def delete_s3_dir(s3_path, skip_confirm=False):
from cosmos.util.helpers import confirm
if not skip_confirm:
confirm("Are you sure you want to delete %s?" % s3_path)
cmd = "aws s3 rm %s --recursive --only-show-errors" % s3_path
print >>sys.stderr, cmd
run(cmd)
def delete_s3_dir(s3_path, skip_confirm=False):
from cosmos.util.helpers import confirm
if not skip_confirm:
confirm("Are you sure you want to delete %s?" % s3_path)
cmd = 'aws s3 rm %s --recursive --only-show-errors' % s3_path
print >> sys.stderr, cmd
run(cmd)
def start(self,
name,
restart=False,
skip_confirm=False,
primary_log_path='workflow.log',
fail_fast=False):
"""
Start, resume, or restart an workflow based on its name. If resuming, deletes failed tasks.
:param str name: A name for the workflow. Must be unique for this Cosmos session.
:param bool restart: If True and the workflow exists, delete it first.
:param bool skip_confirm: (If True, do not prompt the shell for input before deleting workflows or files.
:param str primary_log_path: The path of the primary log to write to. If None, does not write to a file. Log information is always printed to stderr.
:param bool fail_fast: If True, terminate the workflow the first time a Task fails.
:param int default_max_attempts: The default maximum number of times to attempt a Task.
Otherwise, run all Tasks except those downstream of a failure.
:rtype Workflow:
:returns: An Workflow instance.
"""
from .Workflow import Workflow
assert os.path.exists(
os.getcwd()
), "The current working dir of this environment, %s, does not exist" % os.getcwd(
)
# output_dir = os.path.abspath(output_dir)
# output_dir = output_dir if output_dir[-1] != '/' else output_dir[0:] # remove trailing slash
# prefix_dir = os.path.split(output_dir)[0]
# assert os.path.exists(prefix_dir), '%s does not exist' % prefix_dir
from ..util.helpers import mkdir
# assert isinstance(primary_log_path, basestring) and len(primary_log_path) > 0, 'invalid parimary log path'
if primary_log_path is not None and os.path.dirname(primary_log_path):
mkdir(os.path.dirname(primary_log_path))
session = self.session
old_id = None
if restart:
wf = session.query(Workflow).filter_by(name=name).first()
if wf:
old_id = wf.id
msg = 'Restarting %s. Are you sure you want to delete the all sql records?' % wf
if not skip_confirm and not confirm(msg):
raise SystemExit('Quitting')
wf.delete(delete_files=False)
else:
if not skip_confirm and not confirm(
'Workflow with name %s does not exist, '
'but `restart` is set to True. '
'Continue by starting a new Workflow?' % name):
raise SystemExit('Quitting')
# resuming?
wf = session.query(Workflow).filter_by(name=name).first()
# msg = 'Workflow started, Cosmos v%s' % __version__
if wf:
# resuming.
if not skip_confirm and not confirm(
'Resuming %s. All non-successful jobs will be deleted, '
'then any new tasks in the graph will be added and executed. '
'Are you sure?' % wf):
raise SystemExit('Quitting')
# assert ex.cwd == output_dir, 'cannot change the output_dir of an workflow being resumed.'
wf.successful = False
wf.finished_on = None
wf.status = WorkflowStatus.resuming
# if not os.path.exists(wf.output_dir):
# raise IOError('output_directory %s does not exist, cannot resume %s' % (wf.output_dir, wf))
wf.log.info('Resuming %s' % wf)
session.add(wf)
failed_tasks = [
t for s in wf.stages for t in s.tasks if not t.successful
]
n = len(failed_tasks)
if n:
wf.log.info(
'Deleting %s unsuccessful task(s) from SQL database, delete_files=%s'
% (n, False))
for t in failed_tasks:
session.delete(t)
for stage in filter(lambda s: len(s.tasks) == 0, wf.stages):
wf.log.info(
'Deleting stage %s, since it has 0 successful Tasks' %
stage)
session.delete(stage)
else:
# start from scratch
# if check_output_dir:
# assert not os.path.exists(output_dir), 'Workflow.output_dir `%s` already exists.' % (output_dir)
wf = Workflow(id=old_id,
name=name,
manual_instantiation=False,
successful=False)
# mkdir(output_dir) # make it here so we can start logging to logfile
session.add(wf)
wf.info['last_cmd_executed'] = get_last_cmd_executed()
wf.info['cwd'] = os.getcwd()
wf.info['fail_fast'] = fail_fast
wf.primary_log_path = primary_log_path
wf.log.info('Committing SQL session...')
session.commit()
session.expunge_all()
session.add(wf)
wf.log.info('Execution Command: %s' % get_last_cmd_executed())
wf.cosmos_app = self
return wf
def start(self, name, restart=False, skip_confirm=False, primary_log_path='workflow.log', fail_fast=False):
"""
Start, resume, or restart an workflow based on its name. If resuming, deletes failed tasks.
