I wonder if anyone on this list is open to a thought exercise? I'm trying to put together a cogent message about this, but in the meantime:
I've got a situation where my openser box seems to route stuff to itself and it loops around and around until I get the classic "Too Many Hops" error.
This seems to happen most often on BYE, although I've seen an ACK do it too. Here is a small ngrep snippet, after
PHONE has called OPENSER-ONE who statelessly relays to OPENSER-TWO who t_relays to PSTNGW, and then the pstn user hangs-up:
U 23:03:53.254382 PSTNGW:5060 -> OPENSER-TWO:5060 BYE sip:JoeUser@OPENSER-ONE.example.com:5060;user=phone SIP/2.0 Call-ID: callid@PHONE From: sip:18005551212@OPENSER-TWO.example.com:5060;tag=26674 To: "Joe User" sip:JoeUser@OPENSER-ONE.example.com;tag=thetag Content-Length: 0 CSeq: 1 BYE Via: SIP/2.0/UDP PSTNGW:5060;branch=thebranch Route: sip:JoeUser@OPENSER-ONE:5060 Max-Forwards: 70
# U 23:03:53.255013 OPENSER-TWO:5060 -> OPENSER-ONE:5060 BYE sip:JoeUser@OPENSER-ONE:5060 SIP/2.0 Record-Route: sip:JoeUser@OPENSER-TWO;lr=on;ftag=26674 Call-ID: callid@PHONE From: sip:18005551212@OPENSER-TWO.example.com:5060;tag=26674 To: "Joe User" sip:JoeUser@OPENSER-ONE.example.com;tag=thetag Content-Length: 0 CSeq: 1 BYE Via: SIP/2.0/UDP OPENSER-TWO;branch=otherbranch.0 Via: SIP/2.0/UDP PSTNGW:5060;branch=thebranch Max-Forwards: 69 Route: sip:JoeUser@OPENSER-ONE.example.com:5060;user=phone
Right here OPENSER-ONE freaks out, and the delay causes OPENSER-TWO to re-transmit:
# U 23:03:53.611227 OPENSER-TWO:5060 -> OPENSER-ONE:5060 BYE sip:JoeUser@OPENSER-ONE:5060 SIP/2.0 Record-Route: sip:JoeUser@OPENSER-TWO;lr=on;ftag=26674 Call-ID: callid@PHONE From: sip:18005551212@OPENSER-TWO.example.com:5060;tag=26674 To: "Joe User" sip:JoeUser@OPENSER-ONE.example.com;tag=thetag Content-Length: 0 CSeq: 1 BYE Via: SIP/2.0/UDP OPENSER-TWO;branch=otherbranch.0 Via: SIP/2.0/UDP PSTNGW:5060;branch=thebranch Max-Forwards: 69 Route: sip:JoeUser@OPENSER-ONE.example.com:5060;user=phone
OK, so who on this list is good enough to be able to suggest plausible theories about what my openser.cfg flaw is? If you don't want to speculate openly on the list then please send me a direct email.
Thanks, -mark
Probably OPENSER-ONE does not recognize the BYE as targeted to itself (due to misconfiguration of listen|alias parameters) and that's why it keeps on sending to itself the message.. It's like OPENSER-ONE checks the Req-URI|Route domainname|IP and says..hey that's not for me, let's forward it to the next route..which happens to be itself....and this happens until Max-Forwards reaches the lower limit...
Try to add alias=OPENSER-ONE.example.com:5060
in your config file and see what happens..
Samuel.
2006/9/28, Mark Kent mark@noc.mainstreet.net:
I wonder if anyone on this list is open to a thought exercise? I'm trying to put together a cogent message about this, but in the meantime:
I've got a situation where my openser box seems to route stuff to itself and it loops around and around until I get the classic "Too Many Hops" error.
