On 8/24/07, <b class="gmail_sendername">Joshua Colp</b> <<a href="mailto:jcolp@digium.com">jcolp@digium.com</a>> wrote:<div><span class="gmail_quote"></span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
I'm going to end this email with a question myself... how many people<br>have Asterisk on a development/staging server before deployment, test,<br>and isolate the issues they may have in their specific scenario?</blockquote>
<div><br>I do, but many of the problems I have experienced (see #10199 for an example) don't manifest under anything but production loads. In that particular case, I couldn't find a way to replicate the levels of traffic and the nuances of agent pickup / ignore / hangup / etc. in my lab. My current load test consists of a lab box generating about 50-75 concurrent calls to an ITSP that terminate on another * conencted to PRI. But what you do with a call when it hits your box can make a difference. I had a load test that just walked through my IVRs pressing random keys for about 5 minutes. I could load 4 PRI full of calls to that context and the box would be fine. The second I added queueing (so that there was SIP signalling out to agent softphones), I'd get a kernel panic. The agent didn't even have to pick up the phone - just making it ring was enough.
<br><br>Let me ask a question myself: what kind of regression test does * undergo before release, and what level of traffic gets put through stuff like app_queue? I assume it's not real-world scale, else these hard to pin down concurrency issues we're seeing would have been caught in test.
<br></div></div><br>-- <br>j.