[SR-Users] rtimer execution period - round two

[Alex Balashov]

On 07/25/2012 03:32 AM, Alex Balashov wrote:
> On 07/24/2012 02:59 AM, Daniel-Constantin Mierla wrote:
>
>> The way mqueue is consuming items from queue does not ensure queue
>> locking until all items are consumed. So the first item there is removed
>> from queue and cloned in private memory where is accessible via PV, then
>> the queue is unlocked so others can consume or add to it.
>>
>> Not being in office for few days, I cannot run tests myself ... do you
>> have heavy writers? Can you add few more timer processes to consume from
>> queue and see what happens?
>
> I figured it out;  Alex Hermann nailed it.  The number of accounting
> events I was generating was tripping max_while_loops.  I thought I had
> it set high enough, but I guess I didn't.  I didn't catch the log
> message because it was being lost in thousands of lines per second of
> logging.

There is an occasional situation where the queue empties out, 
while(mq_fetch(...)) returns false, and an instant later, something is 
added back to the queue, but now we must wait 1 rtimer task interval for 
event writing to resume.

However, the ability to do microsecond-based intervals in Kamailio 3.3 
has made this problem largely irrelevant.  :-)

-- 
Alex Balashov - Principal
Evariste Systems LLC
235 E Ponce de Leon Ave
Suite 106
Decatur, GA 30030
Tel: +1-678-954-0670
Fax: +1-404-961-1892
Web: http://www.evaristesys.com/, http://www.alexbalashov.com/