This seems to happen most often on BYE, although I've seen an ACK do it too. Here is a small ngrep snippet, after
PHONE has called OPENSER-ONE who statelessly relays to OPENSER-TWO who t_relays to PSTNGW, and then the pstn user hangs-up:
U 23:03:53.254382 PSTNGW:5060 -> OPENSER-TWO:5060 BYE sip:JoeUser@OPENSER-ONE.example.com:5060;user=phone SIP/2.0 Call-ID: callid@PHONE From: sip:18005551212@OPENSER-TWO.example.com:5060;tag=26674 To: "Joe User" sip:JoeUser@OPENSER-ONE.example.com;tag=thetag Content-Length: 0 CSeq: 1 BYE Via: SIP/2.0/UDP PSTNGW:5060;branch=thebranch Route: sip:JoeUser@OPENSER-ONE:5060 Max-Forwards: 70
# U 23:03:53.255013 OPENSER-TWO:5060 -> OPENSER-ONE:5060 BYE sip:JoeUser@OPENSER-ONE:5060 SIP/2.0 Record-Route: sip:JoeUser@OPENSER-TWO;lr=on;ftag=26674 Call-ID: callid@PHONE From: sip:18005551212@OPENSER-TWO.example.com:5060;tag=26674 To: "Joe User" sip:JoeUser@OPENSER-ONE.example.com;tag=thetag Content-Length: 0 CSeq: 1 BYE Via: SIP/2.0/UDP OPENSER-TWO;branch=otherbranch.0 Via: SIP/2.0/UDP PSTNGW:5060;branch=thebranch Max-Forwards: 69 Route: sip:JoeUser@OPENSER-ONE.example.com:5060;user=phone
Right here OPENSER-ONE freaks out, and the delay causes OPENSER-TWO to re-transmit:
# U 23:03:53.611227 OPENSER-TWO:5060 -> OPENSER-ONE:5060 BYE sip:JoeUser@OPENSER-ONE:5060 SIP/2.0 Record-Route: sip:JoeUser@OPENSER-TWO;lr=on;ftag=26674 Call-ID: callid@PHONE From: sip:18005551212@OPENSER-TWO.example.com:5060;tag=26674 To: "Joe User" sip:JoeUser@OPENSER-ONE.example.com;tag=thetag Content-Length: 0 CSeq: 1 BYE Via: SIP/2.0/UDP OPENSER-TWO;branch=otherbranch.0 Via: SIP/2.0/UDP PSTNGW:5060;branch=thebranch Max-Forwards: 69 Route: sip:JoeUser@OPENSER-ONE.example.com:5060;user=phone
OK, so who on this list is good enough to be able to suggest plausible theories about what my openser.cfg flaw is? If you don't want to speculate openly on the list then please send me a direct email.
Thanks, -mark
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
(due to misconfiguration of listen|alias parameters) and that's why it
That can't be it as the box functions normally in less complicated scenarios. Besides, I just added what you suggested and it doesn't make a difference :-)
And, what's the interaction between the "domain" module and how openser internally recognizes messages intended for itself?
And things work in simpler scenarios. Let me tell you what makes this more complicated:
When OPENSER-ONE gets a call that is meant for the pstn it sends back a redirect to the client, redirecting to OPENSER-TWO.
Some clients do start communicating with OPENSER-TWO, but some don't. They send the new R-URI, mentioning OPENSER-TWO, to OPENSER-ONE. I want to handle those messages as easily as possible, so right at the top of route[0] on OPENSER-ONE I have this:
# kludge, sender should have been redirected earlier, # but just in case...
if (uri=~"@OPENSER-TWO.example.com|@192.0.2.2") { log(1, "stateless relay TO pstn gw\n"); forward(); exit; }
if ((src_ip==192.0.2.2) && (!is_uri_host_local())) log(1, "stateless relay FROM pstn gw\n"); forward(); exit; }
The first "if" is for stuff addressed to OPENSER-TWO, but sent to OPENSER-ONE. The second one is for messages coming back the other way.
As I sit and think about this, I'm guessing that I can't get away with just shuffling the packets back and forth. But I could do with a explanation as to how silly this is.
Thanks, -mark
You should (must) use DNS entries for openser1 and openser2 and just use t_relay* calls to forward calls from one to another...
Samuel. 2006/9/29, Mark Kent mark@noc.mainstreet.net:
(due to misconfiguration of listen|alias parameters) and that's why it
That can't be it as the box functions normally in less complicated scenarios. Besides, I just added what you suggested and it doesn't make a difference :-)
And, what's the interaction between the "domain" module and how openser internally recognizes messages intended for itself?
It depends in your config file...if the Request passed the loose_route you have to process it in your config file accordingly...
And things work in simpler scenarios.
Then your "complex" config file is doing inapropriate things....
Let me tell you what makes this more complicated:
When OPENSER-ONE gets a call that is meant for the pstn it sends back a redirect to the client, redirecting to OPENSER-TWO.
Some clients do start communicating with OPENSER-TWO, but some don't. They send the new R-URI, mentioning OPENSER-TWO, to OPENSER-ONE. I want to handle those messages as easily as possible, so right at the top of route[0] on OPENSER-ONE I have this:
# kludge, sender should have been redirected earlier, # but just in case...
if (uri=~"@OPENSER-TWO.example.com|@192.0.2.2") { log(1, "stateless relay TO pstn gw\n"); forward(); exit; }
if ((src_ip==192.0.2.2) && (!is_uri_host_local())) log(1, "stateless relay FROM pstn gw\n"); forward(); exit; }
The first "if" is for stuff addressed to OPENSER-TWO, but sent to OPENSER-ONE. The second one is for messages coming back the other way.
As I sit and think about this, I'm guessing that I can't get away with just shuffling the packets back and forth. But I could do with a explanation as to how silly this is.
