def process_jobreport(payload_report_file, job_scratch_path,
job_communication_point):
"""
Copy job report file to make it accessible by Harvester. Shrink job report file.
:param payload_report_file: name of job report (string).
:param job_scratch_path: path to scratch directory (string).
:param job_communication_point: path to updated job report accessible by Harvester (string).
:raises FileHandlingFailure: in case of IOError.
"""
src_file = os.path.join(job_scratch_path, payload_report_file)
dst_file = os.path.join(job_communication_point, payload_report_file)
try:
logger.info("Copy of payload report [{0}] to access point: {1}".format(
payload_report_file, job_communication_point))
# shrink jobReport
job_report = read_json(src_file)
if 'executor' in job_report:
for executor in job_report['executor']:
if 'logfileReport' in executor:
executor['logfileReport'] = {}
write_json(dst_file, job_report)
except IOError:
logger.error("Job report copy failed, execution terminated': \n %s " %
(sys.exc_info()[1]))
raise FileHandlingFailure("Job report copy from RAM failed")
def write_file(path, contents, mute=True):
"""
Write the given contents to a file.
:param path: full path for file (string).
:param contents: file contents (string).
:param mute: boolean to control stdout info message
:raises PilotException: FileHandlingFailure.
:return: True if successful, otherwise False.
"""
status = False
f = open_file(path, 'w')
if f:
try:
f.write(contents)
except IOError as e:
raise FileHandlingFailure(e)
else:
status = True
f.close()
if not mute:
logger.info('created file: %s' % path)
return status
def write_json(filename,
data,
sort_keys=True,
indent=4,
separators=(',', ': ')):
"""
Write the dictionary to a JSON file.
:param filename: file name (string).
:param data: object to be written to file (dictionary or list).
:param sort_keys: should entries be sorted? (boolean).
:param indent: indentation level, default 4 (int).
:param separators: field separators (default (',', ': ') for dictionaries, use e.g. (',\n') for lists) (tuple)
:raises PilotException: FileHandlingFailure.
:return: status (boolean).
"""
status = False
try:
with open(filename, 'w') as fh:
dumpjson(data,
fh,
sort_keys=sort_keys,
indent=indent,
separators=separators)
except IOError as exc:
raise FileHandlingFailure(exc)
else:
status = True
return status
def postprocess_workdir(workdir):
"""
Post-processing of working directory. Unlink paths.
:param workdir: path to directory to be processed (string).
:raises FileHandlingFailure: in case of IOError.
"""
pseudo_dir = "poolcond"
try:
if os.path.exists(pseudo_dir):
remove(os.path.join(workdir, pseudo_dir))
except IOError:
raise FileHandlingFailure(
"Post processing of working directory failed")
def copy_output(job, job_scratch_dir, work_dir):
cp_start = time.time()
try:
for outfile in job.output_files.keys():
if os.path.exists(outfile):
copy(os.path.join(job_scratch_dir, outfile),
os.path.join(work_dir, outfile))
os.chdir(work_dir)
except IOError:
raise FileHandlingFailure(
"Copy from scratch dir to access point failed")
finally:
cp_time = time.time() - cp_start
logger.info("Copy of outputs took: {0} sec.".format(cp_time))
return 0
def open_file(filename, mode):
"""
Open and return a file pointer for the given mode.
Note: the caller needs to close the file.
:param filename: file name (string).
:param mode: file mode (character).
:raises PilotException: FileHandlingFailure.
:return: file pointer.
"""
f = None
try:
f = open(filename, mode)
except IOError as exc:
raise FileHandlingFailure(exc)
return f
def move(path1, path2):
"""
Move a file from path1 to path2.
:param path1: source path (string).
:param path2: destination path2 (string).
"""
if not os.path.exists(path1):
logger.warning('file copy failure: path does not exist: %s', path1)
raise NoSuchFile("File does not exist: %s" % path1)
try:
import shutil
shutil.move(path1, path2)
except IOError as exc:
logger.warning("exception caught during file move: %s", exc)
raise FileHandlingFailure(exc)
else:
logger.info("moved %s to %s", path1, path2)
def copy(path1, path2):
"""
Copy path1 to path2.
:param path1: file path (string).
:param path2: file path (string).
:raises PilotException: FileHandlingFailure, NoSuchFile
:return:
"""
if not os.path.exists(path1):
logger.warning('file copy failure: path does not exist: %s', path1)
raise NoSuchFile("File does not exist: %s" % path1)
try:
copy2(path1, path2)
except IOError as exc:
logger.warning("exception caught during file copy: %s", exc)
raise FileHandlingFailure(exc)
else:
logger.info("copied %s to %s", path1, path2)
def open_file(filename, mode):
"""
Open and return a file pointer for the given mode.
