def _init_data_uri(self):
"""
Create output data URI for the source context (local).
Args:
self: class instance.
Returns:
On success: True.
On failure: False.
"""
# make sure the source data URI has a compatible scheme (local)
if self._parsed_data_uris[self._source_context]['scheme'] != 'local':
msg = 'invalid data uri scheme for this step: {}'.format(
self._parsed_data_uris[self._source_context]['scheme']
)
Log.an().error(msg)
return self._fatal(msg)
# delete old folder if it exists and clean==True
if (
DataManager.exists(
parsed_uri=self._parsed_data_uris[self._source_context]
)
and self._clean
):
if not DataManager.delete(
parsed_uri=self._parsed_data_uris[self._source_context]
):
Log.a().warning(
'cannot delete existing data uri: %s',
self._parsed_data_uris[self._source_context]['chopped_uri']
)
# create folder
if not DataManager.mkdir(
parsed_uri=self._parsed_data_uris[self._source_context],
recursive=True
):
msg = 'cannot create data uri: {}'.format(
self._parsed_data_uris[self._source_context]['chopped_uri']
)
Log.an().error(msg)
return self._fatal(msg)
# create _log folder
if not DataManager.mkdir(
uri='{}/_log'.format(
self._parsed_data_uris[self._source_context]['chopped_uri']
),
recursive=True
):
msg = 'cannot create _log folder in data uri: {}/_log'.format(
self._parsed_data_uris[self._source_context]['chopped_uri']
)
Log.an().error(msg)
return self._fatal(msg)
return True
def upload_agave_test_data(self):
"""
Upload Agave test data from workflow package.
Args:
self: class instance.
Returns:
None
"""
if (not self._agave or not self._agave_params
or not self._agave_params.get('agave')):
Log.a().warning(
'must provide agave parameters to upload test data')
return False
# create main test data URI
parsed_base_test_uri = URIParser.parse('agave://{}/{}'.format(
self._agave_params['agave']['deploymentSystem'],
self._agave_params['agave']['testDataDir']))
Log.some().info('creating base test data uri: %s',
parsed_base_test_uri['chopped_uri'])
if not DataManager.mkdir(parsed_uri=parsed_base_test_uri,
recursive=True,
agave={
'agave': self._agave,
'agave_config': self._config['agave']
}):
Log.a().warning('cannot create base test data uri: %s',
parsed_base_test_uri['chopped_uri'])
return False
# upload test data
parsed_local_test_uri = URIParser.parse(str(Path(self._path) / 'data'))
parsed_agave_test_uri = URIParser.parse('{}/{}'.format(
parsed_base_test_uri['chopped_uri'],
Path(self._path).name))
Log.some().info('copying test data from %s to %s',
parsed_local_test_uri['chopped_uri'],
parsed_agave_test_uri['chopped_uri'])
if not DataManager.copy(parsed_src_uri=parsed_local_test_uri,
parsed_dest_uri=parsed_agave_test_uri,
local={},
agave={
'agave': self._agave,
'agave_config': self._config['agave']
}):
Log.a().warning('cannot copy test data from %s to %s',
parsed_local_test_uri['chopped_uri'],
parsed_agave_test_uri['chopped_uri'])
return False
return True
def _create_job_uris(self):
"""
Create all work and output URIs.
Args:
self: class instance
Returns:
On success: True.
On failure: False.
"""
# create work URIs. a work URI is required for each workflow context
for context in {
Contexts.mapping[exec_context]['data_scheme']
for exec_context in self._exec_contexts
}:
if not DataManager.mkdir(
parsed_uri=self._parsed_job_work_uri[context],
recursive=True,
**{
context: self._workflow_context[context]\
.get_context_options()
}
):
msg = 'cannot create job work uri for context: {}->{}'.format(
context, self._parsed_job_work_uri[context]['chopped_uri']
)
Log.an().error(msg)
return self._fatal(msg)
# create output URI. output URI scheme must be in the set of data contexts
output_context = self._parsed_job_output_uri['scheme']
if output_context not in self._data_contexts:
msg = 'invalid output context: {}'.format(output_context)
Log.an().error(msg)
return self._fatal(msg)
if not DataManager.mkdir(
parsed_uri=self._parsed_job_output_uri,
recursive=True,
**{
output_context: self._workflow_context[output_context]\
.get_context_options()
}
):
msg = 'cannot create job output uri: {}'.format(
self._parsed_job_output_uri['chopped_uri']
)
Log.an().error(msg)
return self._fatal(msg)
return True
def _create_job_uris(self):
"""
Create all work and output URIs.
Args:
self: class instance
Returns:
On success: True.
On failure: False.
