def get_pipeline_status(cache_id, name):
status = status_mgr(f'{FILE_PATH}/tmp')
status.initialize()
pipeline_status = status.get(f'./{cache_id}/{name}')
start_time = None
finish_time = None
pipeline_id = None
status = None
success = None
error_log = None
if pipeline_status and pipeline_status.last_execution:
last_execution = pipeline_status.last_execution
start_time = last_execution.start_time
pipeline_id = last_execution.pipeline_id
finish_time = last_execution.finish_time
success = last_execution.success
error_log = last_execution.error_log
status = pipeline_status.state()
return {
'start_time': start_time,
'finish_time': finish_time,
'pipeline_id': pipeline_id,
'status': status,
'error_log': error_log,
'success': success,
}
def specs(self, argument, root_dir, ignore_missing_deps=False):
status_manager = status_mgr(root_dir)
specs = []
for spec in pipelines(ignore_missing_deps=ignore_missing_deps,
root_dir=root_dir,
status_manager=status_manager):
if match_pipeline_id(argument, spec.pipeline_id):
specs.append(spec)
return specs
def pipelines(prefixes=None,
ignore_missing_deps=False,
root_dir='.',
status_manager=None):
specs: Iterator[PipelineSpec] = find_specs(root_dir)
hasher = HashCalculator()
if status_manager is None:
status_manager = status_mgr()
if prefixes is None:
prefixes = ('', )
while specs is not None:
deferred = []
found = False
for spec_ in specs:
spec: PipelineSpec = spec_
if not any(
spec.pipeline_id.startswith(prefix)
for prefix in prefixes):
continue
if (spec.pipeline_details is not None and validate_pipeline(
spec.pipeline_details, spec.validation_errors)):
resolve_processors(spec)
process_schedules(spec)
try:
hasher.calculate_hash(spec, status_manager,
ignore_missing_deps)
found = True
except DependencyMissingException as e_:
e: DependencyMissingException = e_
deferred.append((e.spec, e.missing))
continue
yield spec
if found and len(deferred) > 0:
specs = iter((x[0] for x in deferred))
else:
for spec, missing in deferred:
spec.validation_errors.append(
SpecError(
'Missing dependency',
'Failed to find a dependency: {}'.format(missing)))
yield spec
specs = None
# -*- coding: utf-8 -*-
import json
import threading
import time
from http.server import HTTPServer, BaseHTTPRequestHandler
from datapackage_pipelines.manager import execute_pipeline, run_pipelines
from datapackage_pipelines.specs.specs import pipelines
from datapackage_pipelines.utilities.execution_id import gen_execution_id
from datapackage_pipelines.status import status_mgr
called_hooks = []
progresses = 0
status = status_mgr()
class SaveHooks(BaseHTTPRequestHandler):
def do_POST(self):
global progresses
content_len = int(self.headers.get('content-length', 0))
post_body = self.rfile.read(content_len)
hook = json.loads(post_body)
if hook['event'] != 'progress':
called_hooks.append(hook)
else:
progresses += 1
self.send_response(200)
self.end_headers()
return
def run_pipeline(self, cache_id=None, verbose=False, num_rows=-1, background=False):
'''
Starts a thread that runs the datapackage pipelines for this pipeline
- On fail, return error message
- On success, return datapackage.json contents and the resulting csv
- On both fail and success return a status code and a
unique id that can be passed back in to this function
to use the cache
- if run in the background use the static functions get_pipeline_status and
get_pipeline_data to access the results.
'''
if not cache_id:
cache_id = str(uuid.uuid1())
# We have to check the cache_id value since it's
# potentially being passed in from the outside
pattern = re.compile(
r'^[0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}'
r'-[0-9a-fA-F]{4}-[0-9a-fA-F]{12}$'
)
if not pattern.match(cache_id):
raise Exception('The unique ID that was provided was not in uuid format')
''' IMPORTANT '''
# If the file structure between this file and the tmp folder
# ever changes this code must change
cache_dir = f'{ROOT_DIR}/{cache_id}'
results_folder = f'{cache_dir}/results'
# Create the directory and file
if not os.path.exists(cache_dir):
start = time.time()
os.makedirs(cache_dir)
BcodmoPipeline.log_slow_compute(start, cache_id, 'creating the directories')
try:
start = time.time()
self.save_to_file(f'{cache_dir}/pipeline-spec.yaml.original', steps=self._steps)
BcodmoPipeline.log_slow_compute(start, cache_id, 'creating the pipeline-spec.original.yaml file')
# Create a new save step so we can access the data here
new_save_step = {
'run': 'dump_to_path',
'parameters': {
'out-path': results_folder,
'temporal_format_property': 'outputFormat',
}
}
new_steps = self._steps + [new_save_step]
start = time.time()
self.save_to_file(f'{cache_dir}/pipeline-spec.yaml', steps=new_steps)
BcodmoPipeline.log_slow_compute(start, cache_id, 'creating the pipeline-spec.yaml file')
# Remove the results folder
start = time.time()
shutil.rmtree(results_folder, ignore_errors=True)
BcodmoPipeline.log_slow_compute(start, cache_id, 'removing the results folder')
