port to vista tacc system

ESMCI · Aug 2, 2024 · e89176e · e89176e
1 parent 452642c
commit e89176e
Show file tree

Hide file tree

Showing 6 changed files with 21 additions and 18 deletions.
diff --git a/config/cesm/machines/config_batch.xml b/config/cesm/machines/config_batch.xml
@@ -707,17 +707,19 @@
   </batch_system>
 
   <batch_system MACH="vista" type="slurm" >
-    <batch_submit>ssh vista.tacc.utexas.edu cd $CASEROOT ; sbatch</batch_submit>
+    <batch_submit>ssh login1.vista.tacc.utexas.edu sbatch  </batch_submit>
+    <jobid_pattern>Submitted batch job (\d+)$</jobid_pattern>
     <submit_args>
       <arg flag="--time" name="$JOB_WALLCLOCK_TIME"/>
       <arg flag="-p" name="$JOB_QUEUE"/>
       <arg flag="--account" name="$PROJECT"/>
     </submit_args>
     <queues>
-      <queue walltimemax="2:00:00" nodemin="1" nodemax="30">gh-dev</queue>
-      <queue walltimemax="2:00:00" nodemin="1" nodemax="250" default="true">gh-dev-64k</queue>
+      <queue walltimemax="2:00:00" nodemin="1" nodemax="32" default="true">gg</queue>
+      <queue walltimemax="2:00:00" nodemin="1" nodemax="29">gg-4k</queue>
     </queues>
   </batch_system>
+
   <batch_system MACH="zeus" type="lsf">
     <batch_env>-env</batch_env>
     <submit_args>

diff --git a/config/cesm/machines/config_compilers.xml b/config/cesm/machines/config_compilers.xml
@@ -629,8 +629,8 @@ using a fortran linker.
   </FFLAGS>
   <SUPPORTS_CXX>TRUE</SUPPORTS_CXX>
   <SLIBS>
-    <append> -L$ENV{NETCDF_PATH}/lib -lnetcdff -lnetcdf </append>
-    <append MPILIB="openmpi"> -L$ENV{PNETCDF_PATH}/lib -lpnetcdf</append>
+    <append> -Wl,-rpath,$(NETCDF_PATH)/lib -L$ENV{NETCDF_PATH}/lib -lnetcdff -lnetcdf </append>
+    <append MPILIB="openmpi"> -Wl,-rpath,$(PNETCDF_PATH)/lib -L$ENV{PNETCDF_PATH}/lib -lpnetcdf</append>
   </SLIBS>
 </compiler>
 

diff --git a/config/cesm/machines/config_machines.xml b/config/cesm/machines/config_machines.xml
@@ -2905,12 +2905,12 @@ This allows using a different mpirun command to launch unit tests
     <GMAKE_J>4</GMAKE_J>
     <BATCH_SYSTEM>slurm</BATCH_SYSTEM>
     <SUPPORTED_BY></SUPPORTED_BY>
-    <MAX_TASKS_PER_NODE>72</MAX_TASKS_PER_NODE>
-    <MAX_MPITASKS_PER_NODE>72</MAX_MPITASKS_PER_NODE>
+    <MAX_TASKS_PER_NODE>144</MAX_TASKS_PER_NODE>
+    <MAX_MPITASKS_PER_NODE>144</MAX_MPITASKS_PER_NODE>
     <mpirun mpilib="openmpi">
       <executable>mpirun</executable>
       <arguments>
-	<arg name="num_tasks"> -np $TOTALPES</arg>
+	<arg name="num_tasks"> -np {{ total_tasks }}</arg>
       </arguments>
     </mpirun>
     <module_system type="module">