"""
# create work URI for each context
for context in self._job['work_uri']:
if not DataManager.mkdir(
parsed_uri=self._parsed_job_work_uri[context],
recursive=True,
**{
context: self._workflow_context[context]\
.get_context_options()
}
):
msg = 'cannot create job work uri for context: {}->{}'.format(
context, self._parsed_job_work_uri[context]['chopped_uri'])
Log.an().error(msg)
return self._fatal(msg)
# create output URI
output_context = self._parsed_job_output_uri['scheme']
if output_context not in self._workflow_context:
msg = 'invalid output context: {}'.format(output_context)
Log.an().error(msg)
return self._fatal(msg)
if not DataManager.mkdir(
parsed_uri=self._parsed_job_output_uri,
recursive=True,
**{
output_context: self._workflow_context[output_context]\
.get_context_options()
}
):
msg = 'cannot create job output uri: {}'.format(
self._parsed_job_output_uri['chopped_uri'])
Log.an().error(msg)
return self._fatal(msg)
return True
def _get_map_uri_list(self):
"""
Get the contents of the map URI (agave URI).
Args:
self: class instance.
Returns:
Array of base file names in the map URI. Returns False on
exception.
"""
# make sure map URI is compatible scheme (agave)
if self._parsed_map_uri['scheme'] != 'agave':
msg = 'invalid map uri scheme for this step: {}'.format(
self._parsed_map_uri['scheme'])
Log.an().error(msg)
return self._fatal(msg)
# list files from URI
file_list = DataManager.list(parsed_uri=self._parsed_map_uri,
agave=self._agave)
if file_list is False:
msg = 'cannot get contents of map uri: {}'\
.format(self._parsed_map_uri['chopped_uri'])
Log.an().error(msg)
return self._fatal(msg)
return file_list
def _init_archive_uri(self):
"""
Initialize and validate Agave job archive URI.
Args:
None.
Returns:
On success: True.
On failure: False.
"""
if 'agave' not in self._parsed_job_work_uri:
Log.an().error('job work uri must include an agave context')
return False
# construct archive URI
self._parsed_archive_uri = URIParser.parse('{}/_agave_jobs'.format(
self._parsed_job_work_uri['agave']['chopped_uri']))
if not self._parsed_archive_uri:
Log.an().error('invalid job work uri: %s',
self._parsed_job_work_uri['agave'])
return False
# create URI
if not DataManager.mkdir(parsed_uri=self._parsed_archive_uri,
recursive=True,
agave=self.get_context_options()):
Log.an().error('cannot create agave archive uri: %s',
self._parsed_archive_uri['chopped_uri'])
return False
return True
def _init_app_paths(self):
"""
Add app paths to environment PATH for local workflows.
The package path contains the workflow definition YAML file and shell
scripts for calling individual apps used in a workflow.
Args:
None.
Output:
On success: True.
On failure: False.
"""
parsed_uri = URIParser.parse(self._workflow_path)
if not parsed_uri:
Log.an().error('invalid workflow path: %s', self._workflow_path)
return False
apps_uri = ('{}{}' if parsed_uri['folder'] == '/' else '{}/{}')\
.format(parsed_uri['folder'], 'apps')
parsed_apps_uri = URIParser.parse(
('{}{}' if parsed_uri['folder'] == '/' else '{}/{}')\
.format(parsed_uri['folder'], 'apps')
)
if not parsed_apps_uri:
Log.an().error('cannot construct apps uri: %s', apps_uri)
return False
if not DataManager.exists(parsed_uri=parsed_apps_uri):
# no apps directory
return True
for app_dir in DataManager.list(parsed_uri=parsed_apps_uri):
try:
os.environ['PATH'] = '{}{}{}'.format(
os.path.join(parsed_apps_uri['chopped_path'], app_dir,
'assets'), os.pathsep, os.environ['PATH'])
except OSError as err:
Log.an().error('workflow app pathmunge error [%s]', str(err))
return False
return True
def stage(self, **kwargs):
"""
Copy data to all contexts except 'final' from source URI. Source URI can be multiple
locations, but only copy to the first element of dest URIs.
Set _staged indicator to True on success.
Args:
self: class instance.
**kwargs: additional arguments required by DataManager.copy().
Returns:
True or False.
"""
for context in self._parsed_data_uris:
if context != self._source_context:
if self._clean:
# remove target URI first
pass
for i, parsed_source_uri in enumerate(
self._parsed_data_uris[self._source_context]):
Log.some().debug(
'staging data: %s->%s to %s->%s', self._source_context,
parsed_source_uri['chopped_uri'], context,
self._parsed_data_uris[context][i]['chopped_uri'])
if context != 'final':
if not DataManager.copy(
parsed_src_uri=parsed_source_uri,
parsed_dest_uri=self._parsed_data_uris[context]
[i],
**kwargs):
msg = 'cannot stage data by copying from {} to {}'.format(
parsed_source_uri['chopped_uri'],
self._parsed_data_uris[context][i]
['chopped_uri'])
Log.an().error(msg)
return self._fatal(msg)
self._staged = True
return True
def stage(self, **kwargs):
"""
Copy data to all contexts except 'final' from source URI.
Set _staged indicator to True on success.
Args:
self: class instance.
**kwargs: additional arguments required by DataManager.copy().
Returns:
True or False.
