This Bugzilla instance is a read-only archive of historic NetBeans bug reports. To report a bug in NetBeans please follow the project's instructions for reporting issues.
I profiled small web application on Sun AppServer 8.1. I selected 'Analyze Performance' on Entire Application. Profiling slowed down appserver so much that appserver didn't started up correctly. It failed with error: Start server Sun Java System Application Server 8.1 timeout or interrupted. I repeated it without success. When I chose to profile part of the application it worked correctly. Also when I tried to profile web application on weblogic I got a message that instrumentation limit (64k?) was reached. I would recommend to automatically profile only deployed application and not internal appserver code (including startup).
Did you have the "Profile application server code" checkbox turned off? Also, please file a separate issue about the weblogic issue with reaching instrumentation limit, with as much details as possible.
Where is this checkbox? I can't find it.
Regarding the instrumatation limit on weblogic: I don't think it's a bug. Just after reaching of 64k instrumented method a dialog appeared saying that no more methods will be instrumented. And weblogic startup was also very slowed down like Sun AppServer. I think that it's all consequence of one issue that appserver and weblogic are too big for profiling.
Which Profiler build are you using? I most recent builds root methods are automatically determined and set for Entire Application profiling (if the "Profile application server code" checkbox is not checked). You can determine if root methods are set by looking at the NB logfile/console, there is "instrumentationRootMethods =" line containing list of root methods set for profiling - it is written when the profiling session starts. If you can't find the checkbox and the "instrumentationRootMethods =" line is empty, you are probably using M7 or some of older post-M7 builds. In this case please uninstall the Profiler and install the latest build to be able to profile your WebApp with minimal profiling overhead.
Just for your info: the "Profile application server code" checkbox should be next to the Entire Application option in Application Performance task if you are going to profile WebProject.
I uninstalled profiler and installed new version from \\phoenix\public\profiler\1.0\Build050805\installers\profiler-m8-dev-win.exe. BTW, installer still says it's milestone 7. I left checkbox "Profile application server code" turned off and then appserver started properly. When I tried to load a servlet from deployed web application a dialog "instrumenting, please wait" appeared. After some time error appeared in appserver and profiler was detached. [#|2005-08-05T17:02:52.159+0200|WARNING|sun-appserver-pe8.1_01|javax.enterprise. system.stream.err|_ThreadID=11;|Profiler Engine: exception when handling command from client:|#] [#|2005-08-05T17:02:52.179+0200|WARNING|sun-appserver-pe8.1_01|javax.enterprise. system.stream.err|_ThreadID=12;| Profiler Engine: exception when trying to send response or command to client:|#]
Please could you provide more info? Is live results window open during server startup? Is number of profiled classes displayed in Profiler Control Panel (Basic Telemetry) more than 0 when "instrumenting, please wait" dialog is active?
Sorry, by saying "number of profiled classes" I meant number of instrumented methods...
I've filed new issue 61878 - CPU Profiling of WebApp fails if live results are open. Could you please check if this matches the problem you've described?
This issue described problem which has already been fixed in Issue 53645 (Provide automatic/assisted selection of root methods for J2EE projects in the IDE. If the problem with detaching the Profiled during performing instrumentation doesn't match the issue suggested, please create separate issue for it. *** This issue has been marked as a duplicate of 53645 ***
If I didn't open live results window and tried to load a servlet a dialog "instrumenting, please wait" appeared. After while number of instrumented methods increased and stopped at 50970. Then 100% CPU was eaten, used memory grew, computer was unresponsive and after while java.lang.OutOfMemoryError was thrown from appserver. When I increased virtual memory from 500MB to 1.5GB error wasn't thrown, but profiler was detached anyway. If I opened live results window number of instrumented methods was 0 and then profiler was detached.
I don't know whether it's duplicate of issue #53645. I haven't found anything about detaching in issue #53645. And I forgot to say that I used Attach Profiler, not Profile Main Project.
Verification of old issues.
Closing old issues.