Re: [Wrapper-user] Dealing w/ JVM still accepting pings, but process hung

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

Jennifer,
As you have pointed out, the Wrapper is only able to monitor the health 
of the JVM
as a whole.  If a piece of code within the application stops working, 
there is currently
no way for the Wrapper to detect that.

I will give some more thought to how the Wrapper might be able to help 
with this
problem in the future.

For now however, rather than modifying the Wrapper.   I would suggest 
creating
a simple thread which loops and constantly compares the current time 
with a timeout
value set before and after calls to the code you are worried about.   If 
the thread
ever decides that it has timed out then it could simply make a call to
WrapperManager.restart();

Playing with the way the Wrapper and JVM communicate is not advisable.  Long
term, you will probably end up running into other problems.

Cheers,
Leif

Jennifer Kolar wrote:

> Actually- in case anyone is interested,I figured out my own soln...
>
> I added another member variable in WrapperManager  "m_ignorePings" and 
> a setter to set it to true or false (false by default).
> in WrapperManager in the case statement that is responding to input 
> from the native side, if the message is "ping" and the
> value of m_ignorePings is true, the WrapperManager simply doesn't send 
> a ping response back. I also exposed the ability for the child to tell 
> the wrappermanager to
> send a ping to the native side so that when I am in this ignorePings 
> state, my application can still let the native side know the VM is 
> alive on its own terms.
>
> then, in my child application- I can set the ignorePings state to true 
> whenever I am entering a method I think will potentially hang, and set 
> it to false when I leave it.
>
> No C or JNI changes needed.
> Jennifer
>
>
> Jennifer Kolar wrote on 2/18/2004, 11:34 AM:
>
>> I have an application that I am launching via the wrapper as a 
>> service from the simplewrapper app.
>> I have the problem that the JVM pinging in the wrapper is still 
>> happening successfully but the inner process in hung.. and since the 
>> wrapper thinks
>> the JVM is still alive, it allows the process to continue nicely.
>>
>> Leif, anyone, have any ideas on how to work around this.
>>
>> One thought I have had- but I don't see it exposed in any way in the 
>> JNI layer is to tell the wrapper to Stop and Start its internal ping. 
>> It is easy expose a ping method
>> on the wrapper that the child application can call.. and in that 
>> way---  if I think my child app is entering a dangerous method, I can 
>> have the wrapper stop its own pinging, have my
>> app send the ping itself, and then once I am out of dangerous 
>> territory.. restart the internal pinging in the wrapper. Any ideas on 
>> how to do this?
>>
>> thanks
>> Jennifer
>