def _do_stageout(job, xdata, activity, title):
"""
Use the `StageOutClient` in the Data API to perform stage-out.
:param job: job object.
:param xdata: list of FileSpec objects.
:param activity: copytool activity or preferred list of activities to resolve copytools
:param title: type of stage-out (output, log) (string).
:return: True in case of success transfers
"""
log = get_logger(job.jobid)
log.info('prepare to stage-out %d %s file(s)' % (len(xdata), title))
event_type = "put_sm"
#if log_transfer:
# eventType += '_logs'
#if special_log_transfer:
# eventType += '_logs_os'
if job.is_analysis():
event_type += "_a"
rse = get_rse(xdata)
localsite = remotesite = rse
trace_report = TraceReport(pq=os.environ.get('PILOT_SITENAME', ''), localSite=localsite, remoteSite=remotesite, dataset="", eventType=event_type)
trace_report.init(job)
try:
client = StageOutClient(job.infosys, logger=log, trace_report=trace_report)
kwargs = dict(workdir=job.workdir, cwd=job.workdir, usecontainer=False, job=job) #, mode='stage-out')
# prod analy unification: use destination preferences from PanDA server for unified queues
if job.infosys.queuedata.type != 'unified':
client.prepare_destinations(xdata, activity) ## FIX ME LATER: split activities: for astorages and for copytools (to unify with ES workflow)
client.transfer(xdata, activity, **kwargs)
except PilotException as error:
import traceback
error_msg = traceback.format_exc()
log.error(error_msg)
msg = errors.format_diagnostics(error.get_error_code(), error_msg)
job.piloterrorcodes, job.piloterrordiags = errors.add_error_code(error.get_error_code(), msg=msg)
except Exception:
import traceback
log.error(traceback.format_exc())
# do not raise the exception since that will prevent also the log from being staged out
# error = PilotException("stageOut failed with error=%s" % e, code=ErrorCodes.STAGEOUTFAILED)
else:
log.debug('stage-out client completed')
log.info('summary of transferred files:')
for e in xdata:
if not e.status:
status = "(not transferred)"
else:
status = e.status
log.info(" -- lfn=%s, status_code=%s, status=%s" % (e.lfn, e.status_code, status))
remain_files = [e for e in xdata if e.status not in ['transferred']]
log.debug('remain_files=%s' % str(remain_files))
log.debug('xdata=%s' % str(xdata))
return not remain_files
def __init__(self,
infosys_instance=None,
acopytools=None,
logger=None,
default_copytools='rucio',
trace_report=None):
"""
If `acopytools` is not specified then it will be automatically resolved via infosys. In this case `infosys` requires initialization.
:param acopytools: dict of copytool names per activity to be used for transfers. Accepts also list of names or string value without activity passed.
:param logger: logging.Logger object to use for logging (None means no logging)
:param default_copytools: copytool name(s) to be used in case of unknown activity passed. Accepts either list of names or single string value.
"""
super(StagingClient, self).__init__()
if not logger:
logger = logging.getLogger('%s.%s' % (__name__, 'null'))
logger.disabled = True
self.logger = logger
self.infosys = infosys_instance or infosys
if isinstance(acopytools, basestring):
acopytools = {'default': [acopytools]} if acopytools else {}
if isinstance(acopytools, (list, tuple)):
acopytools = {'default': acopytools} if acopytools else {}
self.acopytools = acopytools or {}
if self.infosys.queuedata:
if not self.acopytools: ## resolve from queuedata.acopytools using infosys
self.acopytools = (self.infosys.queuedata.acopytools
or {}).copy()
if not self.acopytools: ## resolve from queuedata.copytools using infosys
self.acopytools = dict(
default=(self.infosys.queuedata.copytools or {}).keys())
if not self.acopytools.get('default'):
if isinstance(default_copytools, basestring):
default_copytools = [default_copytools
] if default_copytools else []
self.acopytools['default'] = default_copytools
if not self.acopytools:
msg = 'failed to initilize StagingClient: no acopytools options found, acopytools=%s' % self.acopytools
logger.error(msg)
self.trace_report.update(clientState='BAD_COPYTOOL',
stateReason=msg)
self.trace_report.send()
raise PilotException("failed to resolve acopytools settings")
logger.info('configured copytools per activity: acopytools=%s' %
self.acopytools)
# get an initialized trace report (has to be updated for get/put if not defined before)
self.trace_report = trace_report if trace_report else TraceReport(
pq=os.environ.get('PILOT_SITENAME', ''))
def create_trace_report(job, label='stage-in'):
"""
Create the trace report object.