"""
for context in self._parsed_data_uris:
if context != self._source_context:
if self._clean:
# remove target URI first
pass
Log.some().debug('staging data: {}->{} to {}->{}'.format(
self._source_context, self._parsed_data_uris[
self._source_context]['chopped_uri'], context,
self._parsed_data_uris[context]['chopped_uri']))
if context != 'final':
if not DataManager.copy(
parsed_src_uri=self._parsed_data_uris\
[self._source_context],
parsed_dest_uri=self._parsed_data_uris[context],
**kwargs
):
msg = 'cannot stage data by copying from {} to {}'.format(
self._parsed_data_uris[self._source_context]\
['chopped_uri'],
self._parsed_data_uris[context]['chopped_uri']
)
Log.an().error(msg)
return self._fatal(msg)
self._staged = True
return True
def _get_map_uri_list(self):
"""
Get the contents of the map URI (agave URI).
Args:
self: class instance.
Returns:
Array of base file names in the map URI. Returns False on
exception.
"""
combined_file_list = []
for uri in self._parsed_map_uris:
# make sure map URI is compatible scheme (agave)
if uri['scheme'] != 'agave':
msg = 'invalid map uri scheme for this step: {}'.format(
uri['scheme']
)
Log.an().error(msg)
return self._fatal(msg)
# get file list from URI
file_list = DataManager.list(
parsed_uri=uri,
globstr=self._step['map']['glob'],
agave=self._agave
)
if file_list is False:
msg = 'cannot get contents of map uri: {}'\
.format(uri['chopped_uri'])
Log.an().error(msg)
return self._fatal(msg)
for f in file_list:
combined_file_list.append({
'chopped_uri': uri['chopped_uri'],
'filename': f
})
return combined_file_list
def clean_up(self):
"""
Copy data from Agave archive location to step output location (data URI).
Args:
self: class instance.
Returns:
On success: True.
On failure: False.
"""
# destination _log directory, common for all map items
dest_log_dir = '{}/{}'.format(
self._parsed_data_uris[self._source_context][0]\
['chopped_uri'],
'_log'
)
# copy data for each map item
for map_item in self._map:
# copy step output
if not self._agave['agave_wrapper'].files_import_from_agave(
self._parsed_data_uris[self._source_context][0]['authority'],
self._parsed_data_uris[self._source_context][0]\
['chopped_path'],
map_item['template']['output'],
'{}/{}'.format(
map_item['run'][map_item['attempt']]['archive_uri'],
map_item['template']['output']
)
):
msg = 'agave import failed for step "{}"'\
.format(self._step['name'])
Log.an().error(msg)
return self._fatal(msg)
# check for any agave log files (*.out and *.err files)
agave_log_list = DataManager.list(
uri=map_item['run'][map_item['attempt']]['archive_uri'],
agave=self._agave
)
if agave_log_list is False:
msg = 'cannot get agave log list for step "{}"'\
.format(self._step['name'])
Log.an().error(msg)
return self._fatal(msg)
# copy each agave log file, the pattern is gf-{}-{}-{}.out or .err
for item in agave_log_list:
if re.match('^gf-\d*-.*\.(out|err)$', item):
if not self._agave['agave_wrapper'].files_import_from_agave(
self._parsed_data_uris[self._source_context][0]\
['authority'],
'{}/{}'.format(
self._parsed_data_uris[self._source_context][0]\
['chopped_path'],
'_log'
),
item,
'{}/{}'.format(
map_item['run'][map_item['attempt']]\
['archive_uri'],
item
)
):
msg = 'cannot copy agave log item "{}"'.format(item)
Log.an().error(msg)
return self._fatal(msg)
# check if anything is in the _log directory
src_log_dir = '{}/{}'.format(
map_item['run'][map_item['attempt']]['archive_uri'],
'_log'
)
if DataManager.exists(
uri=src_log_dir,
agave=self._agave
):
# create dest _log dir if it doesn't exist
if not DataManager.exists(
uri=dest_log_dir,
agave=self._agave
):
if not DataManager.mkdir(
uri=dest_log_dir,
agave=self._agave
):
msg = 'cannot create _log directory for step "{}"'\
.format(self._step['name'])
Log.an().error(msg)
return self._fatal(msg)
# get list of all items in src_log_dir
log_list = DataManager.list(
uri=src_log_dir,
agave=self._agave
)
if log_list is False:
msg = 'cannot get _log list for step "{}"'\
.format(self._step['name'])
Log.an().error(msg)
return self._fatal(msg)
# copy each list item
for item in log_list:
if not self._agave['agave_wrapper'].files_import_from_agave(
self._parsed_data_uris[self._source_context][0]\
['authority'],
'{}/{}'.format(
self._parsed_data_uris[self._source_context][0]\
['chopped_path'],
'_log'
),
item,
'{}/{}/{}'.format(
map_item['run'][map_item['attempt']]\
['archive_uri'],
'_log',
item
)
):
msg = 'cannot copy log item "{}"'.format(item)
Log.an().error(msg)
return self._fatal(msg)
self._update_status_db('FINISHED', '')
return True
def _run_map(self, map_item):
"""
Run a job for each map item and store the job ID.