Note: the caller needs to close the file.
:param filename: file name (string).
:param mode: file mode (character).
:raises PilotException: FileHandlingFailure.
:return: file pointer.
"""
f = None
if not mode == 'w' and not os.path.exists(filename):
raise NoSuchFile("File does not exist: %s" % filename)
try:
f = open(filename, mode)
except IOError as e:
raise FileHandlingFailure(e)
return f
def calculate_checksum(filename, algorithm='adler32'):
"""
Calculate the checksum value for the given file.
The default algorithm is adler32. Md5 is also be supported.
Valid algorithms are 1) adler32/adler/ad32/ad, 2) md5/md5sum/md.
:param filename: file name (string).
:param algorithm: optional algorithm string.
:raises FileHandlingFailure, NotImplementedError: exception raised when file does not exist or for unknown algorithm.
:return: checksum value (string).
"""
if not os.path.exists(filename):
raise FileHandlingFailure('file does not exist: %s' % filename)
if algorithm == 'adler32' or algorithm == 'adler' or algorithm == 'ad' or algorithm == 'ad32':
return calculate_adler32_checksum(filename)
elif algorithm == 'md5' or algorithm == 'md5sum' or algorithm == 'md':
return calculate_md5_checksum(filename)
else:
msg = 'unknown checksum algorithm: %s' % algorithm
logger.warning(msg)
raise NotImplementedError()
def write_file(path, contents, mute=True, mode='w', unique=False):
"""
Write the given contents to a file.
If unique=True, then if the file already exists, an index will be added (e.g. 'out.txt' -> 'out-1.txt')
:param path: full path for file (string).
:param contents: file contents (object).
:param mute: boolean to control stdout info message.
:param mode: file mode (e.g. 'w', 'r', 'a', 'wb', 'rb') (string).
:param unique: file must be unique (Boolean).
:raises PilotException: FileHandlingFailure.
:return: True if successful, otherwise False.
"""
status = False
# add an incremental file name (add -%d if path already exists) if necessary
if unique:
path = get_nonexistant_path(path)
f = open_file(path, mode)
if f:
try:
f.write(contents)
except IOError as exc:
raise FileHandlingFailure(exc)
else:
status = True
f.close()
if not mute:
if 'w' in mode:
logger.info('created file: %s', path)
if 'a' in mode:
logger.info('appended file: %s', path)
return status
def set_scratch_workdir(job, work_dir, args):
"""
Copy input files and some db files to RAM disk.
:param job: job object.
:param work_dir: job working directory (permanent FS) (string).
:param args: args dictionary to collect timing metrics.
:return: job working directory in scratch (string).
"""
scratch_path = config.HPC.scratch
du = disk_usage(scratch_path)
logger.info("Scratch dir available space: {0} used: {1}".format(
du.free, du.used))
job_scratch_dir = os.path.join(scratch_path, str(job.jobid))
for inp_file in job.input_files:
job.input_files[inp_file]["scratch_path"] = job_scratch_dir
logger.debug("Job scratch path: {0}".format(job_scratch_dir))
# special data, that should be preplaced in RAM disk
dst_db_path = 'sqlite200/'
dst_db_filename = 'ALLP200.db'
dst_db_path_2 = 'geomDB/'
dst_db_filename_2 = 'geomDB_sqlite'
tmp_path = 'tmp/'
src_file = '/ccs/proj/csc108/AtlasReleases/21.0.15/DBRelease/current/sqlite200/ALLP200.db'
src_file_2 = '/ccs/proj/csc108/AtlasReleases/21.0.15/DBRelease/current/geomDB/geomDB_sqlite'
if os.path.exists(scratch_path):
try:
add_to_pilot_timing(job.jobid, PILOT_PRE_STAGEIN, time.time(),
args)
logger.debug("Prepare \'tmp\' dir in scratch ")
if not os.path.exists(scratch_path + tmp_path):
os.makedirs(scratch_path + tmp_path)
logger.debug("Prepare dst and copy sqlite db files")
t0 = time.time()
if not os.path.exists(scratch_path + dst_db_path):
os.makedirs(scratch_path + dst_db_path)
shutil.copyfile(src_file,
scratch_path + dst_db_path + dst_db_filename)
logger.debug("")
sql_cp_time = time.time() - t0
