USING RESTART FILE

[w_xuej@163_com]

Dear all,      I want to run for 10 days and every day is initialized using a restart file and resubmit 10 times.There are 10 restart files each day from previews model run and it failed.The error is (shr_sys_abort) ERROR: Incorrect TLT computations(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping I'm running CESM1_2_1 with B_2000 compset and f09_g16 resolution. I create a branch run and  I set in env_run.xml.what should I do .Can you help me?

 

[jedwards]

 

1. Restart depends on text files called rpointer.$MODEL for each component model - these files are read to determine which restart files to read so for each of the 10 sets of restart files you need a seperate set of rpointer files.
2. The error you are getting is from the pop ocean model - look in that models log for details.  Often a problem with POP is that you are reading a netcdf file but the model is expecting a binary file  - check that the namelist settings for file types of pop restart is netcdf.  

 

[w_xuej@163_com]

Thank you for your reply.For your anwsers ,I have two questions.1.  I only have a set of rpointer files and only one restart date.How to read restart files from one moment to the next automatically and what shoud I do for each of the 10 sets of restart files having a seperate set of rpointer files?2.  In ocn.log,it implies : WARNING: Input header file does not exist for file: ./test10.pop.r.0001-01-03-00000.nc.hdr Assuming fields will be read in sequential order
(POP_check) WARNING:  time-averaging and coupling frequency may be incompatible; tavg must be integer multiple of coupling freq(POP_check) No fatal error conditions detected 

 

[jedwards]

The experiment you are attempting is outside the capability of our current restart system, you will need to modify the code to do what you want to do, If I were to do this I think that i would add a timestamp to the rpointer file names - but it's a fairly involved change.   This is on our list to fix prior to the cesm 2 release next year. The ocean model thinks that it is reading a binary restart, but I don't understand why - it looks like you have the namelist set properly - it may be because you are trying a branch run and the namelist setting is ignored?   

 

[njn01]

In regards to  your question (2): the POP2 WARNING messages are intended to flag potential problems and alert the user to inspect these items more closely to make sure everything is ok.  In your case, the input file is read correctly, according to the other information printed in your ocn.log file, and in this instance, the warning messages are only an unfortunate consequence of limitations of the error-checking part of the model, which is unable to analyze all possible combinations of restart-file-reading problems.  The same applies to the tavg POP_check warnings; some combinations of time-averaging and coupling frequency can cause problems, and the code is unable to analyze all possible problematic scenarios, so it prints a message to let you know you should confirm that you have selected compatable values.The "POP_check No fatal error conditions" message is generated when none of the POP2 internal consistency checks has failed.  

 

[w_xuej@163_com]

Thanks for yor reply.But what should I do to avoild this? 

 

[w_xuej@163_com]

Thanks for your reply.I have a another question.I run CESM1_2_1 with B_2000 compset for 1 day.I set start_type="startup",and then I change start_type to "continue" and continue running one day and it successed. But I want to continue running another one day and it failed.It shows the following error:Internal pop2 clock not in sync with Sync Clock  pop2 ymd=       10103   pop2 tod=            0   sync ymd=       10104   sync tod=            0What should I do? I have restart files and all rpointer.* show the same date.Besides, I tried change CONTINUE_RUN to TRUE in the env file,but it does not work.So please help me.

 

[jedwards]

In a B compset the Ocean model lags the rest of the components by 1 day when starting from a "startup" state.   So you should run at least 2 days - 3 might be better.  Once you have the restart files and are continuing then you should be able to cycle with a 1 day period.   

 

[w_xuej@163_com]

Thanks for your reply.I have solved this problem,but I have another quesion. If I change restart files constantly,so I need to create a "continue" run or "branch" run ?What is the difference between them?