RE: [JK2] RedHat 8.0 JNI totaly bogus
-Original Message- From: Henri Gomez I didn't tried RH 8.0 yet, but the experience I've got from RH 7.2 with JNI make me some serious headaches. Before launching Apache 2.0 you need : 1) set the LD_LIBRARY_PATH export LD_LIBRARY_PATH=$jre/bin:$jre/bin/classic:$LD_LIBRARY_PATH Thanks man, that's the cache! For 1.4.1/RH8: export LD_LIBRARY_PATH=$JAVA_HOME/jre/lib/i386:$JAVA_HOME/jre/lib/i386/client:$ LD_LIBRARY_PATH 2) make JDK use the non floating stacks export LD_ASSUME_KERNEL=2.2.5 No need for that to start the JVM, but will check how it works. But important: 'which java' _must_ point to the jre/bin/java so jre's path has to precede jdk one, like $JAVA_HOME/jre/bin:$JAVA_HOME/bin MT. -- To unsubscribe, e-mail: mailto:tomcat-dev-unsubscribe;jakarta.apache.org For additional commands, e-mail: mailto:tomcat-dev-help;jakarta.apache.org
Re: [JK2] RedHat 8.0 JNI totaly bogus
Mladen Turk wrote: -Original Message- From: Henri Gomez I didn't tried RH 8.0 yet, but the experience I've got from RH 7.2 with JNI make me some serious headaches. Before launching Apache 2.0 you need : 1) set the LD_LIBRARY_PATH export LD_LIBRARY_PATH=$jre/bin:$jre/bin/classic:$LD_LIBRARY_PATH Thanks man, that's the cache! Did it works now ? For 1.4.1/RH8: export LD_LIBRARY_PATH=$JAVA_HOME/jre/lib/i386:$JAVA_HOME/jre/lib/i386/client:$ LD_LIBRARY_PATH 2) make JDK use the non floating stacks export LD_ASSUME_KERNEL=2.2.5 No need for that to start the JVM, but will check how it works. There is good information about floating stack in IBM SDK But important: 'which java' _must_ point to the jre/bin/java so jre's path has to precede jdk one, like $JAVA_HOME/jre/bin:$JAVA_HOME/bin Are you using IBM, Sun or Blackdown SDK ? -- To unsubscribe, e-mail: mailto:tomcat-dev-unsubscribe;jakarta.apache.org For additional commands, e-mail: mailto:tomcat-dev-help;jakarta.apache.org
RE: [JK2] RedHat 8.0 JNI totaly bogus
-Original Message- From: Henri Gomez Did it works now ? Yes. For 1.4.1/RH8: export LD_LIBRARY_PATH=$JAVA_HOME/jre/lib/i386:$JAVA_HOME/jre/lib/i386/client :$ LD_LIBRARY_PATH 2) make JDK use the non floating stacks export LD_ASSUME_KERNEL=2.2.5 No need for that to start the JVM, but will check how it works. There is good information about floating stack in IBM SDK Will check. But important: 'which java' _must_ point to the jre/bin/java so jre's path has to precede jdk one, like $JAVA_HOME/jre/bin:$JAVA_HOME/bin Are you using IBM, Sun or Blackdown SDK ? Sun's 1.4.1_01 Still, I found that JVM kills the process that started it when calling JNI_CreateJavaVM (that doesn't happens on WIN32, cause error is returned instead), if the jvm could not be loaded. Makes no sense at all. This forces Apache to restart the child process, calling create again, and all that is going in the infinite loop. Very bad :(. MT. -- To unsubscribe, e-mail: mailto:tomcat-dev-unsubscribe;jakarta.apache.org For additional commands, e-mail: mailto:tomcat-dev-help;jakarta.apache.org
[JK2] Use lock file forJNI startup status WAS: [JK2] RedHat 8.0 JNI totaly bogus
I've made the abort and exit hook for JVM adding two aditional option parameters at the end of config ones. If the JVM isn't set properly (wrong LD_LIBRARY_PATH, PATH pointing to the java outside the jre, etc...) the CreateJavaVM calls the abort hook. I was assuming that this hook will act instead the real abort function, but seems that it is only atexit call, cause the real abort function is called after the abort hook. So the only way is IMO to use the some sort of a lock file that will be created when the abort hook is called from JVM initialization, thus preventing further CreateJavaVM invocations. Perhaps we can use the shm for that from parent process, but the env is created on child_init. Thoughts? MT. -- To unsubscribe, e-mail: mailto:tomcat-dev-unsubscribe;jakarta.apache.org For additional commands, e-mail: mailto:tomcat-dev-help;jakarta.apache.org
[JK2] RedHat 8.0 JNI totaly bogus
Hi, Using JNI on RH8 using either original 2.0.40, Henri's 2.0.43-1.7.2 or build from sources (either prefork or worker). Even tried 2.0.1 rpm, but on all config when trying to use the [vm:] I'm having constant child_init, with cpu at 99%. ErrorLog is growing and seems that entire thing faults at vm_init (vm_default at 445 is the last thing logged). Nothing is passed after the loading of jvm, like the child process exits, but pool cleanup doesn't report that. After stopping httpd, a few childs are reported as 'still did not exit'. I've put some extra ap_logs and the child pids are growing even when set StartServers to 1. The Java is 1.4.0_01 Any clues? MT. -- To unsubscribe, e-mail: mailto:tomcat-dev-unsubscribe;jakarta.apache.org For additional commands, e-mail: mailto:tomcat-dev-help;jakarta.apache.org
