[Dxspider-support] Cluster hangs

[Kelly Jones]

Hi Dirk and all.  See comments below...

At 07:57 AM 3/6/2005, you wrote:
>On Sun, 2005-03-06 at 09:16 -0500, T3 wrote:
> > I think were going tin the wrong direction with this
> > the problem are the hangups, hangups are caused from extreme fast access
> > of cluster files, and the slow perl.
> >  *if* thats what the problem is, im seeing more excess traffic slam the
> > cluster
> > while its trying to process spots
>
>Could I just point out that, over several years now, that DXSpider has
>been the only software that has managed to keep up and stay up during
>CQWW. There are nodes out there with more than 100 users during this
>contest. They cope.

I agree.  DX Spider was a Godsend after using CLX for years.  Up until 
about a month ago I was using a very old build of Spider, circa two or so 
years old.  I never once experienced this 'hang' until replacing the old 
cluster box with a 'new and improved' box.  Obviously changing an entire 
box changes many factors at once so it's difficult to pinpoint any one 
thing.  However, during ARRL CW was the first time I have ever seen Spider 
get hung up during high volumes.  This was on the new box running Mdrk 
10.1/Perl 5.8.5/1.51 bld 58.323.


> >
> > ive been watching backbone traffic and the PC16 and 17 are out of
> > control along with dupe file access
> > its next to impossible to RF backbone link now even at 9.6K to another
> > cluster, My 9.6K radio is
> > having trouble with the excess backbone fat. If this don't get cleared
> > up. I'll be forced to Jump
> > ship to another flavor of cluster. to try to save these expensive radio
> > equipment.
>
>If PC16/17s are truly your problem then other cluster software is not
>going to help as they all use PC16/17 (compatibility issue, you see); I
>would also say that the 'dupefile' is the only thing that is stopping it
>all just falling down in a hopeless mess. You could try doing some
>careful route filtering which will reduce (and maybe eliminate) loops.
>The only alternative you have is (probably)  DXSpider 2 running Aranea
>protocol. And since this is in restricted alpha testing only, that may
>be a problem.

Again, I didn't make any changes to nodes/routes/connections.  For whatever 
reason my old setup ran fine.  But now something gets jammed.  So I'm not 
sure this is the problem unless there is something different with the way 
the PC16/17 is being processed in the new flavors of Spider vs the old.

Everything appears to have run fine overnight here.  My spots are up to 
date and any command request is pretty snappy.  During ARRL CW, however, 
Sunday seemed to be my worse day.  I'll keep watching here.  As has been 
stated, this problem only seems to arise during high traffic volume.

Perhaps Charlie is on to something also.  Maybe Perl 5.8.5 is the 
culprit.  I don't recall what my old cluster was running, but I can dig it 
out of the closet and take a look.  One other thing I notice is that my 
memory usage is at the full 1g level - there's a physical 1g in the 
box.  However, only very little is being swapped.  But keep in mind that I 
have XWin running, my grayline map and a few other small apps.  So I'm not 
sure the mem usage is too unusual.

Kelly - N0VD