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

[McIDAS #EVZ-798953]: It's official...I am clueless sometimes!

Hi Gilbert,

re:
> First, to you and yours at UNIDATA...Merry Christmas and have a great
> 2008!

You too!

> Since I got this machine replaced in early September,
> weather.admin.niu.edu, it's been crashing, as you know. Sometimes it takes
> four hours; sometimes it took a week. Freezes, kernel oopses with
> errors, and sometimes it happens so quick there aren't errors. Last night,
> I think I finally figured out why.

Eureka!

> When I installed Fedora 7 on here, it told me it had successfully
> installed the C and Fortran compilers. And then I installed some extra
> stuff beyond that, to make sure all was in place. Of course, that didn't
> stop it from crashing, kernel oopses and freezes still continued.
> 
> Until yesterday. That's when I really checked to make sure all the GCC,
> C++, FORTRAN and everything else needed had indeed installed, like it
> said it did in the installation.

So, are you saying that your 'yum' updates did not catch the fact that the
development environment was completely installed?  I think that this
would mean that the "registry" that yum uses indicated that the various
packages had been completely installed, but they weren't or pieces had
been deleted.  Unless, of course, you make the cardinal sin if copying
packages from one machine to another.  I say this since I have made the same
mistake myself!

> In the immortal words of Ronald Reagan, "Trust...but verify".

:=)

> I am embarrassed. Most of it didn't install. I don't know why, but
> yesterday I spent all day installing all of it and then some. Anything and
> everything remotely tied into programming languages got thrown in.
> Suddenly, I noticed NLDN data processing on McIDAS. As well as other
> things. And my machine load average dropped HUGE.

The ldm-mcidas NLDN  decoder, nldn2md, needs the shared library that gets
installed when g77 (compat-g77) gets installed.  At one point this was
installed on one of your machines (I think weather.admin)... I specifically
remember corresponding with you about the need to install g77 so that nldn2md
would work.  I _think_, however, that this was before your upgrade to
Fedora 7.

> So today, I went for broke. I turned EVERYTHING back on. Everything for
> McIDAS, everything for GEMPAK. NIMAGE, Level3, level2, bring it on. I'm
> intentionally trying to crash this thing and melt it into the ground. And
> you know what? I can't!!!!

Excellent!

> The highest load average I can get is .99! When I first installed
> everything and you turned it all on, I was pegging 6. Now I know why.
> Stuff wasn't getting processed, I/O collisions were an awful car wreck on
> that highway, and the kernel wasn't liking it. And error messages weren't
> forthcoming.

A load average of around 1 is what I always expected (except at peak data
times) on your systems.  As I said previously, they are pretty good machines,
so seeing excessively high load averages was completely unexpected.

> Granted, I have yet to see how it does in peak times, but that's coming
> up shortly. But I think this has been licked. All errors in my
> ldmd.logs are gone.

Fantastic.  That should help a bunch.

> From McIDAS and from GEMPAK, stuff I've been trying
> to figure out for months. And a few minor errors on weather2/weather3
> have also disappeared when installing a few periphery C++/FORTRAN/GCC
> accessories.

Did you happen to notice exactly which package(s) when installed properly
caused the problems to go away?  I ask since we typically don't run into
these sorts of problems because we install everything related to development.
We do this because it makes our life simpler, but the downside is that
we can not let users know exactly which packages must be installed and
which ones are optional.  I recently had some intense experience in this
on openSUSE while hosting a McIDAS training workshop in Brazil.  Live and
learn...

> So Tom, I tell you...trust, but verify.

Yup.

> And do you happen to have a big enough rock I can crawl under? I am so
> sorry for all the time you lost trying to help me figure this out. And if
> this does work, you can add back weather.admin.niu.edu to the ADDE server
> list.

No worries.  I am hopeful that you did, in fact, find the cause of your
problems and resolved them.

By the way, I made some additional code modifications in McIDAS that I
would like to install on your machines.  The modifications are necessary
for installations built with gfortran, and I think one or two of your
machines used gfortran for their builds.

Cheers,

Tom
****************************************************************************
Unidata User Support                                    UCAR Unidata Program
(303) 497-8642                                                 P.O. Box 3000
address@hidden                                   Boulder, CO 80307
----------------------------------------------------------------------------
Unidata HomePage                       http://www.unidata.ucar.edu
****************************************************************************


Ticket Details
===================
Ticket ID: EVZ-798953
Department: Support McIDAS
Priority: Normal
Status: Closed