Re: [JK2] RedHat 8.0 JNI totaly bogus
Mladen Turk wrote: Hi, Using JNI on RH8 using either original 2.0.40, Henri's 2.0.43-1.7.2 or build from sources (either prefork or worker). Even tried 2.0.1 rpm, but on all config when trying to use the [vm:] I'm having constant child_init, with cpu at 99%. ErrorLog is growing and seems that entire thing faults at vm_init (vm_default at 445 is the last thing logged). Nothing is passed after the loading of jvm, like the child process exits, but pool cleanup doesn't report that. After stopping httpd, a few childs are reported as 'still did not exit'. I've put some extra ap_logs and the child pids are growing even when set StartServers to 1. The Java is 1.4.0_01 I think they just released 1.4.1 - and I remember reading in the release notes about a jni fix that seems related. I had few other problems with 1.4.0 and jni - for example I wasn't able to load the libjk.so ( java loading the C lib, not the reverse ). It did work after I removed the 'endorsed' libs flags - I have no ideea why. Right now 1.3 is much more stable for JNI stuff ( IMO ) - I hope 1.4.1 will solve those problems. -- Costin -- To unsubscribe, e-mail: mailto:tomcat-dev-unsubscribe;jakarta.apache.org For additional commands, e-mail: mailto:tomcat-dev-help;jakarta.apache.org
RE: [JK2] RedHat 8.0 JNI totaly bogus
From: Costin Manolache Using JNI on RH8 using either original 2.0.40, Henri's 2.0.43-1.7.2 or build from sources (either prefork or worker). Even tried 2.0.1 rpm, but on all config when trying to use the [vm:] I'm having constant child_init, with cpu at 99%. ErrorLog is growing and seems that entire thing faults at vm_init (vm_default at 445 is the last thing logged). Nothing is passed after the loading of jvm, like the child process exits, but pool cleanup doesn't report that. After stopping httpd, a few childs are reported as 'still did not exit'. I've put some extra ap_logs and the child pids are growing even when set StartServers to 1. The Java is 1.4.0_01 I think they just released 1.4.1 - and I remember reading in the release notes about a jni fix that seems related. I had few other problems with 1.4.0 and jni - for example I wasn't able to load the libjk.so ( java loading the C lib, not the reverse ). It did work after I removed the 'endorsed' libs flags - I have no ideea why. Right now 1.3 is much more stable for JNI stuff ( IMO ) - I hope 1.4.1 will solve those problems. Will try both. MT. -- To unsubscribe, e-mail: mailto:tomcat-dev-unsubscribe;jakarta.apache.org For additional commands, e-mail: mailto:tomcat-dev-help;jakarta.apache.org
RE: [JK2] RedHat 8.0 JNI totaly bogus
*PLEASE* Post your results!!! -- mailto:Paul.Brzezinski;EDS.com Enterprise Distributed Capabilities EDS Corporation 248-265-8283 : -Original Message- : From: Mladen Turk [mailto:mturk;mappingsoft.com] : Sent: Wednesday, October 23, 2002 2:01 PM : To: 'Tomcat Developers List' : Subject: RE: [JK2] RedHat 8.0 JNI totaly bogus : : : : : From: Costin Manolache : : Using JNI on RH8 using either original 2.0.40, Henri's : 2.0.43-1.7.2 or : build from sources (either prefork or worker). Even tried : 2.0.1 rpm, : but on all config when trying to use the [vm:] I'm having constant : child_init, with cpu at 99%. ErrorLog is growing and seems : that entire : thing faults at vm_init (vm_default at 445 is the last : thing logged). : Nothing is passed after the loading of jvm, like the : child process : exits, but pool cleanup doesn't report that. After : stopping httpd, a : few childs are reported as 'still : did not exit'. : : I've put some extra ap_logs and the child pids are growing : even when : set StartServers to 1. The Java is 1.4.0_01 : : I think they just released 1.4.1 - and I remember reading in : the release notes about a jni fix that seems related. : : I had few other problems with 1.4.0 and jni - for example I : wasn't able to load the libjk.so ( java loading the C lib, : not the reverse ). It did work after I removed the 'endorsed' : libs flags - I have no ideea why. : : Right now 1.3 is much more stable for JNI stuff ( IMO ) - I : hope 1.4.1 will solve those problems. : : : Will try both. : : MT. : : : : -- : To unsubscribe, e-mail: : mailto:tomcat-dev-: [EMAIL PROTECTED] : For : additional commands, : e-mail: mailto:tomcat-dev-help;jakarta.apache.org : -- To unsubscribe, e-mail: mailto:tomcat-dev-unsubscribe;jakarta.apache.org For additional commands, e-mail: mailto:tomcat-dev-help;jakarta.apache.org
