[Users] Dealing with reinvites

Mon Jan 2 18:10:30 CET 2006

I'm not sure what the standard says, but AFAIK the dialog is confirmed 
with 200 Ok from the first INVITE. Thus, Asterisk should accept the new 
reINVITE although the ACK was not received yet. (IMO, receiving an 
reINVITE imply that the previous 200 Ok was ACKnowledged).

IMO Asterisk should accept the reINVITE (just like when the ACK was 
received, thus stopping retransmissions). If then the ACK is received, 
the CSeq does not match, thus it should be ignored.

Which Asterisk version do you use? 1.2.1? Try to discuss the problem on 
the devel mailing list, and create a bug report on bugs.digium.com with 
suggested proposals how to fix the problem.

regards
klaus

Matt Schulte wrote:
> I see what you mean, and was worried that would be the answer .. :-)
> 
> 
>>The ACK is for first INVITE, so the the Asterisk should match it
> 
> against the first transaction it created.
> 
> Since the cseq's are different, Asterisk flips out, since it expects the
> 102 cseq before the 103 .. This is normal behavior, right?
> 
> 
>>It may be logically to drop the re-INVITE if the ACK for the first
> 
> invite didn't arrive or create a temporary transaction and wait for both
> ACKs.
> 
> Well, the cats out of the bag, this carrier doesn't want to disable
> reinvites.. I think they have altered the Asterisk code to send a
> reinvite ASAP as opposed to 'waiting' for the RTP stream to setup. We
> have never ever had this problem before.. For those of you wondering,
> the carrier is TXLink..
> 
> Any other suggestions? Hehe. Thanks much.
> 
> 	Matt
> 
> -----Original Message-----
> From: Daniel-Constantin Mierla [mailto:daniel at voice-system.ro] 
> Sent: Saturday, December 31, 2005 6:55 AM
> To: Matt Schulte
> Cc: users at openser.org
> Subject: Re: [Users] Dealing with reinvites
> 
> Hello,
> 
> without the full grep of the packages, I am doing just assumptions.
> 
> The ACK is for first INVITE, so the the Asterisk should match it against
> the first transaction it created. For the second INVITE, it should wait
> for another ACK. This is a common race in IP world due to different path
> routing and cannot be avoided, even if you fix something on openser, the
> delay may occur on the wire.
> 
> The ACK sent for a 200OK is a separate transaction, and it is routed
> according to Route headers. You cannot change it in OpenSER/SER unless
> you make the proxy to be call stateful, to act as an end User Agent.
> 
> I am not sure what the RFC says exactly about how the end User Agent
> must act. It may be logically to drop the re-INVITE if the ACK for the
> first invite didn't arrive or create a temporary transaction and wait
> for both ACKs.
> 
> Cheers,
> Daniel
> 
> 
> On 12/30/05 17:09, Matt Schulte wrote:
> 
>>All,
>>
>>	We have had a new issue arise within openser, basically we
> 
> connect to 
> 
>>a SIP carrier that uses Asterisk as a proxy. Naturally, this carrier 
>>is hellbent on offloading the RTP traffic, thus the reinvites.
>>My issue is that they appear to be sending the second invite "too 
>>fast", before the initial RTP even gets a chance to establish. This 
>>was working just fine before, and I believe they changed something. 
>>Anyway, they're refusing to help with this issue, so now I must "fix"
> 
> it on my end.
> 
>>OpenSER is in-part, the problem.. See below:
>>
>>(Sorry if formatting is off, dump from tethereal ..)
>>
>>
>>  0.000000 204.13.233.13 -> 206.80.70.47 SIP/SDP Request: INVITE 
>>sip:8886963856 at 206.80.70.47, with session description
>>  0.001752 206.80.70.47 -> 204.13.233.13 SIP Status: 100 trying -- 
>>your call is important to us
>>  0.002723 206.80.70.47 -> 206.80.70.54 SIP/SDP Request: INVITE 
>>sip:+1314xxxxxxx at sip.stl.netlogic.net, with session description
>>  0.010327 206.80.70.54 -> 206.80.70.47 SIP Status: 100 Trying
>>  0.292175 206.80.70.54 -> 206.80.70.47 SIP Status: 180 Ringing
>>  0.292692 206.80.70.47 -> 204.13.233.13 SIP Status: 180 Ringing
>>  2.016246 206.80.70.54 -> 206.80.70.47 SIP/SDP Status: 200 OK, with 
>>session description
>>  2.016893 206.80.70.47 -> 204.13.233.13 SIP/SDP Status: 200 OK, with 
>>session description
>>  2.050871 204.13.233.13 -> 206.80.70.47 SIP Request: ACK
>>sip:+1314xxxxxxx at 206.80.70.54
>>  2.051233 204.13.233.13 -> 206.80.70.47 SIP/SDP Request: INVITE
>>sip:+1314xxxxxxx at 206.80.70.54  2.051827 206.80.70.47 -> 204.13.233.13 
>>SIP Status: 100 trying
>>--<<KABOOM>>--
>>  2.051956 206.80.70.47 -> 206.80.70.54 SIP/SDP Request: INVITE 
>>sip:+1314xxxxxxx at 206.80.70.54,
>>  2.052112 206.80.70.47 -> 206.80.70.54 SIP Request: ACK
>>sip:+13142664000 at 206.80.70.54
>>  2.053028 206.80.70.54 -> 206.80.70.47 SIP/SDP Status: 200 OK, with 
>>session description
>>  2.053336 206.80.70.54 -> 206.80.70.47 SIP Status: 503 Server error
>>
>>
>>Where you see "kaboom", is the problem. If you notice, the time 
>>between the first ACK and second INVITE is about 1ms .. While this 
>>shouldn't be a problem, you'll notice when the ACK/INVITES are getting
> 
> 
>>statefully forwarded, they're getting sent out of order. The end point
> 
> 
>>is an Asterisk machine, thus the 503 error.
>>
>>Carrier_Ast --> OpenSER --> Netlogic_Ast
>>
>>I am using loose_route to forward, just like everyone else, they are 
>>even hitting loose_route in the correct order.. Thoughts? Suggestions?
>>This seems like something internal to OpenSER, I have tried butchering
> 
> 
>>the config to force the ACK out first and it just created more
> 
> problems.
> 
>>Thanks!
>>
>>
>>	Matt S
>>
>>
>>
>>
>>
>>_______________________________________________
>>Users mailing list
>>Users at openser.org
>>http://openser.org/cgi-bin/mailman/listinfo/users
>>
>>  
> 
> 
> 
> 
> 
> _______________________________________________
> Users mailing list
> Users at openser.org
> http://openser.org/cgi-bin/mailman/listinfo/users
> 
> 