logger.debug("Copy of sqlite files took: {0}".format(sql_cp_time))
logger.debug("Prepare dst and copy geomDB files")
t0 = time.time()
if not os.path.exists(scratch_path + dst_db_path_2):
os.makedirs(scratch_path + dst_db_path_2)
shutil.copyfile(src_file_2,
scratch_path + dst_db_path_2 + dst_db_filename_2)
geomdb_cp_time = time.time() - t0
logger.debug(
"Copy of geomDB files took: {0} s".format(geomdb_cp_time))
logger.debug("Prepare job scratch dir")
t0 = time.time()
if not os.path.exists(job_scratch_dir):
os.makedirs(job_scratch_dir)
logger.debug("Copy input file")
for inp_file in job.input_files:
logger.debug("Copy: {0} to {1}".format(
os.path.join(work_dir, inp_file),
job.input_files[inp_file]["scratch_path"]))
shutil.copyfile(
os.path.join(work_dir, inp_file),
os.path.join(job.input_files[inp_file]["scratch_path"],
inp_file))
input_cp_time = time.time() - t0
logger.debug(
"Copy of input files took: {0} s".format(input_cp_time))
except IOError as e:
logger.error("I/O error({0}): {1}".format(e.errno, e.strerror))
logger.error(
"Copy to scratch failed, execution terminated': \n %s " %
(sys.exc_info()[1]))
raise FileHandlingFailure("Copy to RAM disk failed")
finally:
add_to_pilot_timing(job.jobid, PILOT_POST_STAGEIN, time.time(),
args)
else:
logger.info('Scratch directory (%s) dos not exist' % scratch_path)
return work_dir
os.chdir(job_scratch_dir)
logger.debug("Current directory: {0}".format(os.getcwd()))
true_dir = '/ccs/proj/csc108/AtlasReleases/21.0.15/nfs_db_files'
pseudo_dir = "./poolcond"
os.symlink(true_dir, pseudo_dir)
du = disk_usage(scratch_path)
logger.info("Scratch dir available space for job: {0} used: {1}".format(
du.free, du.used))
return job_scratch_dir
def run(args):
"""
Main execution function for the generic HPC workflow.
:param args: pilot arguments.
:returns: traces object.
"""
# set communication point. Worker report should be placed there, matched with working directory of Harvester
if args.harvester_workdir:
communication_point = args.harvester_workdir
else:
communication_point = os.getcwd()
work_report = get_initial_work_report()
worker_attributes_file = config.Harvester.workerAttributesFile
worker_stageout_declaration = config.Harvester.StageOutnFile
payload_report_file = config.Payload.jobreport
payload_stdout_file = config.Payload.payloadstdout
payload_stderr_file = config.Payload.payloadstderr
try:
logger.info('setting up signal handling')
signal.signal(signal.SIGINT, functools.partial(interrupt, args))
logger.info('setting up tracing')
traces = namedtuple('traces', ['pilot'])
traces.pilot = {'state': SUCCESS, 'nr_jobs': 0}
if args.hpc_resource == '':
logger.critical('hpc resource not specified, cannot continue')
traces.pilot['state'] = FAILURE
return traces
# get the resource reference
resource = __import__('pilot.resource.%s' % args.hpc_resource,
globals(), locals(), [args.hpc_resource], -1)
# get the user reference
user = __import__('pilot.user.%s.common' % args.pilot_user.lower(),
globals(), locals(), [args.pilot_user.lower()], -1)
# get job (and rank)
add_to_pilot_timing('0', PILOT_PRE_GETJOB, time.time(), args)
job, rank = resource.get_job(communication_point)
add_to_pilot_timing(job.jobid, PILOT_POST_GETJOB, time.time(), args)
# cd to job working directory
add_to_pilot_timing(job.jobid, PILOT_PRE_SETUP, time.time(), args)
work_dir = resource.set_job_workdir(job, communication_point)
work_report['workdir'] = work_dir
worker_attributes_file = os.path.join(work_dir, worker_attributes_file)
logger.debug("Worker attributes will be publeshied in: {0}".format(
worker_attributes_file))
set_pilot_state(job=job, state="starting")
work_report["jobStatus"] = job.state
publish_work_report(work_report, worker_attributes_file)
# Get HPC specific setup commands
logger.info('setup for resource %s: %s' %
(args.hpc_resource, str(resource.get_setup())))
setup_str = "; ".join(resource.get_setup())
