Re: [OpenSIPStack] Double packets - opensbc

[Jeremy A <je...@el...>]

Not much difference I'm afraid.

I set SIP_TIMER_MANAGER to 20000 (20 milliseconds?)

The trace showed that the only packets sent by openSBC that weren't
duplicated were the DNS queries. The rest were duplicated at typically
10-15 microseconds delay


jo...@op... wrote:
> Hi Jeremy,
>
> I think I found out whats causing this early firing of retransmission.   
> The timer resolution set for SIPTimerManager is too low.  IT's currently 
> set to 500 ms which is as low as it could get.  This would increase the 
> chances of SIP Timer A to fire sooner.  Try setting #define 
> SIP_TIMER_RESOLUTION from 500 to 10 milliseconds in 
> SIPTimerManager.cxx.   Let me know if this solves it.  Thanks for 
> pointing it out.
>
> Joegen
>
> jo...@op... wrote:
>   
>> Ouch!  I guess you are right.   For a second there, I thought I was 
>> looking at milliseconds.    Let me try to replicate this issue.  For the 
>> mean time would you be able to recompile using a higher value for 
>> SIP_TIMER_T1.  Just look for #define SIP_TIMER_T1 500 in SIPTimer.h.  I 
>> want to be sure it's actually  Timer_A firing early instead of a rogue 
>> code in OpenSIPStack.
>>
>> Joegen
>>
>> Jeremy A wrote:
>>   
>>     
>>> Perhaps not?
>>>
>>> In your first example the time difference 0.030086 to 0.066302 /
>>> 0.073700 is of the order of 30-40 milliseconds
>>>
>>> In your second example the time difference 116.150431 to 116.159165 is
>>> 8734 microseconds or 8.734 milliseconds.
>>>
>>> I have looked carefully at the monitoring process to make sure it is not
>>> causing a false result. I can find no obvious problem.
>>>
>>> FYI The opensbc server is a VMWare virtual machine hosted on the
>>> monitoring server. The code running opensbc is the 1.1.4 linux release
>>> compiled natively and is the release version.
>>>
>>>