Re: [xwiki-devs] Does JDK6u14 "Garbage-First garbage collector (G1)" work and/or improve Xwiki performance/size/mem-locality?

NielsMayer | 19.02.2009 at 03:21 PM From <http://nielsmayer.com/xwiki/bin/view/fedora10/64bitJavaAppletIn64BitFirefox#Comments> http://nielsmayer.com/xwiki/bin/view/fedora10/64bitJavaAppletIn64BitFirefox… <http://nielsmayer.com/xwiki/bin/view/fedora10/64bitJavaAppletIn64BitFirefox#Comments> For info on 6u14's new "Garbage First" GC and potential usefulness for use with java web apps, see: <http://n2.nabble.com/Does-JDK6u14-%22Garbage-First-garbage-collector-%28G1%29%22-work-and-or-improve-Xwiki-performance-size-mem-locality--tp2344358p2344358.html> http://n2.nabble.com/Does-JDK6u14-%22Garbage-First-garbage-collector-%28G1%… <http://n2.nabble.com/forum/PrintPost.jtp?post=2344358> http://n2.nabble.com/forum/PrintPost.jtp?post=2344358 (cancel the print dialog: this gives you the complete message w/o extra crap). Unfortunately, when running with VM Arguments: jvm_args: -XX:+UnlockExperimentalVMOptions -XX:+UseG1GC -Xms256m -Xmx768m -XX:PermSize=256m -XX:MaxPermSize=768m -Djava.awt.headless=true … java_command: org.apache.catalina.startup.Bootstrap start You get a SIGSEGV (which obviates the whole point of garbage-collection): A fatal error has been detected by the Java Runtime Environment: SIGSEGV (0xb) at pc=0x0000000000e9c621, pid=13441, tid=114968912 JRE version: 6.0_14-b01 Java VM: OpenJDK 64-Bit Server VM (14.0-b10 mixed mode linux-amd64 ) Problematic frame: V [libjvm.so+0x296621] The SIGSEGV happens because of new Garbage Collector, as invoked by "-XX:+UnlockExperimentalVMOptions -XX:+UseG1GC". It dies after running for less than an hour: Current thread (0x000000004149f000): ConcurrentGCThread [stack: 0x0000000000000000,0x0000000000000000] [id=13448] siginfo:si_signo=SIGSEGV: si_errno=0, si_code=1 (SEGV_MAPERR), si_addr=0x0000001600000050 … elapsed time: 2916 seconds This issue doesn't occur with the default GC invoked with jvm_args: -Xms256m -Xmx768m -XX:PermSize=256m -XX:MaxPermSize=768m -Djava.awt.headless=true" I guess that's why this is an "Early Access" release.

PS: and will a different GC prevent sequences like this from happening?? see http://nielsmayer.com/happy-happy-joy-joy.txt