create_test
create_test is a script in CIMEROOT/scripts.
$ ./create_test --help
usage: create_test [-h] [-d] [-v] [-s] [--no-run] [--no-build] [--no-setup]
[-u] [--save-timing] [--no-batch] [--single-exe]
[--single-submit] [-r TEST_ROOT]
[--output-root OUTPUT_ROOT] [--baseline-root BASELINE_ROOT]
[--clean] [-m MACHINE] [--mpilib MPILIB] [-c COMPARE]
[-g GENERATE] [--xml-machine XML_MACHINE]
[--xml-compiler XML_COMPILER] [--xml-category XML_CATEGORY]
[--xml-testlist XML_TESTLIST] [--driver {nuopc}]
[--compiler COMPILER] [-n] [-p PROJECT] [-t TEST_ID]
[-j PARALLEL_JOBS] [--proc-pool PROC_POOL]
[--walltime WALLTIME] [-q QUEUE] [-f TESTFILE] [-o]
[--wait] [--allow-pnl] [--check-throughput]
[--check-memory] [--ignore-namelists] [--ignore-diffs]
[--ignore-memleak] [--force-procs FORCE_PROCS]
[--force-threads FORCE_THREADS] [-i INPUT_DIR]
[--pesfile PESFILE] [--retry RETRY] [-N]
[--workflow WORKFLOW] [--chksum] [--srcroot SRCROOT]
[--force-rebuild] [--mail-user MAIL_USER] [-M MAIL_TYPE]
[testargs ...]
positional arguments:
testargs Tests to run. Testname form is TEST.GRID.COMPSET[.MACHINE_COMPILER]
options:
-h, --help show this help message and exit
--no-run Do not run generated tests
--no-build Do not build generated tests, implies --no-run
--no-setup Do not setup generated tests, implies --no-build and --no-run
-u, --use-existing Use pre-existing case directories they will pick up at the
latest PEND state or re-run the first failed state. Requires test-id
--save-timing Enable archiving of performance data.
--no-batch Do not submit jobs to batch system, run locally.
If false, this will default to machine setting.
--single-exe Use a single build for all cases. This can
drastically improve test throughput but is currently use-at-your-own risk.
It's up to the user to ensure that all cases are build-compatible.
E3SM tests belonging to a suite with share enabled will always share exes.
--single-submit Use a single interactive allocation to run all the tests. This can
drastically reduce queue waiting but only makes sense on batch machines.
-r TEST_ROOT, --test-root TEST_ROOT
Where test cases will be created. The default is output root
as defined in the config_machines file
--output-root OUTPUT_ROOT
Where the case output is written.
--baseline-root BASELINE_ROOT
Specifies a root directory for baseline datasets that will
be used for Bit-for-bit generate and/or compare testing.
--clean Specifies if tests should be cleaned after run. If set, all object
executables and data files will be removed after the tests are run.
-m MACHINE, --machine MACHINE
The machine for creating and building tests. This machine must be defined
in the config_machines.xml file for the given model. The default is to
to match the name of the machine in the test name or the name of the
machine this script is run on to the NODENAME_REGEX field in
config_machines.xml. WARNING: This option is highly unsafe and should
only be used if you are an expert.
--mpilib MPILIB Specify the mpilib. To see list of supported MPI libraries for each machine,
invoke ./query_config. The default is the first listing .
-c COMPARE, --compare COMPARE
While testing, compare baselines against the given compare directory.
-g GENERATE, --generate GENERATE
While testing, generate baselines in the given generate directory.
NOTE: this can also be done after the fact with bless_test_results
--xml-machine XML_MACHINE
Use this machine key in the lookup in testlist.xml.
The default is all if any --xml- argument is used.
--xml-compiler XML_COMPILER
Use this compiler key in the lookup in testlist.xml.
The default is all if any --xml- argument is used.
--xml-category XML_CATEGORY
Use this category key in the lookup in testlist.xml.
The default is all if any --xml- argument is used.
--xml-testlist XML_TESTLIST
Use this testlist to lookup tests.The default is specified in config_files.xml
--driver {nuopc} Override driver specified in tests and use this one.
--compiler COMPILER Compiler for building cime. Default will be the name in the
Testname or the default defined for the machine.
-n, --namelists-only Only perform namelist actions for tests
-p PROJECT, --project PROJECT
Specify a project id for the case (optional).
Used for accounting and directory permissions when on a batch system.
The default is user or machine specified by PROJECT.
Accounting (only) may be overridden by user or machine specified CHARGE_ACCOUNT.
-t TEST_ID, --test-id TEST_ID
Specify an 'id' for the test. This is simply a string that is appended
to the end of a test name. If no test-id is specified, a time stamp plus a
random string will be used (ensuring a high probability of uniqueness).
If a test-id is specified, it is the user's responsibility to ensure that
each run of create_test uses a unique test-id. WARNING: problems will occur
if you use the same test-id twice on the same file system, even if the test
lists are completely different.
-j PARALLEL_JOBS, --parallel-jobs PARALLEL_JOBS
Number of tasks create_test should perform simultaneously. The default
is min(num_cores, num_tests).
--proc-pool PROC_POOL
The size of the processor pool that create_test can use. The default is
MAX_MPITASKS_PER_NODE + 25 percent.
--walltime WALLTIME Set the wallclock limit for all tests in the suite.
Use the variable CIME_GLOBAL_WALLTIME to set this for all tests.
-q QUEUE, --queue QUEUE
Force batch system to use a certain queue
-f TESTFILE, --testfile TESTFILE
A file containing an ascii list of tests to run
-o, --allow-baseline-overwrite
If the --generate option is given, then an attempt to overwrite
an existing baseline directory will raise an error. WARNING: Specifying this
option will allow existing baseline directories to be silently overwritten.
--wait On batch systems, wait for submitted jobs to complete
--allow-pnl Do not pass skip-pnl to case.submit
--check-throughput Fail if throughput check fails. Requires --wait on batch systems
--check-memory Fail if memory check fails. Requires --wait on batch systems
--ignore-namelists Do not fail if there namelist diffs
--ignore-diffs Do not fail if there history file diffs
--ignore-memleak Do not fail if there's a memleak
--force-procs FORCE_PROCS
For all tests to run with this number of processors
--force-threads FORCE_THREADS
For all tests to run with this number of threads
-i INPUT_DIR, --input-dir INPUT_DIR
Use a non-default location for input files
--pesfile PESFILE Full pathname of an optional pes specification file. The file
can follow either the config_pes.xml or the env_mach_pes.xml format.
--retry RETRY Automatically retry failed tests. >0 implies --wait
-N, --non-local Use when you've requested a machine that you aren't on. Will reduce errors for missing directories etc.
--workflow WORKFLOW A workflow from config_workflow.xml to apply to this case.
--chksum Verifies input data checksums.
--srcroot SRCROOT Alternative pathname for source root directory. The default is /home/runner/work/cime
--force-rebuild When used with 'use-existing' and 'test-id', thetests will have their 'BUILD_SHAREDLIB' phase reset to 'PEND'.
--mail-user MAIL_USER
Email to be used for batch notification.
-M MAIL_TYPE, --mail-type MAIL_TYPE
When to send user email. Options are: never, all, begin, end, fail.
You can specify multiple types with either comma-separated args or multiple -M flags.
Logging options:
-d, --debug Print debug information (very verbose) to file /home/runner/work/cime/cime/scripts/create_test.log
-v, --verbose Add additional context (time and file) to log messages
-s, --silent Print only warnings and error messages