[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20050510: computer freezes



Gabe,

If you run cleanup while a script is running, you will kill the gplt
process and orphan the gf process. When you remove the message
queue (the script does this with the "ipcrm -q" command), you no
longer have a connection to the gf process. 

The script should be run when your scripts are not running.

Another bit of advice is to run each script in its own temporary
directory (such as shown in the csh examples using $$ to obtain
a unique process id for a working directory name) so you don't have
multiple scripts creating race conditions with ipcs and .nts files.

Steve Chiswell
Unidata User Support



On Tue, 2005-05-10 at 12:28, Gabe Langbauer wrote:
> Robert (et al),
> 
> 
> > > Gabe,
> > >
> > > Either one of your scripts isn't running gpend (look into running the
> > > _gf versions), or you have run into another issue (sorry Art, but it's
> >
> > > true)
> >
> > >Hmmm, not sure what you mean here, could you elaborate?
> >
> > Just joking with you..another Linux issue  :))
> >
> > > where gf or gplt will hang if you have too many GEMPAK processes
> > > running sequentially in a GEMPAK script, even if you are running gpend
> >
> > > or using the _gf versions.
> >
> > >Our whole ewall (web map wall) runs primarily on gempak scripts, many
> >of
> >
> > >which run simultaneously.  By-and-large, we see very few problems doing
> > >this.  However, we do run cleanup scripts twice daily that kill-off
> >hung
> >
> > >gplts, etc...  We also increased our ipc queue limits to 72 since many
> > >gempak processes can chew these up pretty fast.
> >
> > We had to resort to cleanup scripts as well.  This is something I don't
> >have to do on Solaris with the same load.
> >
> > >Maybe ours aren't long enough to hit this problem you're seeing... how
> > >long do yours get before they start having trouble?
> >
> > It has been a while, but I believe if we had more than 18 consecutive
> >iterations of say, gdplot2, then we would see issues under Linux (RH7.2,
> >RH9, and
> >Debian (2.4.20) kernel).
> > I also briefly tested under RH Enterprise WS 3.0 and saw the same thing.
> >I could copy the same script over to a Solaris SPARC box (too slow for
> >operational use) and a Solaris x86 box (dedicated to LDM at that time)
> >and
> >it would run without problems using the same GEMPAK version.
> >
> > The cleanup scripts that we both had to resort to under Linux may be
> >Gabe's best bet.
> 
> I am running the cleanup script that comes with gempak (once a day), and
> it appears to me that the time at which my "gf" programs begin running
> away is shortly after the cleanup is called from the cron.  Any thoughts
> on why this might be?
>                                                                               
>                                                                               
>  
> --Gabe
>