what causes the "deadlock" ??
>>>>
>>>> That is part of my question : what can have caused this "deadlock" ?
>>>> I can go on for weeks without any problem, and then suddenly there's
>>>> what appears to be a "deadlock".
>>>>
>>> That's what the backtrace will tell us. If you already have
>>> DONT_OPTIMIZED compiled on your production server, then attach gdb to
>>> the process and generate the backtrace.
>>
>> I can not reproduce a problem which I don't know what causes it.
>>
>> These seems like the chicken and the egg.
>>
>> There is no core-file generated. It's just the CLI that freezes, not a
>> restart of Asterisk.
>>
> You don't need a core-file for a deadlock, you are confusing it was a 
> crash.
>
> Regardless, my previous comment was not complete.  There is nothing 
> you can do until you recompile and enable DEBUG_THREADS (information 
> available from the previous wiki link). And yes, this will cause a 
> performance hit to your system,

But in general, are there no general causes to a "deadlock" or the 
Asterisk-CLI that freezes ??

I read on the wiki :
/A program is in deadlock when it is in a state where one or more of its 
components are waiting on something they will never get. I don't now the 
internal structure of Asterisk but I believe it is multi-threaded. A 
simple example of a deadlock is if a thread A were waiting on a resource 
that thread B has locked, and thread B is waiting for something thread A 
has locked./

But I see no such information in my verbose/messages/debug-logfiles

Can I create a deadlock with bad programming in the dialplan ?? I have 
already replaced all macro's with GoSub where possible.

There were 36 calls... Only 15% CPU (8 cores) and Asterisk uses about 
750000Kb of memory (4 GB available).


Jonas.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-users/attachments/20120203/70a1a795/attachment.htm>