:param job: job object.
:param label: 'stage-[in|out]' (string).
:return: trace report object.
"""
event_type, localsite, remotesite = get_trace_report_variables(job, label=label)
trace_report = TraceReport(pq=os.environ.get('PILOT_SITENAME', ''), localSite=localsite, remoteSite=remotesite,
dataset="", eventType=event_type)
trace_report.init(job)
return trace_report
# scopes = file_list_dictionary.get('scopes')
# filesizes = file_list_dictionary.get('filesizes')
# checksums = file_list_dictionary.get('checksums')
# allowlans = file_list_dictionary.get('allowlans')
# allowwans = file_list_dictionary.get('allowwans')
# directaccesslans = file_list_dictionary.get('directaccesslans')
# directaccesswans = file_list_dictionary.get('directaccesswans')
# istars = file_list_dictionary.get('istars')
# accessmodes = file_list_dictionary.get('accessmodes')
# storagetokens = file_list_dictionary.get('storagetokens')
# guids = file_list_dictionary.get('guids')
# generate the trace report
trace_report = TraceReport(pq=os.environ.get('PILOT_SITENAME', ''),
localSite=args.localsite,
remoteSite=args.remotesite,
dataset="",
eventType=args.eventtype)
job = Job(produserid=args.produserid,
jobid=args.jobid,
taskid=args.taskid,
jobdefinitionid=args.jobdefinitionid)
trace_report.init(job)
try:
infoservice = InfoService()
infoservice.init(args.queuename, infosys.confinfo, infosys.extinfo)
infosys.init(
args.queuename
) # is this correct? otherwise infosys.queuedata doesn't get set
except Exception as e:
def _stage_in(args, job):
"""
:return: True in case of success
"""
log = get_logger(job.jobid)
# tested ok:
#log.info('testing sending SIGUSR1')
#import signal
#os.kill(os.getpid(), signal.SIGUSR1)
# write time stamps to pilot timing file
add_to_pilot_timing(job.jobid, PILOT_PRE_STAGEIN, time.time(), args)
# any DBRelease files should not be staged in
for fspec in job.indata:
if 'DBRelease' in fspec.lfn:
fspec.status = 'no_transfer'
event_type = "get_sm"
#if log_transfer:
# eventType += '_logs'
#if special_log_transfer:
# eventType += '_logs_os'
if job.is_analysis():
event_type += "_a"
rse = get_rse(job.indata)
localsite = remotesite = rse
trace_report = TraceReport(pq='', localSite=localsite, remoteSite=remotesite, dataset="", eventType=event_type)
trace_report.init(job)
# now that the trace report has been created, remove any files that are not to be transferred (DBRelease files) from the indata list
toberemoved = []
for fspec in job.indata:
if fspec.status == 'no_transfer':
toberemoved.append(fspec)
for fspec in toberemoved:
logger.info('removing fspec object (lfn=%s) from list of input files' % fspec.lfn)
job.indata.remove(fspec)
try:
if job.is_eventservicemerge:
client = StageInESClient(job.infosys, logger=log, trace_report=trace_report)
activity = 'es_events_read'
else:
client = StageInClient(job.infosys, logger=log, trace_report=trace_report)
activity = 'pr'
kwargs = dict(workdir=job.workdir, cwd=job.workdir, usecontainer=False, job=job) #, mode='stage-in')
client.transfer(job.indata, activity=activity, **kwargs)
except PilotException as error:
log.error('PilotException caught: %s' % error)
job.piloterrorcodes, job.piloterrordiags = errors.add_error_code(error.get_error_code())
except Exception as error:
log.error('failed to stage-in: error=%s' % error)
log.info('summary of transferred files:')
for e in job.indata:
if not e.status:
status = "(not transferred)"
else:
status = e.status
log.info(" -- lfn=%s, status_code=%s, status=%s" % (e.lfn, e.status_code, status))
# write time stamps to pilot timing file
add_to_pilot_timing(job.jobid, PILOT_POST_STAGEIN, time.time(), args)
