Main menu


create_test issues

2 posts / 0 new
Last post
create_test issues


I am trying to run CCSM3 (3.0 released 2004) test_case scenarios on an SGI and some odd things happen. Firstly, opposite to what the manual says, no "batch.$MACH" file is created in my $CCSM_ROOT/scripts directory when I execute create_test. Though this is not a big issue by any means. Secondly, for the BR and HY test case scenario's two folders are created in the test_root directory, in my case they are THY.01a.T31_gx3v5.B.mclaren.100015 and THY.01a.T31_gx3v5.B.mclaren.100015.ref, I was not entirely sure which folder I should attempt to run the test from (and this is not elaborated on in the manual). In any case, I tried building the model from THY.01a.T31_gx3v5.B.mclaren.100015, however it soon crashes when trying to compile the ice model. The log file shows that it is looking for input data under my THY.01a.T31_gx3v5.B.mclaren.100015 directory, which is clearly wrong (please note I specify the input data directory to the correct directory both in the create_test command and in my env.linux.$MACH file).

The model however does build from the THY.01a.T31_gx3v5.B.mclaren.100015.ref directory. Now I'm getting to the crux of my problem, after I submit the THY.01a.T31_gx3v5.B.mclaren.100015.ref.test file and it supposedly finishes, the Teststatus.out file shows only the following output;

doing a 5 day hybrid test
Hybrid Test log is /scratch/2week/nherold/CCSM/output/ccsm_test/cpl/cpl.log.080721-101929

So it is clearly not comparing both the reference and non-reference runs (perhaps it's not even running both of them). I want to know whether a) I'm supposed to build and run the test scenario from the *.ref/ directory or not, and b) whether the output batch.$MACH file is actually meant to be created or whether this is one of those things that was written up but not implemented. As all of these things seem to indicate that something is not working the way it is meant to straight from step 1. It is also interesting to note that I can run a 10 year simulation via create_newcase with no problems.



Just in case anyone runs across this thread with a similar issue, here are the steps which worked for me in running the tests.

First, I didn't try to use the batch method (which involves setting the -testroot option, I think). I just created and ran all of the tests individually. Also, I found that I had to build the executable for each test (the ****.build script) before running the ****.test script, since there seem to be issues with passing variables to the build script when called from the run script on my machine. So for DB and ER tests,

1) Use create_test
2) cd to the directory
3) ./*****.build
4) ./*****.test
5) look at Teststatus.out

For the branch and hybrid tests, two directories are created. You have to build and run in the .ref directory first, then in the other directory. While the non-ref test script supposedly builds and runs the ref case first, I found this didn't work. So the steps are:

1) ./create_test ...
2) cd to ref directory
3) ./***.build
4) ./*****.test
5) cd to non-ref directory
6) ./***.build
7) ./*****.test
8) Look at teststatus.out

This results in an unnecessary re-run of the ref case, but I didn't feel like messing with the build and run scripts.

Hopefully this will help others.

Log in or register to post comments

Who's new

  • 1658093099@...
  • mborreggine@...
  • kabirtam@...
  • suns@...
  • liangpeng0405@...