Args:
self: class instance.
map_item: map item object (item of self._map).
Returns:
On success: True.
On failure: False.
"""
# load default app inputs overwrite with template inputs
inputs = {}
for input_key in self._app['inputs']:
if input_key in map_item['template']:
if map_item['template'][input_key]:
# only include an input if the value is a non-empty string
inputs[input_key] = urllib.parse.quote(
str(map_item['template'][input_key]),
safe='/:'
)
else:
if self._app['inputs'][input_key]['default']:
# only include an input if the value is a non-empty string
inputs[input_key] = urllib.parse.quote(
str(self._app['inputs'][input_key]['default']),
safe='/:'
)
# load default app parameters, overwrite with template parameters
parameters = {}
for param_key in self._app['parameters']:
if param_key in map_item['template']:
if self._app['parameters'][param_key]['type'] in ['int', 'long']:
parameters[param_key] = int(map_item['template'][param_key])
elif self._app['parameters'][param_key]['type'] == ['float', 'double']:
parameters[param_key] = float(map_item['template'][param_key])
else:
parameters[param_key] = str(map_item['template'][param_key])
else:
if self._app['parameters'][param_key]['default'] not in [None, '']:
parameters[param_key] \
= self._app['parameters'][param_key]['default']
# add execution method as parameter
parameters['exec_method'] = self._step['execution']['method']
# add execution init commands if 'init' param given
if 'init' in self._step['execution']['parameters']:
parameters['exec_init'] = self._step['execution']['parameters']['init']
# construct agave app template
name = 'gf-{}-{}-{}'.format(
str(map_item['attempt']),
slugify(self._step['name'], regex_pattern=r'[^-a-z0-9_]+'),
slugify(map_item['template']['output'], regex_pattern=r'[^-a-z0-9_]+')
)
name = name[:62]+'..' if len(name) > 64 else name
archive_path = '{}/{}'.format(
self._agave['parsed_archive_uri']['chopped_path'],
name
)
app_template = {
'name': name,
'appId': self._app['implementation']['agave']['agave_app_id'],
'archive': True,
'inputs': inputs,
'parameters': parameters,
'archiveSystem': self._agave['parsed_archive_uri']['authority'],
'archivePath': archive_path
}
# specify processors if 'slots' param given
if 'slots' in self._step['execution']['parameters']:
app_template['processorsPerNode'] = int(
self._step['execution']['parameters']['slots']
)
# specify memory if 'mem' param given
if 'mem' in self._step['execution']['parameters']:
app_template['memoryPerNode'] = '{}'.format(
self._step['execution']['parameters']['mem']
)
Log.some().debug(
"[step.%s]: agave app template:\n%s",
self._step['name'],
pprint.pformat(app_template)
)
# delete archive path if it exists
if DataManager.exists(
uri=self._agave['parsed_archive_uri']['chopped_uri']+'/'+name,
agave=self._agave
):
if not DataManager.delete(
uri=self._agave['parsed_archive_uri']['chopped_uri']+'/'+name,
agave=self._agave
):
Log.a().warning(
'cannot delete archive uri: %s/%s',
self._agave['parsed_archive_uri']['chopped_uri'],
name
)
# submit job
job = self._agave['agave_wrapper'].jobs_submit(app_template)
if not job:
msg = 'agave jobs submit failed for "{}"'.format(
app_template['name']
)
Log.an().error(msg)
return self._fatal(msg)
# log agave job id
Log.some().debug(
'[step.%s]: agave job id: %s -> %s',
self._step['name'],
map_item['template']['output'],
job['id']
)
# record job info
map_item['run'][map_item['attempt']]['agave_job_id'] = job['id']
map_item['run'][map_item['attempt']]['archive_uri'] = '{}/{}'\
.format(
self._agave['parsed_archive_uri']['chopped_uri'],
name
)
map_item['run'][map_item['attempt']]['hpc_job_id'] = ''
# set status of process
map_item['status'] = 'PENDING'
map_item['run'][map_item['attempt']]['status'] = 'PENDING'
return True
def _run_map(self, map_item):
"""
Run a job for each map item and store the job ID.
Args:
self: class instance.
map_item: map item object (item of self._map)
Returns:
On success: True.
On failure: False.
