James, Old registrations should not be the problem - your problem (as I understand it) is missing registration, not too many :)...
BTW, for the failed calls, do you get a 404 not found from scrip or a 408 timeout ? If you consider it, I can send you a script with an extension of usrloc to log when a new AOR is added or when a whole AOR is removed, so you can use it to doublecheck if your registrations are continuous in time . Regards, Bogdan-Andrei Iancu OpenSIPS Founder and Developer http://www.opensips-solutions.com On 05/16/2013 09:36 PM, James Tranovich wrote: > Hello Bogdan, > > Thanks for your reply! We are using opensips 1.8.0-notls (x86_64/linux). > > We do not think this issue is load related but perhaps older > registrations have not yet expired. We will try setting the > min_expires parameter to a low number to test this hypothesis; any > other approaches we could try? > > Thanks once again! > > James > > > On Thu, May 16, 2013 at 1:53 AM, Bogdan-Andrei Iancu > <[email protected] <mailto:[email protected]>> wrote: > > Hello James, > > No, there is no such known bug or issue. What I suspect is that > there are short intervals (milisecs) where an AOR is not > registered - this may happen because : > - the test tool is not performing properly under high load and > fails to do re-register before old registration expires. > - OpenSIPS is overloaded (too few processes ?) and it is not > able to process traffic in realtime (check the LOAD related > statistics). > > What OpenSIPs versions are you using ? > > Regards, > > Bogdan-Andrei Iancu > OpenSIPS Founder and Developer > http://www.opensips-solutions.com > > > On 05/15/2013 01:56 AM, James Tranovich wrote: >> Hello all -- >> >> First, we love opensips :) >> >> Lately, we have been running into a strange issue which seems to >> be related to handling a ton of REGISTER messages. Basically, we >> have a test script that tries to simulate about 50 to 100 >> simultaneous calls; they all register en masse and then randomly >> start placing calls to another test number (after a random time >> interval). Every once in a while, though, an INVITE won't go >> through because opensips apparently can't find that phone number. >> Oddly enough, if we do an "opensipsctl online" immediately >> before/after, that command shows that, in fact, the recipient's >> number is present and presumably already registered. SIP >> logs/ngrep tracing confirm this. >> >> I was wondering if this is a known bug. This behavior only >> happens when registering a certain number of calls at once; if we >> test with a low number of calls (10, say), this behavior does not >> happen. It may be that we are spamming opensips with too many >> REGISTER messages (authentication is required, so two REGISTER >> messages are sent, the first w/o auth, the second with auth). But >> I don't see why opensips should have problems with this. >> >> Any thoughts on this? Is this a known issue already? (Searching >> for this issue didn't yield much). Thanks! >> >> James >> >> >> _______________________________________________ >> Users mailing list >> [email protected] <mailto:[email protected]> >> http://lists.opensips.org/cgi-bin/mailman/listinfo/users > >
_______________________________________________ Users mailing list [email protected] http://lists.opensips.org/cgi-bin/mailman/listinfo/users