# Prepare job scratch directory (RAM disk etc.)
job_scratch_dir = resource.set_scratch_workdir(job, work_dir, args)
my_command = " ".join([job.script, job.script_parameters])
my_command = resource.command_fix(my_command, job_scratch_dir)
my_command = setup_str + my_command
add_to_pilot_timing(job.jobid, PILOT_POST_SETUP, time.time(), args)
# Basic execution. Should be replaced with something like 'run_payload'
logger.debug("Going to launch: {0}".format(my_command))
logger.debug("Current work directory: {0}".format(job_scratch_dir))
payloadstdout = open(payload_stdout_file, "w")
payloadstderr = open(payload_stderr_file, "w")
add_to_pilot_timing(job.jobid, PILOT_PRE_PAYLOAD, time.time(), args)
set_pilot_state(job=job, state="running")
work_report["jobStatus"] = job.state
work_report["startTime"] = datetime.utcnow().strftime(
"%Y-%m-%d %H:%M:%S")
start_time = time.asctime(time.localtime(time.time()))
job.startTime = datetime.utcnow().strftime("%Y-%m-%d %H:%M:%S")
publish_work_report(work_report, worker_attributes_file)
stime = time.time()
t0 = os.times()
exit_code, stdout, stderr = execute(my_command,
stdout=payloadstdout,
stderr=payloadstderr,
shell=True)
logger.debug("Payload exit code: {0}".format(exit_code))
t1 = os.times()
exetime = time.time() - stime
end_time = time.asctime(time.localtime(time.time()))
t = map(lambda x, y: x - y, t1, t0)
t_tot = reduce(lambda x, y: x + y, t[2:3])
job.endTime = datetime.utcnow().strftime("%Y-%m-%d %H:%M:%S")
payloadstdout.close()
payloadstderr.close()
add_to_pilot_timing(job.jobid, PILOT_POST_PAYLOAD, time.time(), args)
state = 'finished' if exit_code == 0 else 'failed'
set_pilot_state(job=job, state=state)
job.exitcode = exit_code
work_report["startTime"] = job.startTime
work_report["endTime"] = job.endTime
work_report["jobStatus"] = job.state
work_report["cpuConsumptionTime"] = t_tot
work_report["transExitCode"] = job.exitcode
log_jobreport = "\nPayload exit code: {0} JobID: {1} \n".format(
exit_code, job.jobid)
log_jobreport += "CPU comsumption time: {0} JobID: {1} \n".format(
t_tot, job.jobid)
log_jobreport += "Start time: {0} JobID: {1} \n".format(
start_time, job.jobid)
log_jobreport += "End time: {0} JobID: {1} \n".format(
end_time, job.jobid)
log_jobreport += "Execution time: {0} sec. JobID: {1} \n".format(
exetime, job.jobid)
logger.info(log_jobreport)
log_jobreport = "\nJob report start time: {0}\nJob report end time: {1}".format(
job.startTime, job.endTime)
logger.debug(log_jobreport)
# Parse job report file and update of work report
if os.path.exists(payload_report_file):
payload_report = user.parse_jobreport_data(
read_json(payload_report_file))
work_report.update(payload_report)
resource.process_jobreport(payload_report_file, job_scratch_dir,
work_dir)
resource.postprocess_workdir(job_scratch_dir)
# output files should not be packed with logs
protectedfiles = job.output_files.keys()
# log file not produced (yet), so should be excluded
if job.log_file in protectedfiles:
protectedfiles.remove(job.log_file)
else:
logger.info("Log files was not declared")
logger.info("Cleanup of working directory")
protectedfiles.extend(
[worker_attributes_file, worker_stageout_declaration])
user.remove_redundant_files(job_scratch_dir, protectedfiles)
res = tar_files(job_scratch_dir, protectedfiles, job.log_file)
if res > 0:
raise FileHandlingFailure("Log file tar failed")
add_to_pilot_timing(job.jobid, PILOT_PRE_STAGEOUT, time.time(), args)
# Copy of output to shared FS for stageout
if not job_scratch_dir == work_dir:
copy_output(job, job_scratch_dir, work_dir)
add_to_pilot_timing(job.jobid, PILOT_POST_STAGEOUT, time.time(), args)
logger.info("Declare stage-out")
add_to_pilot_timing(job.jobid, PILOT_PRE_FINAL_UPDATE, time.time(),
args)
declare_output(job, work_report, worker_stageout_declaration)
logger.info("All done")
publish_work_report(work_report, worker_attributes_file)
traces.pilot['state'] = SUCCESS
logger.debug("Final report: {0}".format(work_report))
add_to_pilot_timing(job.jobid, PILOT_POST_FINAL_UPDATE, time.time(),
args)
except Exception as e:
work_report["jobStatus"] = "failed"
work_report["exitMsg"] = str(e)
publish_work_report(work_report, worker_attributes_file)
logging.exception('exception caught:')
traces.pilot['state'] = FAILURE
return traces