Recommended Data Collection for Crashes and Hangs in Java (Windows)
VM crashes
- collect Hotspot Error Logfile. It is written to the current working directory. It's name is e.g. hs_err_pid<pid>.log (where <pid> is the process id of process).
- To decode the hs_err log, please use http://blogs.sun.com/dave/resource/hs_err6.pl or use the web interface - http://openjdk.neojava.org/cgi-bin/hs_err.
( Please note: We cannot guarantee the availability or the support of these tools as they are not Sun products ) - Get the windows logfile (Dr. Watson log or ADplus log) and a dump file (Dr. Watson user.dmp or an ADplus dump).
- Always use full crash dump type for Dr. Watson. See the following sample screenshot:

For details about Dr. Watson, please refer to Description of the Dr. Watson for Windows (Drwtsn32.exe) Tool - http://support.microsoft.com/kb/308538
Please also refer to our troubleshooting document 216912 on sunsolve for details - http://sunsolve.sun.com/search/document.do?assetkey=1-37-216912
- alternatively, you can use ADplus. Please refer to How to use ADPlus to troubleshoot "hangs" and "crashes" - http://support.microsoft.com/default.aspx?scid=kb;en-us;286350
In particular if you run the JVM as a "service", which is similar to a daemon, then ADplus is the only way. Choose these options:
adplus -crash -o <outputdir> -p <pid> -quiet -NoDumpOnFirst
| Note: The option -NoDumpOnFirst is documented (run adplus -?). If the option is omitted, ADplus will create a minidump every minute, which will quickly fill up diskspace. |
- Windows crash dumps for Java Processes - http://blogs.sun.com/poonam/date/20081019
- More on Windows Crash Dumps for Java Processes - http://blogs.sun.com/poonam/date/20081227