Re: [Wrapper-user] "JVM appears hung": Wrapper restarts webserver every week

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

Andreas,
One think that you have to be very careful of is to make sure that the
JVM is never forced to swap any of its memory to disk.  This is
nothing related to the Wrapper.  But Java gets VERY VERY VERY slow
when its memory is being swapped.   It is common for 1/1000 affect on
performance.   It may be possible that your new database is using a
lot more memory and thus forcing the JVM to swap.

Take a look at your task manager and make sure you have enough memory.
 I will see if there is any way that I can tell when this is happening
to Java and warn the user about it in the log.

Cheers,
Leif

On Fri, Apr 17, 2009 at 5:02 PM, Andreas Karlsson
<and...@ya...> wrote:
>
> Hi Leif,
> I've restarted it with the setting you suggested.
> The problem shows up at any time between 7-10 days from last restart.
> Often the application complains about low memory and performs garbage collection
> prior to that.
> I've been fiddling with the memory settings and now they should be the same for the
> server-jvm-wrapper but the error still occurs.
> I'll get back to you next week when it has restarted again.
> brgds
> Andreas
>
> --- Leif Mortenson <le...@ta...> wrote:
>
>> Andreas,
>> Most likely the Java process is being starved of CPU and that is
>> leading to the Wrapper thinking it is frozen.   If that is the case
>> then you can easily resolve the problem by setting the following
>> property:
>> wrapper.ping.timeout=120
>> (http://wrapper.tanukisoftware.org/doc/english/prop-ping-timeout.html)
>>
>> That will tell the Wrapper to allow the JVM to be frozen for 120
>> seconds before it is killed.  The default is 30 seconds.
>>
>> However, your web site would be unresponsive during that period still.
>>  Most likely it is now for at least 30 seconds.  Do the restarts
>> happen at any particular time each week?  Is there some large DB
>> processing being done that is consuming all of the CPU?  It may be
>> necessary to move the Database over to another physical server to make
>> sure your app server is not being affected by that processing.
>>
>> If you set the wrapper.debug=TRUE property and then reproduce this, I
>> might be able to tell you a little more information about what is
>> happening prior to the restart.  I would only need the 3-5 minutes of
>> time prior to the Wrapper timing out and killing the JVM.
>>
>> There have also been several bugs fixed since 3.0.5 which could affect
>> this.  Please review the release-notes:
>> http://wrapper.tanukisoftware.org/doc/english/release-notes.html
>>
>> Cheers,
>> Leif
>>
>> On Thu, Apr 16, 2009 at 5:55 PM, Andreas Karlsson
>> <and...@ya...> wrote:
>> >
>> > Hi wrapper-folks!
>> > I would love some hints and ideas on how to debug this problem and find a
>> solution.
>> >
>> > The system is a
>> > - windows server 2003
>> > - jettyserver
>> > - java 1.4.1
>> > - wrapper 3.0.5
>> >
>> > It has been running for years without any problems. But after some updates of
>> the
>> > data in the database where the web-content is stored and a recompile of the code
>> it
>> > started to produce these messages (pasted below). It happens approximately once
>> > every week.
>> > How can I debug this?
>> > Is there a way to trigg a dump when it happens?
>> >
>> >
>> > ERROR  | wrapper  | 2009/04/15 10:17:07 | JVM appears hung: Timed out waiting
>> for
>> > signal from JVM.
>> > ERROR  | wrapper  | 2009/04/15 10:17:07 | Java Virtual Machine did not exit on
>> > request, terminated
>> > STATUS | wrapper  | 2009/04/15 10:17:13 | Launching a JVM...
>> > INFO   | jvm 2    | 2009/04/15 10:17:14 | Wrapper (Version 3.0.5)
>> >
>> > brgds
>> > Andreas