start = time.time()
pipeline_id = f'./{cache_id}/{self.name}'
status = status_mgr(ROOT_DIR)
status.initialize()
pipeline_status = status.get(pipeline_id)
last_execution = pipeline_status.last_execution
BcodmoPipeline.log_slow_compute(start, cache_id, 'checking the status before creating a thread')
old_start_time = None
if last_execution:
old_start_time = last_execution.start_time
start = time.time()
x = threading.Thread(target=self.run_pipeline_thread, args=(cache_id, verbose,), daemon=True)
BcodmoPipeline.log_slow_compute(start, cache_id, 'creating the thread')
start = time.time()
x.start()
BcodmoPipeline.log_slow_compute(start, cache_id, 'starting the thread')
if background:
while True:
# Loop until the next pipeline has started
start = time.time()
status = status_mgr(ROOT_DIR)
status.initialize()
pipeline_status = status.get(pipeline_id)
last_execution = pipeline_status.last_execution
BcodmoPipeline.log_slow_compute(start, cache_id, 'checking the status after creating the thread')
if last_execution and last_execution.start_time != old_start_time:
break
if x.is_alive():
time.sleep(0.1)
else:
return {
'status_code': 1,
'cache_id': cache_id,
'yaml': self.get_yaml(),
'error_text': 'There was an unknown error in starting the pipeline',
}
return {
'status_code': 0,
'cache_id': cache_id,
'yaml': self.get_yaml(),
}
else:
# Join the thread
x.join()
status_dict = BcodmoPipeline.get_pipeline_status(cache_id, self.name)
if status_dict['success']:
pipeline_data = BcodmoPipeline.get_pipeline_data(cache_id, num_rows)
return {
'status_code': 0,
'cache_id': cache_id,
'yaml': self.get_yaml(),
'datapackage': pipeline_data['datapackage'],
'resources': pipeline_data['resources'],
}
else:
return {
'status_code': 1,
'cache_id': cache_id,
'yaml': self.get_yaml(),
'error_text': status_dict.error_log,
}
finally:
try:
start = time.time()
# Clean up the directory, deleting old folders
cur_time = time.time()
dirs = [
folder_name for folder_name in os.listdir(f'{FILE_PATH}/tmp')
if not folder_name.startswith('.')
]
for folder_name in dirs:
folder = f'{FILE_PATH}/tmp/{folder_name}'
st = os.stat(folder)
modified_time = st.st_mtime
age = cur_time - modified_time
if age > DAY * 30:
shutil.rmtree(folder)
BcodmoPipeline.log_slow_compute(start, cache_id, 'checking age status of folders after complete')
except Exception as e:
logger.info(f'There was an error trying to clean up folder: {str(e)}')
logger.error(vars(e))
def run_pipeline_thread(self, cache_id, verbose):
cache_dir = f'{ROOT_DIR}/{cache_id}'
pipeline_spec_path = f'{cache_dir}/pipeline-spec.yaml'
pipeline_id = f'./{cache_id}/{self.name}'
dpp_command_path, processor_path = self._get_version_paths(self.version)
os.environ['DPP_PROCESSOR_PATH'] = processor_path
try:
# Activate the correct virtual environment
start = time.time()
self._activate_virtualenv(self.version)
BcodmoPipeline.log_slow_compute(start, cache_id, 'activating the virtualenv')
# Set the verbose string if necessary
if verbose:
command_list = [dpp_command_path, 'run', '--verbose', pipeline_id]
else:
command_list = [dpp_command_path, 'run', pipeline_id]
# Start the dpp process
start = time.time()
p = subprocess.Popen(
command_list,
stderr=subprocess.DEVNULL,
stdout=subprocess.DEVNULL,
cwd=ROOT_DIR,
)
BcodmoPipeline.log_slow_compute(start, cache_id, 'creating the process')
sleep_timer = 1
start = time.time()
while p.poll() is None:
BcodmoPipeline.log_slow_compute(start, cache_id, 'polling the process')
time.sleep(1)
if sleep_timer != 5:
sleep_timer += 1
# The pipeline-spec.yaml was deleted, need to end the process now
if not os.path.exists(pipeline_spec_path):
# Get the chilren of the dpp process (the dpp slave process)
children = [child.pid for child in psutil.Process(p.pid).children()]
# Terminate the parent process
p.terminate()
# Terminate all of the children processes
for child in children:
os.kill(child, signal.SIGTERM)
# Invalidate the pipeline in the dpp backend
status = status_mgr(ROOT_DIR)
status.initialize()
pipeline_status = status.get(pipeline_id)
if pipeline_status:
last_execution = pipeline_status.last_execution
if last_execution:
last_execution.finish_execution(
False,
{},
['This pipeline was stopped by laminar'],
)
# One last try
if p.poll() is None:
p.kill()
break
start = time.time()
finally:
# Deactivate the virtualenv - not sure if this is necessary since it is a thread
start = time.time()
self._deactivate_virtualenv()
BcodmoPipeline.log_slow_compute(start, cache_id, 'deactivating the virtualenv')
# If the pipeline-spec.yaml file has been deleted since this thread started, the
# whole cache_id folder should be deleted
if not os.path.exists(pipeline_spec_path) and os.path.exists(cache_dir):
shutil.rmtree(cache_dir)
def register_all_pipelines(root_dir='.'):
for spec in pipelines(root_dir=root_dir):
ps = status_mgr().get(spec.pipeline_id)
ps.init(spec.pipeline_details, spec.source_details,
spec.validation_errors, spec.cache_hash)
ps.save()
