Re: [Wrapper-user] Hung JVM

[Leif M. <le...@ta...>]

Paul,
    Is this something that you have been able to reproduce?  If so could 
you make me
a log file with the wrapper.debug=true property set?  I would need that 
to be able to
tell you the cause of the timeout.  I would like the wrapper.conf file 
at the same time.
It you are concerned with posting info to the list, then send the log 
file to me directly
with an abbreviated message to the list.   I prefer to keep most emails 
on list so that
others can use them in the future.  But for large log files etc, It is 
usually better to
send them to be directly.

    You mentioned that your application is quite large and that it 
pauses for long
periods of time due to GC?   If so, that is most likely your problem.  
If a single
GC sweep is taking longer than 30 seconds then the default 
wrapper.ping.timeout
is expiring and the Wrapper will kill the JVM.  It may be necessary for 
you to
lengthen this timeout or use a different GC method.
    If your JVM is using more memory than is available without using 
disk swapping
then the GC sweep can easily take an order of magnitude longer than it 
would if there
was enough real memory.   I highly advise you to never try to run a Java 
application
it in a situation where it is being paged to swap memory.  Performance 
in general
is dreadful.

    If it isn't GC related then I will need to see the debug log file.

Cheers,
Leif

Paul Casanova wrote:

>Hi all,
>
>I have the wrapper set to timeout the JVM at 10 minutes as the application
>is quite huge, and sometimes idle as it garbage collects.
>
>The wrapper determined that the JVM had hung as follows:
>
>ERROR  | wrapper  | 2003/11/11 14:41:01 | JVM appears hung: Timed out
>waiting for signal from JVM.
>STATUS | wrapper  | 2003/11/11 14:41:01 | Dumping JVM state.
>ERROR  | wrapper  | 2003/11/11 14:41:01 | Unable to send BREAK event to JVM
>process.  Err(6 : The handle is invalid. (0x6))
>ERROR  | wrapper  | 2003/11/11 14:41:02 | Java Virtual Machine did not exit
>on request, terminated
>STATUS | wrapper  | 2003/11/11 14:41:08 | Launching a JVM...
>INFO   | jvm 2    | 2003/11/11 14:41:08 | Wrapper (Version 3.0.3)
>INFO   | jvm 2    | 2003/11/11 14:41:08 |
>
>I have to be careful about including preceding log entries due to privacy
>and security, but basically there where just RMI reflection exceptions when
>the server prompted the client to update a job status window, but this had
>been happening all day, so I doubt it's the cause of the hang.
>  
>