Strange 503 (service unavailable) error of "call controller 2"
Hello every experts,
I am doing some load testing of "call controller 2", and found a really strange problem.
I am using "mobicents-all-1_2_0_BETA1" and Sipp 3.1.
The problem is:
If I run "sipp -sf uac_pcap.xml -r 1 -rp 200" (Sipp will generate 1 call per 200 ms), after receiving about 2200 INVITE requests, mobicents server will start to return 503 (service unavailable) errors and not process most of the new coming INVITE requests.
That is, for a new starting mobicents server, it can correctly process about first 2200 calls without returning any errors. But after that, it rejects to provide services for most of new coming requests (it will correctly process some of them) and return 503 errors.
There are no exceptions showing up on the JBoss console and it can receive those INVITE requests with no problem. However, it just does not redirect them to the UAS.
I tried to stop the loading test at 2100, and then restart Sipp again. However, when Sipp reached about 100, the mobicents server will behave like I described above again.
The more surprising thing is, if I change the Sipp generator to "sipp -sf uac_pcap.xml -r 1 -rp 300" (Sipp will generate 1 call per 300 ms), the mobicents server will not start to return errors after the first 2200 calls, but after 4000 calls.
And then I tried to change the "call per ms" to 1/400ms, 1/500/ms, 1/1s, 1/3s. The mobicents will always go crazy at about 4000th calls.
Another interesting point is, to make the mobicents be able to serve request again, I can not just "undeploy-all" and then "deploy-all". I have to terminate JBoss, and restart the whole thing again!!!
I have been testing it for all day. So I am pretty sure it is not an incident case. Could anyone please explain to me if I did anything wrong, or this is a limitation of mobicents/call controller 2"?
Thank you very much~