Re: [Wrapper-user] Wrapper access violation

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

Alex,
    Yup, this is a known bug that has been fixed in 3.1.1.   I had been 
hoping to get a point
release out fairly quickly but have run into some problems on FreeBSD.   
Thanks to the
help of a couple users, I think we just about have that solved.  So a 
release should be coming
along soon.

    In your direct email, you mentioned a link to a development build.  
I do not usually keep
such builds up for long as I can not support them.   It sounds like you 
are stuck so I put up
the latest snapshot build for you to use.  Please be sure to upgrade to 
the release as soon
as it is out.  I am using this build without any problems, but it should 
not be considered
well tested.  Do let me know about any problems you find and be sure to 
read over the
change log.

http://wrapper.tanukisoftware.org/tmp/3.1.1e/wrapper_win32_3.1.1e.zip

    You also asked why the Wrapper did not restart when it encountered 
this crash.  The
fact of the matter is that this was a very fatal crash that the Wrapper 
was not able to recover
from.  It is not possible for the Wrapper to restart itself, it is only 
capable of restarting the
JVM.   Normally the Wrapper does not crash.  The tick timer was labeled 
as experimental
for exactly this reason until it could be well tested.   It seems to be 
quite stable in 3.1.1 now.
But it will not become the default timer until it has undergone much 
more testing by the
user base.

Alex Kritikos wrote:

> Hi,
>
>  
>
> We are using wrapper 3.1.0 with the tick based timer. In one of our 
> production servers we get random crashes with the following in the log:
>
>  
>
> DEBUG  | wrapperp | 2004/07/09 11:50:31 | send a packet PING : ping
>
> INFO   | jvm 1    | 2004/07/09 11:50:31 | Received a packet PING : ping
>
> INFO   | jvm 1    | 2004/07/09 11:50:31 | Send a packet PING : ok
>
> DEBUG  | wrapperp | 2004/07/09 11:50:31 | read a packet PING : ok
>
> DEBUG  | wrapper  | 2004/07/09 11:50:31 | Got ping response from JVM
>
> INFO   | jvm 1    | 2004/07/09 11:49:11 | The system clock fell behind 
> the timer by 83700ms.
>
> FATAL  | wrapper  | 2004/07/09 11:49:11 | encountered a fatal error in 
> Wrapper
>
> FATAL  | wrapper  | 2004/07/09 11:49:11 |   exceptionCode    = 
> EXCEPTION_ACCESS_VIOLATION
>
> FATAL  | wrapper  | 2004/07/09 11:49:11 |   exceptionFlag    = 
> EXCEPTION_NONCONTINUABLE_EXCEPTION
>
> FATAL  | wrapper  | 2004/07/09 11:49:11 |   exceptionAddress = 0040E42A
>
> FATAL  | wrapper  | 2004/07/09 11:49:11 |   Write access exception to 
> 00000000
>
> FATAL  | wrapper  | 2004/07/09 11:49:11 | Fatal error in the Timer thread.
>
>  
>
> This is a win2k box using JDK 1.3.1
>
>  
>
> Attached is our conf file
>
>  
>
> I have read in the bugs section about other people experiencing this 
> and you provided a dev version for testing. Unfortunately the links 
> are broken now, so I was wondering If this has been fixed and if so on 
> what release etc. Also is there a tactical solution to stop the crashes?
>
> We can not stop using the tick timer as we always get the restart 
> problem due to the system clock falling behind quite a lot. The 
> machine has an ntp client synchronizing every 10 mins.
>
>  
>
> Thanks,
>
>  
>
> Alex Kritikos
>