> I am now working on issue #1662.

GREAT!

> ...
> BTW: Benedikt, what did you use, if you did, to trace memory leaks?
>


1. I used "top" to watch the size of the process.
   (I think this doesn't help you to break the problem down)
2. I tried to work with mcheck. (At that time you was "involved" by giving a tip ...)
   I made it run to produce this mcheck output. I spent effort in analyzing the
   mcheck-outputfile. At the end it wasn't helpful. I forgot the reason.  
3. Here in my company I have access to rational purify on Windows.
   With the approach you described in bug#1664 you will end up with a small(!)
   C-program. Maybe I could analyze this program with purify.
   (BUT: The project where I worked with mapserver is over. So
    it is a problem for me to find time and I don't want to promise
    anything to be done "in time". But I'd *like* to do so ...)

Benedikt
   
> Regards,
> Umberto
>
> On 2/13/06, Umberto Nicoletti <[EMAIL PROTECTED]> wrote:
> > On 2/13/06, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
> > >
> > > Umberto
> > >
> > > Thank you for taking care onon this old but still open issue!
> > >
> > > I created
> > > http://mapserver.gis.umn.edu/bugs/show_bug.cgi?id=1661 and
> > > http://mapserver.gis.umn.edu/bugs/show_bug.cgi?id=1662
> > >
> > > Benedikt
> > >
> > > (Sorry! I was not able to add your e-mail-adress to cc. Bugzilla denied to
> > > do so. I suppose
> > > you can handle this anyway.)
> >
> > I just did. I will send you a patch for the
> > msConnPoolCloseUnreferenced function asap o that you can test it.
> >
> > Regards,
> > Umberto
> >
> > >
> > > Umberto Nicoletti <[EMAIL PROTECTED]> schrieb am 13.02.2006
> > > 10:25:02:
> > >
> > >
> > >  > Benedikt,
> > >  > since I am looking this issue would'n t you mind opening a bug for the
> > >  > inclusion of msConnPoolCloseUnreferenced in mapscript and one for the
> > >  > memory leaks (add me to the cc list)?
> > >  >
> > >  >
> > >  > Reagrds,
> > >  > Umberto
> > >  >
> > >  > On 1/2/06, Benedikt Rothe <[EMAIL PROTECTED]> wrote:
> > >  > >
> > >  > > Hi list members,
> > >  > >
> > >  > > From former threads I got the impression, that there are some folks
> > >  > > interested in the Oracle/Mapserver/Java/Tomcat.
> > >  > >
> > >  > > Therfore I'd like to share experiences I made with using
> > > Connection-Pooling
> > >  > > of
> > >  > > Oracle-Connections inside Java/Tomcat.
> > >  > >
> > >  > > Testenvironment: Mapserver 4.6.2; Suse-Linux; Tomcat 4.1.31; Sun-Java
> > > 1.4.2
> > >  > > Simulating 5 Browsers, which produce maps, query features, make
> > > selections,
> > >  > > query-legend-pics frequently.
> > >  > >
> > >  > > - After using synchronized "enough" I didn't have crashes of Tomcat.
> > >  > >
> > >  > > - Big memory leak: Between the first 5 requests and the next 100
> > > requests
> > >  > >    the Tomcat-process became about 400MB bigger.  (I use "top" for
> > >  > >    watching memory-footprint.)
> > >  > >
> > >  > > - Cleaning the Connection-Pool "by hand". This means:
> > >  > >   Opening the function msConnPoolCloseUnreferenced in mappool.c
> > >  > >   for use in Java and call it after every request.
> > >  > >
> > >  > > - After this I still have memory leaks:  About 100MB for 30.000
> > > requests.
> > >  > >   (I also made a test: 25.000 requests without Connection pooling.
> > > Memory
> > >  > > increased
> > >  > >     and decreased as expected  in this case.)
> > >  > >
> > >  > > - Performancecomparison in my testcase:
> > >  > >   Without use of connection-pooling: ~ 50 Request per minute
> > >  > >   With use of connection-pooling: ~75 Request per minute
> > >  > >
> > >  > >
> > >  > > As a result I have the following encouragements:
> > >  > > - Making msConnPoolCloseUnreferenced  availabe for mapscript via swig.
> > >  > >   (I made a hack by directly editing mapscript/java/mapscript_wrap.c
> > > and
> > >  > >    Java-Files in
> > > mapscript/java/edu/umn/gis/mapscript.) I
> > >  > > think this
> > >  > >   function could be part of the mapscript-Object?
> > >  > >
> > >  > > - Investigations on the memory leaks. Both leaks shouldn't occure.
> > > (I'll do
> > >  > >   this, if I find time. but ...)
> > >  > >
> > >  > > - Fernando Simon: What about using OCI-Connection-Pooling  for oracle
> > >  > > instead the mappool.c?
> > >  > >
> > >  > >
> > > http://oraclesvca2.oracle.com/docs/cd/B14117_01/appdev.101/b10779/oci09adv.htm#452244
> > >  > >   (If you don't have time, I maybe could help coding. But would it
> > > become
> > >  > > part uf Mapserver?)
> > >  > >
> > >  > > Happy new year to everybody
> > >  > > Benedikt Rothe
> > >
> >

Reply via email to