[asterisk-users] Registrations, how many is too many?

[Yuan LIU]

>From: voiplist <gotvoip at gmail.com>
>Date: Wed, 28 Feb 2007 21:57:08 -0600
>
>That is interesting.. Not sure though how getting rid of IAX could
>have fixed your SIP issues, seems odd.
>
>We can't really get rid of IAX, our customers would flip their lids.

Maybe you can try to add another box separating SIP and IAX, just for 
problem isolation purposes.

Yuan Liu

>The big difference we have is that this has happened on more than one
>occasion when there was little to no call volume.
>
>On 2/28/07, Alejandro Kauffmann <akauffma at prodigy.net.mx> wrote:
>> > Anyone have any idea if there is some sort of limitation to
>> > the number of SIP or IAX end points which can register to an
>> > Asterisk system (2.8Ghz dual processor, 2GB ram) while also
>> > handling 30-50 simultaneous calls without getting into trouble?
>> >
>> > Of course the 30-50 simultaneous calls end up being 60-100
>> > channels of mostly G711 VoIP.
>> >
>> > We have seen issues where our Asterisk just gets all crazy
>> > and SIP quits working all together, to the point sometimes
>> > where we can't even fix it with a restart.
>> >
>> > At one point we were using all realtime for IAX and SIP
>> > clients, then we went to text files (more or less), still we
>> > are seeing this issue.
>> >
>> > When this happens we can't even do simple things with SIP
>> > like "sip show peers" etc. because Asterisk just says that
>> > the application doesn't exist.
>> >
>> > This has been a battle for a few months and we can't put our
>> > finger on it. Can't seem to figure out when it's going to
>> > happen either which is VERY tough on the nerves to say the least.
>> >
>> > This happens during peak times but also in the middle of the
>> > night when call volume is slow to non existent.
>> >
>> > The only thing that's constant during both peak and non peak
>> > times is the amount of registrations the system deals with.
>> >
>> > We have approx 1500-1800 end points registering to this
>> > particular system at any one time. This is a split between
>> > IAX and SIP not sure what the percentage of each is at the moment.
>> >
>> > It's been a long time since a problem has beat me/us and this
>> > one has won so far.
>> >
>> > Any help in getting my sanity back would be REALLY
>> > appreciated. _______________________________________________
>>
>>We've seen the same behavior since the 1.0.x version and have been unable 
>>to
>>track it down.  What I can tell you is that we used to peak around 40 
>>calls
>>(SIP/IAX to zap over E1 PRI all using alaw) and SIP would crash.  I saw a
>>peak of 97 calls today (216 channels with chanspy accounting for most of 
>>the
>>difference) and this is on a single core 3.2Ghz with 1Gb ram.  What 
>>changed?
>>We got rid of IAX.  I can't tell you if this will work for you or not, but
>>my nerves are doing better now that I don't get calls at random hours of 
>>the
>>day telling me that "we have no phones".  As a side note, inbound ZAP(PRI)
>>would still work, but only to IAX endpoints.