Just to add to this, when I'm developing locally in development mode I'm still seeing the memory leak. This, combined with my theory on the timed cache, makes me think that the GC issue and this development memory leak might be in or related to the objectFactory.
Doug -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Doug Hughes Sent: Tuesday, June 06, 2006 11:22 AM To: [email protected] Subject: RE: [Reactor for CF] Reactor Performance Part II Mike - A few points: "I think Doug made some significant code changes to get this result." - Yes, I rewrote the OO queries and removed the pooling code. It's faster and seems to have had a positive impact on memory usage. About the 7 instances: I know! I know! It's bad! I'm going to see if two of three of them can be consolidated. The 7th is the one that my blog is running in and is only there because if it's unstable nature. Once this problems is resolved this instance will be disabled. I'll get back to you on the consolidation. Also, let's say that I had a timedcache cfc which worked like a traditional collection, except that each time it's accessed it looks for elements in its instance data that have expired and it removes them from it's underlying structure. This CFC exists and these sorts of checks could be going on quite frequently (with each page request at least). I wonder if that could trigger the full GCs and if maybe that's the GC problem. Doug -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Mike Brunt - Webapper Services, LLC Sent: Tuesday, June 06, 2006 11:11 AM To: [email protected] Subject: [Reactor for CF] Reactor Performance Part II After Doug's trials and tribulations with the hard-drive failure and server move we are putting together the suspected memory-leak investigations again. Yesterday we re-instated verbose logging of Garbage Collection and also JRun Metrics logging. No other changes-tweaks have been applied as yet so we are running with the pre-change JVM settings that Doug had on his previous system (pre crash) before we did anything. Once again we are seeing many Full Garbage Collections, in a 23 hour period there were 2,125 which is one every 39.4 seconds which is far too frequent. I do however see a difference in overall memory consumption. This relates to far less memory being needed-used; for instance here is a snap-shot from the metrics logging showing the highest amount of memory in use that I could find in that 24 hour period: 06/05 20:54:50 metrics Web threads (busy/total/listen/idle/delay/handled): 0/7/1/6/0/128 Sessions:0 Total Memory=121984 Free=53363 Before we were seeing memory consumption regularly in the 400MB+ memory range. I think Doug made some significant code changes to get this result. However, I am seeing Free memory getting very low at times. We are seeing this behavior because we have no start memory setting for the JVM so when JRun starts it takes a minimal amount of memory, usually around 45MB and gets the rest as needed. We have found that this can cause more work than had we given JRun a good start memory size. However, there are 7 instances on this box and if we simply add the start size to the arguments in the jvm.config file then all instances will get that memory, so with a 512MB start we are already into 3.5GB of ram being consumed. So the next steps will be to control Full GC's which is always a good thing to do, install the SeeFusion utility and to up the start size of the JVM. As I mentioned above we are certainly seeing much better memory size needs and use. Kind Regards - Mike Brunt CTO and Lead Architect - Webapper Services LLC Tel 562.243.6255 Web Site: http://www.webapper.com Blog: http://www.webapper.net Tools: http://www.seefusion.com -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- Reactor for ColdFusion Mailing List [email protected] Archives at: http://www.mail-archive.com/reactor%40doughughes.net/ -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- Reactor for ColdFusion Mailing List [email protected] Archives at: http://www.mail-archive.com/reactor%40doughughes.net/ -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- Reactor for ColdFusion Mailing List [email protected] Archives at: http://www.mail-archive.com/reactor%40doughughes.net/ -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