"""
# load default app inputs overwrite with template inputs
inputs = {}
for input_key in self._app['inputs']:
if input_key in map_item['template']:
inputs[input_key] = urllib.parse.quote(str(
map_item['template'][input_key] or ''),
safe='/:')
else:
inputs[input_key] = urllib.parse.quote(str(
self._app['inputs'][input_key]['default'] or ''),
safe='/:')
# load default app parameters, overwrite with template parameters
parameters = {}
for param_key in self._app['parameters']:
if param_key in map_item['template']:
parameters[param_key] = map_item['template'][param_key]
else:
parameters[param_key] \
= self._app['parameters'][param_key]['default']
# add execution method as parameter
parameters['exec_method'] = self._step['execution']['method']
# construct agave app template
name = 'gf-{}-{}-{}'.format(str(map_item['attempt']),
slugify(self._step['name']),
slugify(map_item['template']['output']))
name = name[:62] + '..' if len(name) > 64 else name
archive_path = '{}/{}'.format(
self._agave['parsed_archive_uri']['chopped_path'], name)
app_template = {
'name': name,
'appId': self._app['definition']['agave']['agave_app_id'],
'archive': True,
'inputs': inputs,
'parameters': parameters,
'archiveSystem': self._agave['parsed_archive_uri']['authority'],
'archivePath': archive_path
}
Log.some().debug("agave app template:\n%s",
pprint.pformat(app_template))
# delete archive path if it exists
if DataManager.exists(
uri=self._agave['parsed_archive_uri']['chopped_uri'] + '/' +
name,
agave=self._agave):
if not DataManager.delete(
uri=self._agave['parsed_archive_uri']['chopped_uri'] +
'/' + name,
agave=self._agave):
Log.a().warning(
'cannot delete archive uri: %s/%s',
self._agave['parsed_archive_uri']['chopped_uri'], name)
# submit job
job = self._agave['agave_wrapper'].jobs_submit(app_template)
if not job:
msg = 'agave jobs submit failed for "{}"'.format(
app_template['name'])
Log.an().error(msg)
return self._fatal(msg)
# log agave job id
Log.some().debug('agave job id: %s -> %s',
map_item['template']['output'], job['id'])
# record job info
map_item['run'][map_item['attempt']]['agave_job_id'] = job['id']
map_item['run'][map_item['attempt']]['archive_uri'] = '{}/{}'\
.format(
self._agave['parsed_archive_uri']['chopped_uri'],
name
)
map_item['run'][map_item['attempt']]['hpc_job_id'] = ''
# set status of process
map_item['status'] = 'PENDING'
map_item['run'][map_item['attempt']]['status'] = 'PENDING'
return True
def _init_context_uris(self):
"""
Generate all context URIs for this workflow run.
Context URIs are generated based on contexts given in
_parsed_job_work_uri, and the "final" context for steps given in the
_parsed_job_output_uri.
Args:
None.
Returns:
On failure: Raises WorkflowDAGException.
"""
self._context_uris['inputs'] = {}
self._context_uris['steps'] = {'final': {}}
self._parsed_context_uris['inputs'] = {}
self._parsed_context_uris['steps'] = {'final': {}}
# init contexts in parsed_job_work_uri for inputs and steps
for context in self._parsed_job_work_uri:
self._context_uris['inputs'][context] = {}
self._context_uris['steps'][context] = {}
self._parsed_context_uris['inputs'][context] = {}
self._parsed_context_uris['steps'][context] = {}
for node_name in self._topo_sort:
node = self._graph.nodes[node_name]
if node['type'] == 'input':
if node['source_context'] == context:
# use original input URI
parsed_uri = URIParser.parse(
self._workflow['inputs'][node['name']]['value'])
if not parsed_uri:
msg = 'invalid input uri: {}'.format(
self._workflow['inputs'][
node['name']]['value'])
raise WorkflowDAGException(msg)
self._context_uris['inputs'][context][node['name']]\
= parsed_uri['chopped_uri']
self._parsed_context_uris['inputs'][context]\
[node['name']] = parsed_uri
else:
# switch context of input URI
new_base_uri = '{}/_input-{}'.format(
self._parsed_job_work_uri[context]['chopped_uri'],
slugify(node['name']))
# create new base URI
if not DataManager.mkdir(
uri=new_base_uri,
recursive=True,
**{context: self._context_options[context]}):
msg = 'cannot create new base uri for input: {}'\
.format(new_base_uri)
Log.an().error(msg)
raise WorkflowDAGException(msg)
# switch input URI base
switched_uri = URIParser.switch_context(
self._workflow['inputs'][node['name']]['value'],
new_base_uri)
if not switched_uri:
msg = (
'cannot switch input uri context to '
'new base URI: {}->{}'
).format(
self._workflow['inputs'][node['name']]\
['value'],
new_base_uri
)
Log.an().error(msg)
raise WorkflowDAGException(msg)
self._context_uris['inputs'][context][node['name']]\
= switched_uri['chopped_uri']
self._parsed_context_uris['inputs'][context]\
[node['name']] = switched_uri
else: # node['type'] == 'step'
self._context_uris['steps'][context][node['name']]\
= '{}/{}'.format(
self._parsed_job_work_uri[context]['chopped_uri'],
slugify(node['name'])
)
self._parsed_context_uris['steps'][context][node['name']]\
= URIParser.parse(
self._context_uris['steps'][context][node['name']]
)
# init final contexts for steps
for node_name in self._topo_sort:
node = self._graph.nodes[node_name]
if node['type'] == 'step':
self._context_uris['steps']['final'][node['name']]\
= '{}/{}'.format(
self._parsed_job_output_uri['chopped_uri'],
slugify(node['name'])
)
self._parsed_context_uris['steps']['final'][node['name']]\
= URIParser.parse(
self._context_uris['steps']['final'][node['name']]
)
def register_agave_app(self, agave, agave_config, agave_params, agave_publish):
"""
Register app in Agave.