remain_files = [e for e in job.indata if e.status not in ['remote_io', 'transferred', 'no_transfer']]
if not remain_files:
log.info("stage-in finished")
else:
log.info("stage-in failed")
return not remain_files
def _stage_in(args, job):
"""
:return: True in case of success
"""
log = get_logger(job.jobid)
# tested ok:
#log.info('testing sending SIGUSR1')
#import signal
#os.kill(os.getpid(), signal.SIGUSR1)
# write time stamps to pilot timing file
add_to_pilot_timing(job.jobid, PILOT_PRE_STAGEIN, time.time(), args)
# any DBRelease files should not be staged in
for fspec in job.indata:
if 'DBRelease' in fspec.lfn:
fspec.status = 'no_transfer'
event_type = "get_sm"
if job.is_analysis():
event_type += "_a"
rse = get_rse(job.indata)
localsite = remotesite = rse
trace_report = TraceReport(pq=os.environ.get('PILOT_SITENAME', ''),
localSite=localsite,
remoteSite=remotesite,
dataset="",
eventType=event_type)
trace_report.init(job)
# now that the trace report has been created, remove any files that are not to be transferred (DBRelease files) from the indata list
toberemoved = []
for fspec in job.indata:
if fspec.status == 'no_transfer':
toberemoved.append(fspec)
for fspec in toberemoved:
logger.info('removing fspec object (lfn=%s) from list of input files' %
fspec.lfn)
job.indata.remove(fspec)
########### bulk transfer test
# THE FOLLOWING WORKS BUT THERE IS AN ISSUE WITH TRACES, CHECK STAGEIN SCRIPT IF STORED CORRECTLY
#filename = 'initial_trace_report.json'
#tpath = os.path.join(job.workdir, filename)
#write_json(tpath, trace_report)
#lfns, scopes = get_filedata_strings(job.indata)
#script = 'stagein.py'
#srcdir = os.environ.get('PILOT_SOURCE_DIR')
#scriptpath = os.path.join(os.path.join(srcdir, 'pilot/scripts'), script)
#copy(scriptpath, srcdir)
#cmd = 'python %s --lfns=%s --scopes=%s --tracereportname=%s -w %s -d -q %s' %\
# (os.path.join(srcdir, script), lfns, scopes, tpath, job.workdir, args.queue)
#logger.debug('could have executed: %s' % script)
#exit_code, stdout, stderr = execute(cmd, mode='python')
#logger.debug('exit_code=%d' % exit_code)
#logger.debug('stdout=%s' % stdout)
#logger.debug('stderr=%s' % stderr)
########### bulk transfer test
try:
if job.is_eventservicemerge:
client = StageInESClient(job.infosys,
logger=log,
trace_report=trace_report)
activity = 'es_events_read'
else:
client = StageInClient(job.infosys,
logger=log,
trace_report=trace_report)
activity = 'pr'
kwargs = dict(workdir=job.workdir,
cwd=job.workdir,
usecontainer=False,
job=job,
use_bulk=False)
client.prepare_sources(job.indata)
client.transfer(job.indata, activity=activity, **kwargs)
except PilotException as error:
import traceback
error_msg = traceback.format_exc()
log.error(error_msg)
msg = errors.format_diagnostics(error.get_error_code(), error_msg)
job.piloterrorcodes, job.piloterrordiags = errors.add_error_code(
error.get_error_code(), msg=msg)
except Exception as error:
log.error('failed to stage-in: error=%s' % error)
log.info('summary of transferred files:')
for e in job.indata:
status = e.status if e.status else "(not transferred)"
log.info(" -- lfn=%s, status_code=%s, status=%s" %
(e.lfn, e.status_code, status))
# write time stamps to pilot timing file
add_to_pilot_timing(job.jobid, PILOT_POST_STAGEIN, time.time(), args)
remain_files = [
e for e in job.indata
if e.status not in ['remote_io', 'transferred', 'no_transfer']
]
if not remain_files:
log.info("stage-in finished")
else:
log.info("stage-in failed")
return not remain_files
def test_copy_out_rucio(self):
from pilot.copytool.rucio import copy_out
trace_report = TraceReport()
trace_report.update(eventType='unit test')
copy_out(self.outdata, trace_report=trace_report)
os.remove(self.outdata[0].pfn)