Thanks, -mark
Thanks very much for the detailed answers. You have supplied and confirmed lots of information for me.
I don't completely get what you mean by:
You should (must) use DNS entries for openser1 and openser2
Note that, for me, it's not DNS that's a mystery, but the exact sequence of words you used in this context.
The snippet of code used the name of openser2 as it appears in messages from openser2. It's pattern matching, not DNS. It also used the IP address of openser2 (src_ip), which also doesn't (or shouldn't) rely on dns.
I run the DNS, I run openser[12]. I figure I should be able to control things so that messages from openser2 look like I want them to look, and so should be able to do such comparisons.
Or, did you mean something else?
BTW, your comments, plus the nudge from Klaus, have convinced me to trash that code segment (the forward() kludge), so this is just a curiosity now.
Thanks, -mark
Hi Mark!
Does both gateways do record_route? If yes, then there should be 2 route headers in the BYE sent from the GW.
If server 2 does record_route, then the header Route: sip:JoeUser@OPENSER-TWO:5060 is missing.
If server 2 does not record_route. then the GW must not send to proxy 2.
Further, in the the BYE from GW to openser2, the proxy 1 is addressed 2 times (RURI and route header). proxy 1 should be referenced only 1 time, on the other hand there is no reference to the SIP client.
I guess the problem is how you handle the INVITE. Thus a trace of the initial INVITE would be more useful. I also suspect you to use fix_nated_contact in proxy 2 - this must not happen!
regards klaus
Mark Kent wrote:
I wonder if anyone on this list is open to a thought exercise? I'm trying to put together a cogent message about this, but in the meantime:
I've got a situation where my openser box seems to route stuff to itself and it loops around and around until I get the classic "Too Many Hops" error.
This seems to happen most often on BYE, although I've seen an ACK do it too. Here is a small ngrep snippet, after
PHONE has called OPENSER-ONE who statelessly relays to OPENSER-TWO who t_relays to PSTNGW, and then the pstn user hangs-up:
U 23:03:53.254382 PSTNGW:5060 -> OPENSER-TWO:5060 BYE sip:JoeUser@OPENSER-ONE.example.com:5060;user=phone SIP/2.0 Call-ID: callid@PHONE From: sip:18005551212@OPENSER-TWO.example.com:5060;tag=26674 To: "Joe User" sip:JoeUser@OPENSER-ONE.example.com;tag=thetag Content-Length: 0 CSeq: 1 BYE Via: SIP/2.0/UDP PSTNGW:5060;branch=thebranch Route: sip:JoeUser@OPENSER-ONE:5060 Max-Forwards: 70
# U 23:03:53.255013 OPENSER-TWO:5060 -> OPENSER-ONE:5060 BYE sip:JoeUser@OPENSER-ONE:5060 SIP/2.0 Record-Route: sip:JoeUser@OPENSER-TWO;lr=on;ftag=26674 Call-ID: callid@PHONE From: sip:18005551212@OPENSER-TWO.example.com:5060;tag=26674 To: "Joe User" sip:JoeUser@OPENSER-ONE.example.com;tag=thetag Content-Length: 0 CSeq: 1 BYE Via: SIP/2.0/UDP OPENSER-TWO;branch=otherbranch.0 Via: SIP/2.0/UDP PSTNGW:5060;branch=thebranch Max-Forwards: 69 Route: sip:JoeUser@OPENSER-ONE.example.com:5060;user=phone
Right here OPENSER-ONE freaks out, and the delay causes OPENSER-TWO to re-transmit:
# U 23:03:53.611227 OPENSER-TWO:5060 -> OPENSER-ONE:5060 BYE sip:JoeUser@OPENSER-ONE:5060 SIP/2.0 Record-Route: sip:JoeUser@OPENSER-TWO;lr=on;ftag=26674 Call-ID: callid@PHONE From: sip:18005551212@OPENSER-TWO.example.com:5060;tag=26674 To: "Joe User" sip:JoeUser@OPENSER-ONE.example.com;tag=thetag Content-Length: 0 CSeq: 1 BYE Via: SIP/2.0/UDP OPENSER-TWO;branch=otherbranch.0 Via: SIP/2.0/UDP PSTNGW:5060;branch=thebranch Max-Forwards: 69 Route: sip:JoeUser@OPENSER-ONE.example.com:5060;user=phone
OK, so who on this list is good enough to be able to suggest plausible theories about what my openser.cfg flaw is? If you don't want to speculate openly on the list then please send me a direct email.
Thanks, -mark
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
Klaus,
Your dissecting of the initial BYE, plus the help from samuel, tells me that I'm shooting myself in the foot with the forwarding I put in.
Also, you may be right about the double fix_nated_contact, which, if it's not happening now, would certainly have been happening once I remove the kludge that I included in an earlier email.
Thanks, -mark
-
Klaus Darilion -
Mark Kent -
samuel