Args:
self: class instance
Returns:
On success: True.
On failure: False.
"""
Log.some().info('registering agave app %s', str(self._path))
Log.some().info('app version: %s', self._config['version'])
# compile agave app template
if not TemplateCompiler.compile_template(
self._path,
'agave-app-def.json.j2',
self._path / 'agave-app-def.json',
version=self._config['version'],
agave=agave_params['agave']
):
Log.a().warning(
'cannot compile agave app "%s" definition from template',
self._app['name']
)
return False
# create main apps URI
parsed_agave_apps_uri = URIParser.parse(
'agave://{}/{}'.format(
agave_params['agave']['deploymentSystem'],
agave_params['agave']['appsDir']
)
)
Log.some().info(
'creating main apps uri: %s',
parsed_agave_apps_uri['chopped_uri']
)
if not DataManager.mkdir(
parsed_uri=parsed_agave_apps_uri,
recursive=True,
agave={
'agave': agave,
'agave_config': agave_config
}
):
Log.a().warning('cannot create main agave apps uri')
return False
# delete app uri if it exists
parsed_app_uri = URIParser.parse(
'agave://{}/{}/{}'.format(
agave_params['agave']['deploymentSystem'],
agave_params['agave']['appsDir'],
self._app['folder']
)
)
Log.some().info(
'deleting app uri if it exists: %s',
parsed_app_uri['chopped_uri']
)
if not DataManager.delete(
parsed_uri=parsed_app_uri,
agave={
'agave': agave,
'agave_config': agave_config
}
):
# log warning, but ignore.. deleting non-existant uri returns False
Log.a().warning(
'cannot delete app uri: %s', parsed_app_uri['chopped_uri']
)
# upload app assets
parsed_assets_uri = URIParser.parse(str(self._path / 'assets'))
Log.some().info(
'copying app assets from %s to %s',
parsed_assets_uri['chopped_uri'],
parsed_app_uri['chopped_uri']
)
if not DataManager.copy(
parsed_src_uri=parsed_assets_uri,
parsed_dest_uri=parsed_app_uri,
local={},
agave={
'agave': agave,
'agave_config': agave_config
}
):
Log.a().warning(
'cannot copy app assets from %s to %s',
parsed_assets_uri['chopped_uri'],
parsed_app_uri['chopped_uri']
)
return False
# upload test script
parsed_test_uri = URIParser.parse(
'{}/{}'.format(
parsed_app_uri['chopped_uri'],
'test'
)
)
Log.some().info(
'creating test uri: %s', parsed_test_uri['chopped_uri']
)
if not DataManager.mkdir(
parsed_uri=parsed_test_uri,
recursive=True,
agave={
'agave': agave,
'agave_config': agave_config
}
):
Log.a().warning(
'cannot create test uri: %s', parsed_test_uri['chopped_uri']
)
return False
parsed_local_test_script = URIParser.parse(
str(self._path / 'test' / 'test.sh')
)
parsed_agave_test_script = URIParser.parse(
'{}/{}'.format(parsed_test_uri['chopped_uri'], 'test.sh')
)
Log.some().info(
'copying test script from %s to %s',
parsed_local_test_script['chopped_uri'],
parsed_agave_test_script['chopped_uri']
)
if not DataManager.copy(
parsed_src_uri=parsed_local_test_script,
parsed_dest_uri=parsed_agave_test_script,
local={},
agave={
'agave': agave,
'agave_config': agave_config
}
):
Log.a().warning(
'cannot copy test script from %s to %s',
parsed_local_test_script['chopped_uri'],
parsed_agave_test_script['chopped_uri']
)
return False
# update existing app, or register new app
Log.some().info('registering agave app')
app_definition = self._yaml_to_dict(
str(self._path / 'agave-app-def.json')
)
if not app_definition:
Log.a().warning(
'cannot load agave app definition: %s',
str(self._path / 'agave-app-def.json')
)
return False
agwrap = AgaveAppsAddUpdate(
agave, agave_config
)
app_add_result = agwrap.call(app_definition)
if not app_add_result:
Log.a().warning(
'cannot register agave app:\n%s', pprint.pformat(app_definition)
)
return False
register_result = {}
# publish app
if agave_publish:
Log.some().info('publishing agave app')
agwrap = AgaveAppsPublish(
agave, agave_config
)
app_publish_result = agwrap.call(app_add_result['id'])
if not app_publish_result:
Log.a().warning(
'cannot publish agave app: %s', app_add_result['id']
)
return False
# return published id and revision
register_result = {
'id': app_publish_result['id'],
'version': self._config['version'],
'revision': 'u{}'.format(app_publish_result['revision'])
}
else:
# return un-published id and blank revision
register_result = {
'id': app_add_result['id'],
'version': self._config['version'],
'revision': ''
}
return register_result
def _get_map_uri_list(self):
"""
Get the contents of the map URI (agave URI).
