Enrico Maistro ha scritto:
> Hi all,
> 
> 
> ** STUN Packet SENT 1439629545 1439629545

> ** STUN request not matching. ID: 1208953606 done at: 1221144741
> gotresponse 0:

> All of this result in a jump of memory utilization from 3% to 26%, 100%
> cpu utilization for 10 seconds and after few calls (3 to 5) in a nice
> core dump (300+ mb).
> 
> Incoming calls work flawless with no memory consumption, no strange stun
> messages, no 100% cpu usage.
> 
> Has anyone ever seen something like this?
> Suggestion on how to resolve or at least further debug the problem?
> Should i post a message in the -dev mailing list? or maybe open a ticket?


Techically speaking, for every invite, callweaver sends a stun requests
and stores it, parks the SIP message until a stun response is received
and, whenever the stun response for a request is avalaible, the SIP
message is "unlocked", modified and delivered.
This is a dirty hack because of chan_sip limitations.

It may happen that the list of stun requests that don't have a response
are kept in memory for a long time.

So if you do a lot of calls, you may have a large chain of stun requests
using memory that are useless.

You should see something like:

cw_log(LOG_DEBUG,"** STUN Sending request after acquiring STUN\n");

or

cw_log(LOG_DEBUG, "deleting this request of reqresp
%d\n",rr->p->stun_resreq_id);

in your logs.

If you're sure that your stun servers work well, try to lower down, in
stun.c, the 300 value to something like 10 or similar.

->> here        if (req_queue->whendone + 300 < now)

And tell if it works better (unless there is a leak in the newest code,
but i haven't seen it yet)

In all cases, i don't think that stun can fill the memory in some
seconds. There must me something else going on.

Max


_______________________________________________
Callweaver-users mailing list
[email protected]
http://lists.callweaver.org/mailman/listinfo/callweaver-users

Reply via email to