Re: Strange interaction between RPG and Java

[Scott Klement <klemscot@xxxxxxxxxxxx>]

On Wed, 14 Jan 2004 cb_ng@xxxxxxx wrote:

> I have an RPG program that is calling a java program that runs fine when
> run once.  Now, this RPG program is being called from a CL program that
> has basically an infinite loop with a couple seconds delay between
> iterations.  The first headache this caused was the job would slow down
> over a couples hours time.  It was as if some resources were not being
> released.  I have corrected this by adding a call that destroys the JVM
> each iteration.

The JVM cannot be restarted in V5R2 and later once it's destroyed, so I
wouldn't use this method if I were you.

There's obviously a bug in your program that's causing it to not release
resources.   I suggest fixing that bug rather than ending the JVM.


> If anyone has comments or experience with that I would appreciate them.
> Now, my problem and reason for the post is that after 6 - 8 hours of
> iterations I get a SocketException: too many open files from my java
> program.  My program opens/closes 2 sockets per iteration.  I have
> created a java wrapper program that does what the CL/RPG is doing with
> hardcoded parameters and after 24+ hours I have yet to see this error.
> BTW, I used RUNJVAPGM to run the java on the 400.  To me it would appear
> to be something in the interaction of RPG with Java!

There are some special considerations when calling Java w/RPG and
vice-versa.   Basically, the standard Java cleanup routines do not work
with objects created in a call from an RPG program.   Make sure that
you're specifically cleaning up any Java objects created in your RPG.

> .  Increasing the descriptors(ulimit) for the job just delays the error
> from occuring.

That's what happens when you have resources that aren't freed up.

> I have a workaround in place that submits limited iterations as jobs...
> but I would like to find an answer as to what/why this is happening.

Well, if you REALLY want to make sure things are cleaned up on each
iteration, you could do a PWRDWNSYS RESTART(*YES)... :)  (I'm kidding --
don't really do that!)