[Users] system busy - scenario

[samuel]

2006/5/16, unplug <maillisting at gmail.com>:
> I have experienced an error with using openser 1.0.1 + mediaproxy 1.4.2.
>
> The system works fine in the following loading.
> User login to the system: 20 hardware phones
> concurrent RTP session: 5
>
> However, when I increase number of user to 50 and concurrent RTP
> session to 30.  After a few minutes, the system is very unstable.  SIP
> phones display "wait logon" message as they can't logon to the system
> in some reasons and no one can make call even some users can logon.
> This situation will remain few minutes and it will resume normal.
> After a while, system will become unstable again.
>
This probably happens because all the listeners are busy (DNS lookup,
mysql queries, exec_*) and the incoming requests are not answered.
That's why your UA shows the can't logon because they get no answer.
This requests are "queued" and they are processed when the load
decreases, and that's why the system "reverts to normality".


> Anyone has such experience before?  I have grepped the log using ngrep
> but it is very difficult to troubleshoot as it is a log of whole
> system with 50 users.  In my simple analyst, I found that the most
> error replies happened in the log are listed below.

> 407 - proxy authentication required
This is returned everytime the route reaches a proxy_challenge
statement. It's used to obtain the user credentials in the digest
algorithm. Usually the first time a UA registers is challenged and the
challenge parameters are valid for a period of time. After this
period, the digest challenge has expired and a new challenge will be
issued.


> 408 - request timeout
This usually happens when openSER has relayed a request and does not
receive any response within fr_inv_timeout (this name may have changed
since the time I checked and now).

> 481 - call leg transaction does not exit
This error is returned by UAs when they receive a request within a
transaction which does not exists| has ended.
As an example:
when the ACK (after the INV-OK) is lost, the UA terminates the invite
transaction. When the other UA sends a BYE, since there is no
transaction, this BYE is answered with a "481 call leg transaction
does not exist".

> Do above errors will make the system unstable?  Which one is the most
> critical?  When and how the above errors will occur?
>
> To repeat the situation above, it involves a lot of resources.  Is
> there any sip phone simulator which can be used to simulate as a large
> number of user to create calls simultaneously?
>
Take a look at sipp for high load performance evaluation.

I guess there's some bottleneck either in your config file or in your
network that leads to this "network congestion" .

Hope it helps,

Samuel.


> Thanks!!
>
> _______________________________________________
> Users mailing list
> Users at openser.org
> http://openser.org/cgi-bin/mailman/listinfo/users
>