def submit_jobs(self, job_files):
task = self.task
pool = get_param(task.htcondor_pool)
scheduler = get_param(task.htcondor_scheduler)
# prepare objects for dumping intermediate submission data
dump_freq = task.htcondor_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_ids):
job_num = i + 1
job_id = job_ids[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()
return self.job_manager.submit_batch(job_files,
pool=pool,
scheduler=scheduler,
retries=3,
threads=task.threads,
callback=progress_callback)
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 submit_jobs(self, job_files):
task = self.task
# delegate the voms proxy to all endpoints
if self.delegation_ids is None and callable(task.glite_delegate_proxy):
self.delegation_ids = []
for ce in task.glite_ce:
endpoint = get_ce_endpoint(ce)
self.delegation_ids.append(task.glite_delegate_proxy(endpoint))
# prepare objects for dumping intermediate submission data
dump_freq = task.glite_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
# 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()
return self.job_manager.submit_batch(job_files,
ce=task.glite_ce,
delegation_id=self.delegation_ids,
retries=3,
threads=task.threads,
callback=progress_callback)