Yeah sure, see below but I do stress though that each Java application - and hence each CF application - is different so these settings may need to be tweaked to cater for your situation. Unfortunately unless you have an identical test or UAT environment (load etc) to test on then you will have to trial and error settings.
We run 3 load balanced web servers each with this config: Dual Xeon 3.6Ghz CPU's 4GB RAM 6 x 36GB Hard Drives Java 1.4.2_05 By running JVMStat (http://java.sun.com/performance/jvmstat/) we managed to get a better understanding on what was going on under the hood of the apps in as far as Java is concerned. We were seeing things like locked out servers, high CF queues etc and an optimal server config seems to be : CF Related ======= Maximum number of simultaneous requests: 12 Timeout Requests after (seconds): 120 Java Related ======== Minimum JVM Heap Size (MB): 1024 Maximum JVM Heap Size (MB): 1024 MaxPermSize=128m PermSize=64 UseConcMarkSweepGC (to utilise dual processors GC) NewSize=48m We are on IIS and when CF crashes we do indeed see IIS crash so we need to review this - which we have not done as of yet, but the Java settings seems to have certainly stabilised Java/CFMX. When I get a chance, I am going to write a technote for MM.com if they will have it on the subject... Neil -----Original Message----- From: gabriel l smallman [mailto:[EMAIL PROTECTED] Sent: 23 November 2004 15:18 To: CF-Talk Subject: RE: JVM heap creep Neil, Could you share your jvm settings? I would be interested to see what you have as compared to the "best practices" I have read. Thanks gabe -----Original Message----- From: Robertson-Ravo, Neil (RX) [mailto:[EMAIL PROTECTED] Sent: Tuesday, November 23, 2004 9:50 AM To: CF-Talk Subject: RE: JVM heap creep Micha, I agree wholeheartedly on this front, as I think that with the step to CFMX as an application on top of Java, MM has not helped or provided the necessary tools to help debug applications - now this could be down to the fact that Java is not an MM product and you are left to your own devices to solve performance issues. We have recently had performance issues which could have been caused by a numerous amount of small things - we upgraded our hardware and also had to run extensive debug sessions using JVMStat as we found that with the old hardware and settings that Garbage Collection was taking over every 5 mins and indeed was leading to a "stop the world" situation a few tweaks to PermSize etc and we were god to go. We are still getting the odd large spike in activity, especially when we get around 1500 users at any once time. There is a definite problem with IIS in particular I think - Michael reckons it may be down to IISLockdown but I am not not sure as of yet. N -----Original Message----- From: Micha Schopman [mailto:[EMAIL PROTECTED] Sent: 23 November 2004 14:26 To: CF-Talk Subject: RE: JVM heap creep It just depends on logging + monitoring, adjusting, logging + monitoring, adjusting and do this about 10x and keep your fingers crossed. A very bad approach for running enterprise apps, but there is no other way. My personal opinion about this; these stability issues (and there are an enormous amounts of people with these issues) should be the no. 1 top priority of macromedia. They cause a lot of distrust in the product. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~| Special thanks to the CF Community Suite Gold Sponsor - CFHosting.net http://www.cfhosting.net Message: http://www.houseoffusion.com/lists.cfm/link=i:4:185169 Archives: http://www.houseoffusion.com/cf_lists/threads.cfm/4 Subscription: http://www.houseoffusion.com/lists.cfm/link=s:4 Unsubscribe: http://www.houseoffusion.com/cf_lists/unsubscribe.cfm?user=89.70.4 Donations & Support: http://www.houseoffusion.com/tiny.cfm/54
