Hello Daniel!
 
We have managed to track the problem down.
What causes our troubles lies in the file timer.c in the function timer_handler.
Or to be more exact in the for-loop:
 
for (prev_ticks=prev_ticks+1; prev_ticks!=saved_ticks; prev_ticks++)
  timer_run(prev_ticks);
 
In our case, after the NTP time change, this tries to bring the prev_ticks (representing the year 1970) up to the saved_ticks (our current time provided by NTP) tick by tick.
This is quite some calculating for 45 years. Especially on an embedded device.
 
As a workaround we have added an additional if-clause in the adjust_ticks function that simply ignores if the 'delta' between system time and ticks gets too high.
We have run a first test where this workaround seems to be ok but we'll run some further tests.
 
Do you think this workaround might cause some problems? I.e. might there be a problem if the system time jumps ahead too far but we do not adjust the ticks?
(We hope not because a time-leap backwards is ignored too :) )
 
Apart from our workaround:
Do you think there might be a way to handle this problem less CPU consuming (or to bring the ticks up to date in a bigger bulk)?
 
Best regards,
Stefan


Gesendet: Donnerstag, 22. Oktober 2015 um 09:50 Uhr
Von: "Daniel-Constantin Mierla" <miconda@gmail.com>
An: "Kamailio (SER) - Users Mailing List" <sr-users@lists.sip-router.org>
Betreff: Re: [SR-Users] 100% CPU usage after NTP date update
Hello,

can you do 'top' and see what is the pid of the process using a lot of
cpu, then get the backtrace with gdb?

gdb /path/to/kamailio PID
bt full

It will help to see what kamailio is trying to do.

Cheers,
Daniel

On 22/10/15 10:45, Stefan Kohlhauser wrote:
> Hello everyone!
>
> Setup:
> I am currently using a Kamailio 4.2.3 on an embedded device (low CPU and RAM). The device is not able to store the current date, so after a reboot its 1. Jan 1970 until NTP updates the date.
>
> Problem:
> If the date is updated after the Kamailio has started (a leap of 45 years) the Kamailio uses up all the CPU and renders the device useless for a very long time.
>
> Notes:
> There are no incomming requests at the time. (But the Kamailio is not responsive during that time anyway.)
> Adjusting the internal Kamailio-time seems to take longer when the date update leap is bigger.
>
> My questions:
> 1) Can I prevent the Kamailio from using up the entire CPU after a date update? Or maybe do the internal ticks-update more gracefully?
> 2) For my better understanding: What happens within the Kamailio after adjust_ticks has updated the internal date?
>
> Thanks in advance for your help!
>
> Best regards,
> Stefan
>
> _______________________________________________
> SIP Express Router (SER) and Kamailio (OpenSER) - sr-users mailing list
> sr-users@lists.sip-router.org
> http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-users

--
Daniel-Constantin Mierla
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda
Book: SIP Routing With Kamailio - http://www.asipto.com


_______________________________________________
SIP Express Router (SER) and Kamailio (OpenSER) - sr-users mailing list
sr-users@lists.sip-router.org
http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-users