Project

General

Profile

problems with cclm runs

Added by Sinclair Chinyoka almost 5 years ago

I have been trying to set up my cosmo-clm model on our cluster. now im carrying out a test run which runs from 1980010100 to 1980030100.

Int2lm runs well but the problem is cclm. It runs up to the last day and complete successfully but when i checked the outputs it has been repeating same dates.

It runs up to 14Z11JAN1980 instead of 1980030100. even if i get the output file. it is named as the lffd198003100 but the details of that file are those of 14Z11JAN1980
. Im using subchain script. The only file which has proper details is the first file (1980010100) all other files have wrong outputs and cant display
i have attached the cclm script
i need help


Replies (18)

RE: problems with cclm runs - Added by Burkhardt Rockel almost 5 years ago

It is not possible to answer without knowing the content of the files in the joblogs, joboutputs and EXPID directories. Please tar and gzip these directories and upload them to the forum.

RE: problems with cclm runs - Added by Burkhardt Rockel almost 5 years ago

The joblogs directories are empty (except the finish_joblist file). You probably did some error in the job submitting part. Please tar and gzip your sp001 directory and upload it to the forum.

RE: problems with cclm runs - Added by Sinclair Chinyoka almost 5 years ago

It also like that in my spool. It only contains one file.

RE: problems with cclm runs - Added by Sinclair Chinyoka almost 5 years ago

Here is the sp001 directory from the work directory

RE: problems with cclm runs - Added by Burkhardt Rockel almost 5 years ago

I need the cclm-sp_2.4/chain/gcm_to_cclm/sp001/ directory not the work directory.

RE: problems with cclm runs - Added by Burkhardt Rockel almost 5 years ago

It seems that the cclm job not even has been submitted.
Can you please:
in int2lm.job.tmpl replace

$RUNDIR/int2lm.log   by  @{JOBLOGFILE}

and in cclm.job.tmpl replace
$RUNDIR/cosmo-clm.log   by  @{JOBLOGFILE}

this writes the log files where they are supposed to be find.
After this please make a new clean start:
subchain clean
subchain start

after the end of the job, please upload the new work/sp001 and cclm-sp_2.4/chain/gcm_to_cclm/sp001/ to the forum.

RE: problems with cclm runs - Added by Sinclair Chinyoka almost 5 years ago

Thank you very much. let me do that now

RE: problems with cclm runs - Added by Sinclair Chinyoka almost 5 years ago

attached log files. Also i noted that instead of it ending at 1980020100 it stopped at 1980013118

RE: problems with cclm runs - Added by Sinclair Chinyoka almost 5 years ago

when proccessing the cclm out eg lffd1980010606 it opens lbfd1980010612.nc
its like its 6 hours ahead and becoz of that it stops at lffd1980013118 instead of lffd1980020100.

RE: problems with cclm runs - Added by Burkhardt Rockel almost 5 years ago

The CCLM run seems to be OK, but the next subchain command was not executed:

more sp001_wk/joblogs/cclm/cclm_sp001_1980_01.o%j
/opt/tsce/share/cu100/mom_priv/jobs/6911.mu01.SC: line 309: subchain: command not found

Can't find file /lustre/home/uon_schinyoka/.vnc/mu01:81.pid
You'll have to kill the Xvnc process manually

I guess you need to put a ./ before your subchain command.

Therefore please change in cclm.job.tmplsubchain arch” to ”./subchain arch
and
in arch.job.tmplsubchain post” to ”./subchain post” and ”subchain cclm" to ”./subchain cclm
In prep.job.tmpl and int2lm.job.tmpl this is already OK.

After that you may either try a

subchain arch

or start clean again with

subchain clean
subchain start

RE: problems with cclm runs - Added by Sinclair Chinyoka almost 5 years ago

Thank you very much.

But when i use gribapi2ctl on any output of cclm and use grads. all the variables are not displaying and the last time step is 14Z11JAN1980 instead of 00Z02FEB1980.

no2t,tmax_2m and tmin_2m are all not displaying.

RE: problems with cclm runs - Added by Sinclair Chinyoka almost 5 years ago

please can you take a look at the following files. It seems like i can only be able to display data from lffd1980010100.nc only. from all other files i cant. Its either i can a message that its a constant value of 0 or data is missing especially for the tot_prec,Tmax_2m and tmin_2m. Also the time steps are wrong

RE: problems with cclm runs - Added by Sinclair Chinyoka almost 5 years ago

arch and post are incurring some errors. please take a look at the log files

RE: problems with cclm runs - Added by Burkhardt Rockel almost 5 years ago

The errors do not seem to result from the chain itself. There are very likely false namelist settings and/or compiler options.
You are downscaling from about 200km (NCEP-RA) to 3km which is a factor of 67 !!!!!!! At least you need to perform a double nesting!
Regarding the compiler options: Please choose an option that stops the program in case of producing NaN. Your job already produces NaNs after the first few time steps but runs still on for the whole month producing always just NaNs.
It is very hard to help you from remote with so many mistakes. Do you plan to take part in the training course? I am sure that would be of valuable help for you.

RE: problems with cclm runs - Added by Sinclair Chinyoka almost 5 years ago

Thank you very much for your help. I final managed to display some interesting outputs. I changed the resolution (dlon) from 0.025 to 0.125.
The only challenge remaining is that when processing the cclm outputs. its finishing at 1980013118 instead of 198002100. Also the file 1980013118 has the time step for 1980011114.

RE: problems with cclm runs - Added by Sinclair Chinyoka almost 5 years ago

Also for the training course. I missed the registration. I wont be coming this year but the next training i will be available. I once attended last year’s training but i was on NWP Group.

    (1-18/18)