JCSDA / spack-stack

Creative Commons Zero v1.0 Universal
26 stars 44 forks source link

JEDI CI container specs update and FMS update for JEDI-FV3 #1099

Closed climbfuji closed 4 months ago

climbfuji commented 5 months ago

Summary

  1. Add missing packages to JEDI-CI specs for container builds (needed by JEDI CI system)
  2. Update FMS from release-jcsda to 2023.04 for jedi-fv3-env and gmao-swell-env (the latter to be confirmed by GMAO). This is needed for the FMS + FV3-DYCORE updates in fv3-jedi etc.

Testing

  1. [x] Built gcc-openmpi container manually on EC2
  2. [x] Needs testing in JEDI-CI framework after test containers have been built
  3. [ ] Needs testing with gmao-swell-env on Discover (if change is ok for GMAO)

Applications affected

JEDI CI (JEDI-FV3)

Systems affected

Container builds

Dependencies

n/a

Issue(s) addressed

n/a

Checklist

climbfuji commented 5 months ago

@dooruk and/or @rtodling Do you know if it is ok to switch gmao-swell-env to fms@2023.04 ? I believe it is, because even the JCSDA version used in gmao-swell-env now is newer than what GEOS uses. Thanks!

Dooruk commented 4 months ago

@climbfuji this shouldn't impact Swell. However @rtodling mentioned there could be some implications in terms of linking GEOSgcm and JEDI (running GEOS through JEDI) work. That is in a different env (geos-gcm-env) but let me tag @mathomp4 here just in case fms change somehow impacts his work?

climbfuji commented 4 months ago

@climbfuji this shouldn't impact Swell. However @rtodling mentioned there could be some implications in terms of linking GEOSgcm and JEDI (running GEOS through JEDI) work. That is in a different env (geos-gcm-env) but let me tag @mathomp4 here just in case fms change somehow impacts his work?

You are right, geos-gcm-env doesn't load any fms by default, so that shouldn't be a problem either

climbfuji commented 4 months ago

@eap Is this working for you? Should I open it pul for reviews and merging? Thanks!

eap commented 4 months ago

@climbfuji I've been building these today (and likely this will continue overnight) - I should be able to test the lot of them in our "ci-next" environment on Tuesday and I'll report back.

eap commented 4 months ago

Well, this is working for CI but the fms change looks like it has broken the tests. All three build environments are failing in the same way;

[ 22%] Built target test_iodaconv_obserror.x
Scanning dependencies of target fv3
[ 22%] Building Fortran object fv3/CMakeFiles/fv3.dir/model/fv_arrays.F90.o
/workdir/test_root/jedi-bundle/fv3/model/fv_arrays.F90:26:2:

   26 |   use mpp_domains_mod,       only: domain2d
      |  1~~~~~~~~~~~~~~~
Fatal Error: fms_platform.h: No such file or directory
compilation terminated.
make[2]: *** [fv3/CMakeFiles/fv3.dir/build.make:114: fv3/CMakeFiles/fv3.dir/model/fv_arrays.F90.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:28576: fv3/CMakeFiles/fv3.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....
[ 22%] Linking CXX executable ../../../bin/print_queries.x
[ 22%] Built target print_queries.x
make: *** [Makefile:166: all] Error 2

Here's a direct link to the failures: https://github.com/JCSDA-internal/mpas-jedi/pull/966/checks?check_run_id=24712175023

climbfuji commented 4 months ago

That is expected. We need this version of fms as "next" containers for a set of PRs

climbfuji commented 4 months ago

Well, this is working for CI but the fms change looks like it has broken the tests. All three build environments are failing in the same way;

[ 22%] Built target test_iodaconv_obserror.x
Scanning dependencies of target fv3
[ 22%] Building Fortran object fv3/CMakeFiles/fv3.dir/model/fv_arrays.F90.o
/workdir/test_root/jedi-bundle/fv3/model/fv_arrays.F90:26:2:

   26 |   use mpp_domains_mod,       only: domain2d
      |  1~~~~~~~~~~~~~~~
Fatal Error: fms_platform.h: No such file or directory
compilation terminated.
make[2]: *** [fv3/CMakeFiles/fv3.dir/build.make:114: fv3/CMakeFiles/fv3.dir/model/fv_arrays.F90.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:28576: fv3/CMakeFiles/fv3.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....
[ 22%] Linking CXX executable ../../../bin/print_queries.x
[ 22%] Built target print_queries.x
make: *** [Makefile:166: all] Error 2

Here's a direct link to the failures: https://github.com/JCSDA-internal/mpas-jedi/pull/966/checks?check_run_id=24712175023

@shlyaeva Note that @eap has new containers with the updated FMS. We need those for the FV3 dycore update branches.

eap commented 4 months ago

@shlyaeva - in order to use the new containers add jedi-ci-next=true to a line in your pull request description.

climbfuji commented 4 months ago

@shlyaeva - in order to use the new containers add jedi-ci-next=true to a line in your pull request description.

I don't think this works. It needs a different branch of jedi-bundle, too.