diff --git a/scripts/lib/CIME/XML/env_batch.py b/scripts/lib/CIME/XML/env_batch.py
@@ -603,7 +603,7 @@ def _submit_single_job(self, case, job, dep_jobs=None, allow_fail=False,
         if batch_system is None or batch_system == "none" or no_batch:
             logger.info("Starting job script {}".format(job))
             function_name = job.replace(".", "_")
-            job_name = "."+job
+            job_name = os.path.join(self._caseroot,job)
             if not dry_run:
                 args = self._build_run_args(job, True, skip_pnl=skip_pnl, set_continue_run=resubmit_immediate,
                                             submit_resubmits=not resubmit_immediate)
@@ -688,16 +688,17 @@ def _submit_single_job(self, case, job, dep_jobs=None, allow_fail=False,
         if batch_system == 'lsf' and batch_env_flag == 'none':
             sequence = (run_args, batchsubmit, submitargs, batchredirect, get_batch_script_for_job(job))
         elif batch_env_flag:
-            sequence = (batchsubmit, submitargs, run_args, batchredirect, get_batch_script_for_job(job))
+            sequence = (batchsubmit, submitargs, run_args, batchredirect, os.path.join(self._caseroot,get_batch_script_for_job(job)))
         else:
-            sequence = (batchsubmit, submitargs, batchredirect, get_batch_script_for_job(job), run_args)
+            sequence = (batchsubmit, submitargs, batchredirect, os.path.join(self._caseroot,get_batch_script_for_job(job)), run_args)
 
         submitcmd = " ".join(s.strip() for s in sequence if s is not None)
         if dry_run:
             return submitcmd
         else:
             logger.info("Submitting job script {}".format(submitcmd))
-            output = run_cmd_no_fail(submitcmd, combine_output=True)
+#            output = run_cmd_no_fail(submitcmd, combine_output=True)
+            s, output, e = run_cmd(submitcmd, combine_output=True)
             jobid = self.get_job_id(output)
             logger.info("Submitted job id is {}".format(jobid))
             return jobid

diff --git a/scripts/lib/CIME/provenance.py b/scripts/lib/CIME/provenance.py
@@ -83,7 +83,7 @@ def _save_build_provenance_cesm(case, lid): # pylint: disable=unused-argument
     manic = os.path.join("manage_externals","checkout_externals")
     manic_full_path = os.path.join(srcroot, manic)
     out = None
-    if os.path.exists(manic_full_path):
+    if shutil.which('svn') and os.path.exists(manic_full_path):
         args = " --status --verbose --no-logging"
         stat, out, err = run_cmd(manic_full_path + args, from_dir=srcroot)
         errmsg = """Error gathering provenance information from manage_externals.

diff --git a/scripts/tests/scripts_regression_tests.py b/scripts/tests/scripts_regression_tests.py
@@ -303,21 +303,21 @@ def test_a_createnewcase(self):
         self.assertTrue(os.path.exists(os.path.join(testdir, "case.setup")))
 
         run_cmd_assert_result(self, "./case.setup", from_dir=testdir)
-        run_cmd_assert_result(self, "./case.build --skip-provenance-check", from_dir=testdir)
+        run_cmd_assert_result(self, "./case.build", from_dir=testdir)
 
         with Case(testdir, read_only=False) as case:
             ntasks = case.get_value("NTASKS_ATM")
             case.set_value("NTASKS_ATM", ntasks+1)
         # this should fail with a locked file issue
-        run_cmd_assert_result(self, "./case.build --skip-provenance-check",
+        run_cmd_assert_result(self, "./case.build",
                               from_dir=testdir, expected_stat=1)
 
         run_cmd_assert_result(self, "./case.setup --reset", from_dir=testdir)
-        run_cmd_assert_result(self, "./case.build --skip-provenance-check", from_dir=testdir)
+        run_cmd_assert_result(self, "./case.build", from_dir=testdir)
         with Case(testdir, read_only=False) as case:
             case.set_value("CHARGE_ACCOUNT", "fred")
         # this should not fail with a locked file issue
-        run_cmd_assert_result(self, "./case.build --skip-provenance-check",from_dir=testdir, expected_stat=0)
+        run_cmd_assert_result(self, "./case.build",from_dir=testdir, expected_stat=0)
 
         run_cmd_assert_result(self, "./case.st_archive --test-all", from_dir=testdir)
 
@@ -488,7 +488,7 @@ def test_f_createnewcase_with_user_compset(self):
 
         run_cmd_assert_result(self, "%s/create_newcase %s"%(SCRIPT_DIR, args), from_dir=SCRIPT_DIR)
         run_cmd_assert_result(self, "./case.setup", from_dir=testdir)
-        run_cmd_assert_result(self, "./case.build --skip-provenance-check", from_dir=testdir)
+        run_cmd_assert_result(self, "./case.build", from_dir=testdir)
 
         cls._do_teardown.append(testdir)