org.mobicents.slee.resource.sip11.wrappers.ServerTransactionWrapper stays attached forever

breton

unread,

Dec 21, 2009, 7:45:40 PM12/21/09

to mobicents-public

Hi,

if anyone can help me out what might I be doing wrong.

I have sbb with onInvite event method(INVITE is an initial event with
initial event selector). So onInvite method calls some ejb part of an
app, informing it that invite arrived. Before that sender is informed
with 100 (Trying) by runnig this code:

Response response = messageFactory.createResponse(100, request);
ServerTransaction st = event.getServerTransaction();
st.sendResponse(response);

Problem is, that i can see on the management-console that activities
count of type
'org.mobicents.slee.resource.sip11.wrappers.ServerTransactionWrapper'
grows with every arrived INVITE.
It is all okay till ~5000 Invites are received (with a little help of
some sipp scenario). Than mobicents silently goes away, everything
seems to work, no error in the log, sipRa seems to listen to 5060
(netstat), but nothing is received further into the app.

Problem goes away(ServerTransactionWrapper does not grow in count) by
putting st.terminate() after sendResponse but I don't know if that is
correct way?

Thank you in advance for any advice & brg

Eduardo Martins

unread,

Dec 21, 2009, 9:45:38 PM12/21/09

to mobicent...@googlegroups.com

Replying 1xx does not terminates the server tx, only after a final response reply (and after a specific timer to handle retransmissions) will jain sip terminate the transaction. Regarding the non responsive container, it may be the heap getting out of memory or JAIN SIP stack limiting the number for these server transactions.

-- Eduardo

-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-
http://emmartins.blogspot.com
http://www.redhat.com/solutions/telco

aayush

unread,

Dec 22, 2009, 12:46:27 AM12/22/09

to mobicents-public

MAX_SERVER_TRANSACTION table size of JAIN SIP is 5000. So after 5000
odd invites, JAIN SIP is unconditionally dropping your requests.

80% of 5000 is the low water mark. JAIN SIP selectively drops requests
when the table size is between low water mark and high water mark.

On Dec 22, 7:45 am, Eduardo Martins <emmart...@gmail.com> wrote:
> Replying 1xx does not terminates the server tx, only after a final response
> reply (and after a specific timer to handle retransmissions) will jain sip
> terminate the transaction. Regarding the non responsive container, it may be
> the heap getting out of memory or JAIN SIP stack limiting the number for
> these server transactions.
>
> -- Eduardo
>

> -:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-http://emmartins.blogspot.comhttp://www.redhat.com/solutions/telco

breton

unread,

Dec 22, 2009, 10:39:10 PM12/22/09

to mobicents-public

Thank you both for pointing those facts out.

One further question - why those opened st are not closed at some
point (I kept the app running for several hours and those st are never
cleaned)
Is there something to set to timeout and clean those st as it seems
for me easier to do than checking 3rd party call controller behavior.
I saw there is some activity liveness check in the ActivityManager
(set to 1800s) but I dont know if and how does it work (and what is
the threshold to mark some activity as timeouted).

And FYI - container does not go unresponsive, it behaves exactly as
aayush pointed. It silently drops requests without any log and there
is no OutOfMemory info in the log.

laurent

unread,

Dec 30, 2009, 3:49:33 AM12/30/09

to mobicents-public

hello,

I remember to have the same problem in the past and I have done some
change in SIP stack and SIP config properties file.

If I remember correctly , the timer in the SIP stack in not configured
so he dont destroy not terminated transaction.

you can check this old thread:
http://groups.google.com/group/mobicents-public/browse_thread/thread/8e799953f099fba2/6ae48e8b76428ca5?lnk=gst&q=transaction+not+removed+in+#6ae48e8b76428ca5