Inline again On Mon, Jan 19, 2009 at 10:27 AM, Gang Liu <gangban.lau at gmail.com> wrote: > inline. > > In realworld, we need take care of buggy UAs. I have added many checking > source code to pjsip app, for exampe, my own timer to prevent a dialog can't > be released when remote side reply CANCLE with 200 OK but don't reply 487 > finial respond for INVITE. And some UA will send re-INVITE after they sent > BYE. Such actions maybe cause related dialog not released. > pjsip_endpt_dump() is also suggested by benny when I trace memory issue and > it help me fix many leak at real world. > Ok, good points, I'll keep this in mind and start using pjsip_endpt_dump(). However, in our situation we are lucky -- almost all the traffic is from Cisco AS5400s and the outbound legs are to proprietary bridge hardware, so we don't have to deal with unknown UAs at the moment. I can't get anywhere near 1000 cps (but the hardware is 3-4 years old). See >> above - when I do a normal kind of performance test, the memory usage is >> stable. It's only under real world load that I see the problem. >> > > my test linux box is HP DL380 with Dual Quard core CPU. > Ah, some old G3 DL360s here (in production we'll be using much stronger boxes) > pls take a look caching pool implementation. It will re-use old pools if > you set memory limit is high. > Yeah, I was pretty sure that's what the caching pool does. I'm going to have to go through the code carefully to find out why my silly pool caching algorithm seems to "hide" the problem. I've been running it live for a while now and the memory usage seems very stable, compared to last week when I experienced this crash the first time. Anyway I'll take your advice and try to get to the bottom of this. I'll get back to the list with my findings later in the week. Thanks again. -Mark -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.pjsip.org/pipermail/pjsip_lists.pjsip.org/attachments/20090119/4012c1f9/attachment.html>
