def _init_configs(self, section, default_fs_option, default_section,
init_kwargs):
cfg = Config.instance()
# get the proper section
if not section:
section = cfg.get_expanded("target", default_fs_option)
# try to read it and fill configs to pass to the file system and the remote file interface
fs_config = {}
fi_config = {}
if isinstance(section, six.string_types):
# when set, the section must exist
if not cfg.has_section(section):
raise Exception(
"law config has no section '{}' to read {} options".format(
section, self.__class__.__name__))
# extend options of sections other than the default one with its values
self._update_section_defaults(default_section, section)
# read the configs from the section for both the file system and remote interface
fs_config = self.parse_config(section)
fi_config = self.file_interface_cls.parse_config(section)
# update both configs with init kwargs
fs_config = merge_dicts(fs_config, init_kwargs, deep=True)
fi_config = merge_dicts(fi_config, init_kwargs, deep=True)
return section, fs_config, fi_config
def arc_create_job_file_factory(self, **kwargs):
# job file fectory config priority: config file < class defaults < kwargs
get_prefixed_config = self.workflow_proxy.get_prefixed_config
cfg = {
"dir":
get_prefixed_config("job", "job_file_dir"),
"mkdtemp":
get_prefixed_config("job", "job_file_dir_mkdtemp", type=bool),
"cleanup":
get_prefixed_config("job", "job_file_dir_cleanup", type=bool),
}
kwargs = merge_dicts(cfg, self.arc_job_file_factory_defaults, kwargs)
return ARCJobFileFactory(**kwargs)
def submit_jobs(self, job_files, **kwargs):
task = self.task
# prepare objects for dumping intermediate submission data
dump_freq = self._get_task_attribute(
"dump_intermediate_submission_data")()
if dump_freq and not is_number(dump_freq):
dump_freq = 50
# progress callback to inform the scheduler
def progress_callback(i, job_id):
job_num = i + 1
# some job managers respond with a list of job ids per submission (e.g. htcondor, slurm)
# batched submission is not yet supported, so get the first id
if isinstance(job_id, list):
job_id = job_id[0]
# set the job id early
self.submission_data.jobs[job_num]["job_id"] = job_id
# log a message every 25 jobs
if job_num in (1, len(job_files)) or job_num % 25 == 0:
task.publish_message("submitted {}/{} job(s)".format(
job_num, len(job_files)))
# dump intermediate submission data with a certain frequency
if dump_freq and job_num % dump_freq == 0:
self.dump_submission_data()
# get job kwargs for submission and merge with passed kwargs
submit_kwargs = self._get_job_kwargs("submit")
submit_kwargs = merge_dicts(submit_kwargs, kwargs)
return self.job_manager.submit_batch(job_files,
retries=3,
threads=task.threads,
callback=progress_callback,
**submit_kwargs)
def _submit(self, job_files, **kwargs):
task = self.task
# job_files is an ordered mapping job_num -> {"job": PATH, "log": PATH/None}, get keys and
# values for faster lookup by numeric index
job_nums = list(job_files.keys())
job_files = [f["job"] for f in six.itervalues(job_files)]
# prepare objects for dumping intermediate submission data
dump_freq = self._get_task_attribute("dump_intermediate_submission_data", True)()
if dump_freq and not is_number(dump_freq):
dump_freq = 50
# get job kwargs for submission and merge with passed kwargs
submit_kwargs = self._get_job_kwargs("submit")
submit_kwargs = merge_dicts(submit_kwargs, kwargs)
# progress callback to inform the scheduler
def progress_callback(i, job_id):
job_num = job_nums[i]
# some job managers respond with a list of job ids per submission (e.g. htcondor, slurm)
# so get the first id as long as batched submission is not yet supported
if isinstance(job_id, list) and not self.job_manager.chunk_size_submit:
job_id = job_id[0]
# set the job id early
self.submission_data.jobs[job_num]["job_id"] = job_id
# log a message every 25 jobs
if i in (0, len(job_files) - 1) or (i + 1) % 25 == 0:
task.publish_message("submitted {}/{} job(s)".format(i + 1, len(job_files)))
# dump intermediate submission data with a certain frequency
if dump_freq and (i + 1) % dump_freq == 0:
self.dump_submission_data()
return self.job_manager.submit_batch(job_files, retries=3, threads=task.threads,
callback=progress_callback, **submit_kwargs)
def arc_create_job_manager(self, **kwargs):
kwargs = merge_dicts(self.arc_job_manager_defaults, kwargs)
return ARCJobManager(**kwargs)
def lsf_create_job_file_factory(self, **kwargs):
# job file fectory config priority: kwargs > class defaults
kwargs = merge_dicts({}, self.lsf_job_file_factory_defaults, kwargs)
return LSFJobFileFactory(**kwargs)
def lsf_create_job_manager(self, **kwargs):
kwargs = merge_dicts(self.lsf_job_manager_defaults, kwargs)
return LSFJobManager(**kwargs)
def htcondor_create_job_file_factory(self, **kwargs):
# job file fectory config priority: kwargs > class defaults
kwargs = merge_dicts({}, self.htcondor_job_file_factory_defaults,
kwargs)
return HTCondorJobFileFactory(**kwargs)
def htcondor_create_job_manager(self, **kwargs):
kwargs = merge_dicts(self.htcondor_job_manager_defaults, kwargs)
return HTCondorJobManager(**kwargs)
def glite_create_job_manager(self, **kwargs):
kwargs = merge_dicts(self.glite_job_manager_defaults, kwargs)
return GLiteJobManager(**kwargs)
def glite_create_job_file_factory(self, **kwargs):
# job file fectory config priority: kwargs > class defaults
kwargs = merge_dicts({}, self.glite_job_file_factory_defaults, kwargs)
return self.glite_job_file_factory_cls()(**kwargs)
def glite_create_job_manager(self, **kwargs):
kwargs = merge_dicts(self.glite_job_manager_defaults, kwargs)
return self.glite_job_manager_cls()(**kwargs)
