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

20030702: 20030630: ruc20 model output



Celia,

The CONDUIT feed contains the bgrb (hybrid level), pgrb (pressure level)
and sgrb (surface grid) ruc files, as you will see from the LDM product
headers, so filing all three will be larger than just the bgrb that you are 
ftp'ing.

The CONDUIT feed is very reliable. Most data reaches us within 20-40 seconds.

I don't see that chisel.rap.ucar.edu is sending in rtstats, so I can't
diagnose your feed quality for latency. Please add the rtstats line to your 
ldmd.conf
file so that we can monitor the quality of your feed. 

Since you are feeding the entire CONDUIT feed, that alone will peak at 1.8GB an 
hour. 
You also have CRAFT and other feeds going to chisel, so I would recommend that 
your queue be 
at least 2GB. If your queue is too small, you could lose products before pqact 
gets to filing them.
If you are doing a lot of pqact processing, you might also want to check to 
make sure your
system isn't bogging down on file I/O. You can issue a "kill -USR2 process_id" 
using
the pqact process id to start verbose logging, and a second time to
go to debug logging. You will see pqact log messages in ldmd.log that state the 
"Delay"
in local processing. Issue the kill command 1 more time to revert to normal 
logging
(you don't want to run too long in debug mode because it will be writing copious
amounts of logging information).

If you see Delay times longer than a few seconds, then your pqact process is 
waining
on disk I/O. If the times are large and approach the lenth of time you
have room for data in your queue, you will loose data.

Steve Chiswell





>From: Celia Chen <address@hidden>
>Organization: UCAR/Unidata
>Keywords: 200307021815.h62IFsLd014044

>Chis,
>
>Thanks for the pqact line for ruc20 data feed.  I am able to file
>by forecast runs. Do you happen to know why the
>the file size is larger than the one that I FTP directly from NCEP? 
>
>**** from ldm feed ****
>-rw-r--r--    1 ldm      rap      85295736 Jul  2 11:43 2003070217_fh.0002
>-rw-r--r--    1 ldm      rap      85981462 Jul  2 11:47 2003070217_fh.0003
>***** FTP from NCEP ***** 
>-rw-rw-r--    1 celia    rap      57890226 Jul  1 15:50 ruc2.t15z.bgrb20f03   
>                               
>
>And, is the time (yyyymmddhh) on the file name the model run time?  
>Also, how reliable is the LDM feed of the CONDUIT datasets? I did notices
>some files are much smaller and some forecasts were missing.
>
>
>Thanks in advance.
>
>Celia
>
>
>On Tue, Jul 01, 2003 at 04:47:34PM -0600, Unidata Support wrote:
>> >From: Celia Chen <address@hidden>
>> >Organization: UCAR/Unidata
>> >Keywords: 200306302030.h5UKUmLd002830
>> 
>> >Hi,
>> >
>> >I was wondering if we can get the RUC20 model output files via the LDM.
>> >If so, do you have a sample pqact line that I can use?
>> >
>> >Thanks.
>> >
>> >Celia
>> >
>> 
>> Celia,
>> 
>> The CONDUIT data feed carries the RUC20 grids (aka grid #252).
>> 
>> If you just want to PIPE all the data, then
>> 
>> NMC2 RUC2/#252
>>      PIPE    pipe_action
>> 
>> to file by forecast run
>> 
>> NMC2 MT.ruc_CY.(..)/RD.(........)/PT.grid_DF.gr1/(fh.*)_tl.*RUC2/#252
>>      FILE    data/conduit/ruc20/\2\1_\3
>> 
>> in the above, \2 is YYYYMMDD, \1 is HH, and \3 is fh.anal, fh.0001, fh.0002 
> etc.
>> 
>> 
>> Steve Chiswell
>> ****************************************************************************
>> Unidata User Support                                    UCAR Unidata Program
>> (303)497-8643                                                  P.O. Box 3000
>> address@hidden                                   Boulder, CO 80307
>> ----------------------------------------------------------------------------
>> Unidata WWW Service              http://my.unidata.ucar.edu/content/support 
>> ****************************************************************************
>