Re: [Wrapper-user] JVM hang - Xms / Xmx

[Paul C. <cas...@au...>]

Yep - looks like pushing for a 1.4.x upgrade is the way to go.

Our app is a government GIS roads mapping tool for an Australian state
(that's as much as I can say).  On startup it loads all the data relating
to each layer of the map (eg freeways / local raods / railway lines etc
etc) into memory so that they are all there ready for 300+ users to turn
on/off whichever layer(s) they want - when they want.

I have a feeling that some of the problems with our app center around the
printing method used.  When the app was developed (around 5 years ago) Java
printing was slow and problematic (according to code comments) and so the
app generates a postscript or HPGL file using an OS installed print driver,
then Java calls the Windows print command to send the file to the user's
network printer.  Since setting the min/max RAM to different values I've
been able to get some more pieces to the puzzle when the app crashes.  A
couple of days ago someone ran a plot (print) and the system appeared to
get into an infinite loop.  The Windows CPU time for the Java process kept
ticking over (second for second) - although CPU for the process was only
6-7% (which indicates that it's a complicated loop - possibly between the
client and server - there was no network activity between the app server
and the database server which have a dedicated network card).  However, RAM
for the Java process steadily increased towards the maximum from about
1300MB when I first saw it to 1730MB before the client requested a restart.
I wanted to let it go till it crashed, but the client comes first.

The app was initially written to transfer files from the server to the
user's PC via FTP (yuck!), which I've scrapped and now just uses Tomcat
(which was in use for another component of the app anyway).

I'm thinking that the whole approach to printing needs to be overhauled,
and I've read somewhere that 1.4.x has a new print api.  Any ideas /
pointers / suggestions (from anyone)?

Has any reader had experience with network printing from Java?  Any
feedback would be greatly appreciated (as always).

Regards,

Paul Casanova



|---------+---------------------------------------->
|         |           Leif Mortenson               |
|         |           <le...@ta...>    |
|         |           Sent by:                     |
|         |           wra...@li...|
|         |           ceforge.net                  |
|         |                                        |
|         |                                        |
|         |           08/04/2004 12:09 AM          |
|         |           Please respond to            |
|         |           wrapper-user                 |
|---------+---------------------------------------->
  >--------------------------------------------------------------------------------------------------------------|
  |                                                                                                              |
  |       To:       wra...@li...                                                           |
  |       cc:                                                                                                    |
  |       Subject:  Re: [Wrapper-user] JVM hang - Xms / Xmx                                                      |
  >--------------------------------------------------------------------------------------------------------------|




Paul,

Paul Casanova wrote:

>I tried adding -Xrunhprof:depth=8 to the config file, but the JVM always
>suffers a runtime error r6017 ("unexpected multithread lock error") and
>restarts.  Any ideas?
>
>
Doing a Google search on that error message.  It looks like a C runtime
error.
In other words, Java is having a bad day.

Searching the Sun Bug database, the following issue came up:
http://developer.java.sun.com/developer/bugParade/bugs/4724252.html

That bug says it has been fixed in Java 1.4.1.  But the original post
says that it not
been a problem in 1.3.1_04.  That just means that their app had not been
able to
reproduce it however.

>Tried starting with initial heap size of 100MB and it took around the same
>time to start as normal, but when I added the -Xrun... it started doing
>full GC's when starting up the app, so it took longer.  Mind you, before I
>added -Xrun... it was doing around 10 GC's (not full) during startup with
>100MB initial!!
>
>Does -Xrunhprof usually have such overhead / impact?
>
>
Profiling takes a bunch of memory and also has a very noticeable effect of
performance.  Depending on the application of course, but it is usually
still
usable if you normally have enough CPU / Memory to spare.

Most likely your app is using a bunch more memory than normal and is thus
having problems hitting some upper memory limit.   Have you tried
increasing
the max memory setting?   If that is not possible, is it possible to
test with a
smaller data set?

Cheers,
Leif



-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
_______________________________________________
Wrapper-user mailing list
Wra...@li...
https://lists.sourceforge.net/lists/listinfo/wrapper-user