:param str name: A name for the workflow. Must be unique for this Cosmos session.
:param bool restart: If True and the workflow exists, delete it first.
:param bool skip_confirm: (If True, do not prompt the shell for input before deleting workflows or files.
:param str primary_log_path: The path of the primary log to write to. If None, does not write to a file. Log information is always printed to stderr.
:param bool fail_fast: If True, terminate the workflow the first time a Task fails.
:param int default_max_attempts: The default maximum number of times to attempt a Task.
Otherwise, run all Tasks except those downstream of a failure.
:rtype Workflow:
:returns: An Workflow instance.
"""
from .Workflow import Workflow
assert os.path.exists(
os.getcwd()), "The current working dir of this environment, %s, does not exist" % os.getcwd()
# output_dir = os.path.abspath(output_dir)
# output_dir = output_dir if output_dir[-1] != '/' else output_dir[0:] # remove trailing slash
# prefix_dir = os.path.split(output_dir)[0]
# assert os.path.exists(prefix_dir), '%s does not exist' % prefix_dir
from ..util.helpers import mkdir
# assert isinstance(primary_log_path, basestring) and len(primary_log_path) > 0, 'invalid parimary log path'
if primary_log_path is not None and os.path.dirname(primary_log_path):
mkdir(os.path.dirname(primary_log_path))
session = self.session
old_id = None
if restart:
wf = session.query(Workflow).filter_by(name=name).first()
if wf:
old_id = wf.id
msg = 'Restarting %s. Are you sure you want to delete the all sql records?' % wf
if not skip_confirm and not confirm(msg):
raise SystemExit('Quitting')
wf.delete(delete_files=False)
else:
if not skip_confirm and not confirm('Workflow with name %s does not exist, '
'but `restart` is set to True. '
'Continue by starting a new Workflow?' % name):
raise SystemExit('Quitting')
# resuming?
wf = session.query(Workflow).filter_by(name=name).first()
# msg = 'Workflow started, Cosmos v%s' % __version__
if wf:
# resuming.
if not skip_confirm and not confirm('Resuming %s. All non-successful jobs will be deleted, '
'then any new tasks in the graph will be added and executed. '
'Are you sure?' % wf):
raise SystemExit('Quitting')
# assert ex.cwd == output_dir, 'cannot change the output_dir of an workflow being resumed.'
wf.successful = False
wf.finished_on = None
wf.status = WorkflowStatus.resuming
# if not os.path.exists(wf.output_dir):
# raise IOError('output_directory %s does not exist, cannot resume %s' % (wf.output_dir, wf))
wf.log.info('Resuming %s' % wf)
session.add(wf)
failed_tasks = [t for s in wf.stages for t in s.tasks if not t.successful]
n = len(failed_tasks)
if n:
wf.log.info('Deleting %s unsuccessful task(s) from SQL database, delete_files=%s' % (n, False))
for t in failed_tasks:
session.delete(t)
for stage in filter(lambda s: len(s.tasks) == 0, wf.stages):
wf.log.info('Deleting stage %s, since it has 0 successful Tasks' % stage)
session.delete(stage)
else:
# start from scratch
# if check_output_dir:
# assert not os.path.exists(output_dir), 'Workflow.output_dir `%s` already exists.' % (output_dir)
wf = Workflow(id=old_id, name=name, manual_instantiation=False, successful=False)
# mkdir(output_dir) # make it here so we can start logging to logfile
session.add(wf)
wf.info['last_cmd_executed'] = get_last_cmd_executed()
wf.info['cwd'] = os.getcwd()
wf.info['fail_fast'] = fail_fast
wf.primary_log_path = primary_log_path
wf.log.info('Committing SQL session...')
session.commit()
session.expunge_all()
session.add(wf)
wf.log.info('Execution Command: %s' % get_last_cmd_executed())
wf.cosmos_app = self
return wf