Args:
self: class instance.
Returns:
Array of base file names in the map URI. Returns False on
exception.
"""
combined_file_list = []
for uri in self._parsed_map_uris:
# make sure map URI is compatible scheme (agave)
if uri['scheme'] != 'agave':
msg = 'invalid map uri scheme for this step: {}'.format(
uri['scheme']
)
Log.an().error(msg)
return self._fatal(msg)
# get file list from URI
file_list = DataManager.list(
parsed_uri=uri,
globstr=self._step['map']['glob'],
agave=self._agave
)
if file_list is False:
msg = 'cannot get contents of map uri: {}'\
.format(uri['chopped_uri'])
Log.an().error(msg)
return self._fatal(msg)
if self._step['map']['inclusive']:
# filter with glob
if glob.globfilter(
[uri['name']],
self._step['map']['glob'],
flags=glob.EXTGLOB|glob.GLOBSTAR
):
combined_file_list.append({
'chopped_uri': '{}://{}{}'.format(
uri['scheme'],
uri['authority'],
uri['folder']
),
'filename': uri['name']
})
for f in file_list:
if '/' in f:
# reparse uri to correctly represent recursive elements
new_uri = URIParser.parse('{}/{}'.format(uri['chopped_uri'], f))
combined_file_list.append({
'chopped_uri': '{}://{}{}'.format(
new_uri['scheme'],
new_uri['authority'],
new_uri['folder']
),
'filename': new_uri['name']
})
else:
combined_file_list.append({
'chopped_uri': uri['chopped_uri'],
'filename': f
})
return combined_file_list
def _init_context_uris(self):
"""
Generate all context URIs for this workflow run.
Context URIs are generated based on contexts given in
_parsed_job_work_uri, and the "final" context for steps given in the
_parsed_job_output_uri.
Args:
None.
Returns:
On failure: Raises WorkflowDAGException.
"""
self._context_uris['inputs'] = {}
self._context_uris['steps'] = {'final': {}}
self._parsed_context_uris['inputs'] = {}
self._parsed_context_uris['steps'] = {'final': {}}
# init all data contexts
for context in {
Contexts.get_data_scheme_of_exec_context(con)
for con in self._exec_contexts
} | self._data_contexts:
self._context_uris['inputs'][context] = {}
self._parsed_context_uris['inputs'][context] = {}
for node_name in self._topo_sort:
node = self._graph.nodes[node_name]
if node['type'] == 'input':
if node['source_context'] == context:
# use original input URI
parsed_uri = URIParser.parse(
self._workflow['inputs'][node['name']]['value'])
if not parsed_uri:
msg = 'invalid input uri: {}'.format(
self._workflow['inputs'][
node['name']]['value'])
raise WorkflowDAGException(msg)
self._context_uris['inputs'][context][node['name']]\
= parsed_uri['chopped_uri']
self._parsed_context_uris['inputs'][context]\
[node['name']] = parsed_uri
else:
# skip if _parsed_job_work_uri is not defined for this context
# this implies that there is no execution defined for that context,
# so no need to setup the data staging location at the work_uri
if context not in self._parsed_job_work_uri:
continue
# switch context of input URI
new_base_uri = '{}/_input-{}'.format(
self._parsed_job_work_uri[context]['chopped_uri'],
slugify(node['name'],
regex_pattern=r'[^-a-z0-9_]+'))
# create new base URI
if not DataManager.mkdir(
uri=new_base_uri,
recursive=True,
**{context: self._context_options[context]}):
msg = 'cannot create new base uri for input: {}'\
.format(new_base_uri)
Log.an().error(msg)
raise WorkflowDAGException(msg)
# switch input URI base
switched_uri = URIParser.switch_context(
self._workflow['inputs'][node['name']]['value'],
new_base_uri)
if not switched_uri:
msg = (
'cannot switch input uri context to '
'new base URI: {}->{}'
).format(
self._workflow['inputs'][node['name']]\
['value'],
new_base_uri
)
Log.an().error(msg)
raise WorkflowDAGException(msg)
self._context_uris['inputs'][context][node['name']]\
= switched_uri['chopped_uri']
self._parsed_context_uris['inputs'][context]\
[node['name']] = switched_uri
for context in {
Contexts.get_data_scheme_of_exec_context(con)
for con in self._exec_contexts
}:
self._context_uris['steps'][context] = {}
self._parsed_context_uris['steps'][context] = {}
for node_name in self._topo_sort:
node = self._graph.nodes[node_name]
if node['type'] == 'step':
self._context_uris['steps'][context][node['name']]\
= '{}/{}'.format(
self._parsed_job_work_uri[context]['chopped_uri'],
slugify(node['name'], regex_pattern=r'[^-a-z0-9_]+')
)
self._parsed_context_uris['steps'][context][node['name']]\
= URIParser.parse(
self._context_uris['steps'][context][node['name']]
)
# init final contexts for steps
for node_name in self._topo_sort:
node = self._graph.nodes[node_name]
if node['type'] == 'step':
self._context_uris['steps']['final'][node['name']]\
= '{}/{}'.format(
self._parsed_job_output_uri['chopped_uri'],
slugify(node['name'], regex_pattern=r'[^-a-z0-9_]+')
)
self._parsed_context_uris['steps']['final'][node['name']]\
= URIParser.parse(
self._context_uris['steps']['final'][node['name']]
)
def stage_final(self, **kwargs):
"""
Move data to final context from source URI.
