Hi Daniel!

Thanks for your reply!

I managed to get the kamctl trap output, it is attached here. This is a case where Kamailio is unresponsive with the UDP workers stuck in the mutex...

I agree with your observations, it goes along with what we were thinking...

Kind regards,

Patrick Wakano

On Tue, 27 Oct 2020 at 19:17, Daniel-Constantin Mierla <miconda@gmail.com> wrote:

The increase of memory is very likely the side effect of processes not handing the traffic, sip_msg_shm_clone() is manly used when creating the transactions. They can accumulate if the timer processes are stuck or other callback functions block the transactions hash table. As said in the other email, try to get the output of kamctl trap to have information about what all processes do at that moment.

Cheers,
Daniel