20000428: update on 17 April McIDAS bugfix and AVN 12Z run (cont.)

>From: weather <address@hidden>
>Organization: NMSU/NSBF
>Keywords: 199909242206.QAA23500 McIDAS-XCD DMGRID AVN grid missing SC5.0

Robert,

Well, the 12 Z AVN run is in, and the message is much the same as it
was for the 0 Z run except that I am including numbers from a different
Sun SPARC where dmgrid.k was compiled using SC5.0:

machine        dataset         #MERC  #PS    Total  OSystem  Compilers
--------------+---------------+------+------+------+--------+---------
adde.unidata   RTGRIDS/AVN.39                1587   x86      SC5.0
wxmcidas.nsbf  RTGRIDS/AVN.39                1587   x86      SC5.0
windfall.evsc  RTGRIDS/AVN.39                1587   SPARC    SC4.2
stratus.al     RTGRIDS/AVN.39                1587   SPARC    SC5.0
psnldm.nsbf    RTGRIDS/AVN.39                1587   SPARC    SC5.0

adde.unidata   RTGRIDS/AVN.49                1016   x86      SC5.0
wxmcidas.nsbf  RTGRIDS/AVN.49                1016   x86      SC5.0
windfall.evsc  RTGRIDS/AVN.49                1016   SPARC    SC4.2
stratus.al     RTGRIDS/AVN.49                1016   SPARC    SC5.0
psnldm.nsbf    RTGRIDS/AVN.49                1016   SPARC    SC5.0

adde.unidata   RTGRIDS/AVN.59                 773   x86      SC5.0
wxmcidas.nsbf  RTGRIDS/AVN.59                 773   x86      SC5.0
windfall.evsc  RTGRIDS/AVN.59                 773   SPARC    SC4.2
stratus.al     RTGRIDS/AVN.59                 773   SPARC    SC5.0
psnldm.nsbf    RTGRIDS/AVN.59                 638   SPARC    SC5.0

I decided to use NMCAMT to look at all of the realtime GRID files to
see what grids might be incomplete.  Here is the comparison between
adde.unidata and psnldm.nsbf:

/home/mcidas/upcworkdata% nmcamt.k FORM=ALL
 GridF Grid Parm Level  RT VT Mod  location  nr  nc received  sectors
  5149  260 T    2      0 108 MRF  18049459  73  73  084435  |xx*x|
total number of grids= 1865
total number of partial grids= 1
nmcamt.k: done

/export/home/mcidas/workdata% nmcamt.k FORM=ALL
 GridF Grid Parm Level  RT VT Mod  location  nr  nc received  sectors
  5149  260 T    2      0 108 MRF  22513647  73  73  091122  |xx*x|
  5419  484 Z    150    0  36 AVN   5071922 145 289  051912  |xxxx****|
  5419  634 ABV  500    0  42 AVN   9656136 145 289  052225  |xxxx****|
  5419  768 V    700    0  48 AVN  11594018 145 289  052338  |xxx*xxx*|
  5419  769 V    850    0  48 AVN  11585726 145 289  052338  |****x**x|
  5009  246 V    850    0  72 UKMT  5082369 145 289  051912  |*xxxxxxx|
  5009  247 T    850    0  30 UKMT  4838395 145 289  051911  |*****x**|
  5009  248 T    100    0  60 UKMT  5086516 145 289  051912  |x*******|
  5009  249 V    400    0  72 UKMT  5090663 145 289  051912  |x*******|
  5009  251 T    700    0  30 UKMT  5162616 145 289  051922  |xxxx****|
  5009  275 W    700    0  48 UKMT 11454550 145 289  052327  |*xxxxxxx|
  5009  277 T    150    0  30 UKMT 12165513 145 289  052350  |xxxx****|
  5009  278 U    200    0  60 UKMT 11938110 145 289  052349  |xxxx****|
  5009  292 U    300    0  30 UKMT 19903989 145 289  052909  |*****xx*|
  5009  293 V    850    0  60 UKMT 19908569 145 289  052909  |*****xx*|
  5009  294 U    SFC    0  48 UKMT 20723479 145 289  052940  |xxxxx***|
  5009  296 U    250    0  30 UKMT 20842191 145 289  052950  |xxxx****|
  5009  301 V    250    0  60 UKMT 23189280 145 289  053056  |*****xxx|
  5009  922 U    850   12  60 UKMT 24743202 145 289  173114  |*xx*xxxx|
  5009  923 U    150   12  42 UKMT 24738508 145 289  173114  |*****xx*|
  5009  924 V    150   12  42 UKMT 24798317 145 289  173234  |x**x****|
total number of grids= 38069
total number of partial grids= 21
nmcamt.k: done