Set _staged_final indicator to True on success.
Args:
self: class instance.
**kwargs: additional arguments required by DataManager.move().
Returns:
True or False.
"""
for context in self._parsed_data_uris:
if context != self._source_context:
if self._clean:
# remove target URI first
pass
Log.some().debug('staging final data: {}->{} to {}->{}'.format(
self._source_context, self._parsed_data_uris[
self._source_context]['chopped_uri'], context,
self._parsed_data_uris[context]['chopped_uri']))
if context == 'final':
if (self._parsed_data_uris[self._source_context]['scheme']
== 'local'
and self._parsed_data_uris[context]['scheme']
== 'local'):
# move final data instead of copy, only for local-->local schemes
if not DataManager.move(
parsed_src_uri=self._parsed_data_uris\
[self._source_context],
parsed_dest_uri=self._parsed_data_uris[context],
**kwargs
):
msg = 'cannot stage final data by copying from {} to {}'.format(
self._parsed_data_uris[self._source_context]\
['chopped_uri'],
self._parsed_data_uris[context]['chopped_uri']
)
Log.an().error(msg)
return self._fatal(msg)
else:
if not DataManager.copy(
parsed_src_uri=self._parsed_data_uris\
[self._source_context],
parsed_dest_uri=self._parsed_data_uris[context],
**kwargs
):
msg = 'cannot stage final data by copying from {} to {}'.format(
self._parsed_data_uris[self._source_context]\
['chopped_uri'],
self._parsed_data_uris[context]['chopped_uri']
)
Log.an().error(msg)
return self._fatal(msg)
self._staged_final = True
return True
def clean_up(self):
"""
Copy data from Agave archive location to step output location (data URI).
Args:
self: class instance.
Returns:
On success: True.
On failure: False.
"""
# destination _log directory, common for all map items
dest_log_dir = '{}/{}'.format(
self._parsed_data_uris[self._source_context]\
['chopped_uri'],
'_log'
)
# create instance of agave wrapper class for data import
agwrap = AgaveFilesImportDataFromAgave(
self._agave['agave'],
self._config['agave']
)
# copy data for each map item
for map_item in self._map:
# copy step output
if not agwrap.call(
self._parsed_data_uris[self._source_context]['authority'],
self._parsed_data_uris[self._source_context]\
['chopped_path'],
map_item['template']['output'],
'{}/{}'.format(
map_item['run'][map_item['attempt']]['archive_uri'],
map_item['template']['output']
)
):
msg = 'agave import failed for step "{}"'\
.format(self._step['name'])
Log.an().error(msg)
return self._fatal(msg)
# check if anything is in the _log directory
src_log_dir = '{}/{}'.format(
map_item['run'][map_item['attempt']]['archive_uri'],
'_log'
)
if DataManager.exists(
uri=src_log_dir,
agave={
'agave': self._agave['agave'],
'agave_config': self._config['agave']
}
):
# create dest _log dir if it doesn't exist
if not DataManager.exists(
uri=dest_log_dir,
agave={
'agave': self._agave['agave'],
'agave_config': self._config['agave']
}
):
if not DataManager.mkdir(
uri=dest_log_dir,
agave={
'agave': self._agave['agave'],
'agave_config': self._config['agave']
}
):
msg = 'cannot create _log directory for step "{}"'\
.format(self._step['name'])
Log.an().error(msg)
return self._fatal(msg)
# get list of all items in src_log_dir
log_list = DataManager.list(
uri=src_log_dir,
agave={
'agave': self._agave['agave'],
'agave_config': self._config['agave']
}
)
if not log_list:
msg = 'cannot get _log list for step "{}"'\
.format(self._step['name'])
Log.an().error(msg)
return self._fatal(msg)
# copy each list item
for item in log_list:
if not agwrap.call(
self._parsed_data_uris[self._source_context]\
['authority'],
'{}/{}'.format(
self._parsed_data_uris[self._source_context]\
['chopped_path'],
'_log'
),
item,
'{}/{}/{}'.format(
map_item['run'][map_item['attempt']]\
['archive_uri'],
'_log',
item
)
):
msg = 'cannot copy log item "{}"'.format(item)
Log.an().error(msg)
return self._fatal(msg)
self._update_status_db('FINISHED', '')
return True