These listings show that there were a lot more incomplete grids on
psnldm than on adde.unidata.  The times of the incomplete grids
also are clustered around 5:20 Z and 17:30 Z.  These are the times
that the AVN model grids should be streaming in on the IDD.

Working on a hypothesis that there is still something wrong with your
copy of dmgrid.k, I decided to run a test.  Since the UVa machine is
using a version of dmgrid.k that was built with SC4.2 compilers, I
decided to check out more thoroughly the new machine I included in the
list above: stratus.al.  Here is what I found:

stratus.al:

uname -a
SunOS cumulus 5.7 Generic_106541-09 sun4u sparc SUNW,Ultra-Enterprise

f77 -V
f77: WorkShop Compilers 5.0 99/09/16 FORTRAN 77 5.0 patch 107596-03
Usage: f77 [ options ] files.  Use 'f77 -flags' for details


psnldm:

uname -a
SunOS psnldm 5.7 Generic_106541-09 sun4u sparc SUNW,UltraSPARC-IIi-Engine

f77 -V
f77: WorkShop Compilers 5.0 99/07/22 FORTRAN 77 5.0 patch 107596-02

Apparently, the patch level for Fortran on stratus.al is newer than the
one on psnldm.  I have no idea if this is significant or not.

I FTPed the copy of dmgrid.k that is running on stratus.al to psnldm
and then did some comparisons:

/export/home/mcidas% ls -l dmgrid.k
-rwxrwxr-x   1 mcidas   staff     351088 Apr 28 21:34 dmgrid.k*
/export/home/mcidas% ls -l bin/dmgrid.k
-rwxrwxr-x   2 mcidas   staff     350860 Apr 28 01:58 bin/dmgrid.k*

You can see that the one built on psnldm is smaller than the one
built on stratus.al.  Again, I don't know if this means anything or
not, but it is something at least.

Since I havn't heard back from you on what you would like to try on
psnldm, I decided to see if the copy of dmgrid.k that I FTPed would
do better at decoding grids than the natively built one.  I hope that
you don't mind.

We will have to wait until the next batch of AVN grids arrives to
see if the stratus.al copy of dmgrid.k does better than yours.
I will check in later to see...

Tom

>From address@hidden Fri Apr 28 16:10:55 2000
>Subject: 20000428: update on 17 April McIDAS bugfix and AVN 12Z run (cont.)

>Thanks for checking into that Tom.  Do to the machine whatver you think
>is necessary.  I appreciate you looking into this.  I will get the
>latest SC5.0 patch and put it on psnldm.

>BTW psnldm is now being fed by wxmcidas.  This was not the case until
>this morning.

>Robert
>NSBF Meteorology

>From address@hidden Fri Apr 28 16:20:10 2000

>I think I see the problem.  I was getting RECLASS statements in my
>ldmd.log when these partial grids occurred.  My usual problems.
>Is a T1 really not a fat enough pipe?  Does everyone else have 
>multiple T1's or T3's?  The other problem is the routing to any feed
>I have tried is always circuitous.  So it looks like it is decoding
>correctly.  Maybe I will get a clean feed tonight and will see for sure.

>Robert
>NSBF Meteorology