def get_db(container, reset_connection=False):
# type: (AnySettingsContainer, bool) -> MongoDatabase
"""
Obtains the database connection from configured application settings.
If :paramref:`reset_connection` is ``True``, the :paramref:`container` must be the application :class:`Registry` or
any container that can retrieve it to accomplish reference reset. Otherwise, any settings container can be provided.
"""
registry = get_registry(container, nothrow=True)
if not reset_connection and registry and registry.db:
return registry.db
database = MongoDatabase(container)
if reset_connection:
registry = get_registry(container)
registry.db = database
return database
def get_pywps_service(environ=None, is_worker=False):
# type: (SettingsType, bool) -> WorkerService
"""
Generates the PyWPS Service that provides WPS-1/2 XML endpoint.
"""
environ = environ or {}
try:
# get config file
registry = get_registry()
settings = get_settings(registry)
pywps_cfg = environ.get("PYWPS_CFG") or settings.get(
"PYWPS_CFG") or os.getenv("PYWPS_CFG")
if not isinstance(
pywps_cfg,
ConfigParser) or not settings.get("weaver.wps_configured"):
load_pywps_config(settings, config=pywps_cfg)
# call pywps application with processes filtered according to the adapter's definition
process_store = get_db(registry).get_store(
StoreProcesses) # type: StoreProcesses
processes_wps = [
process.wps() for process in process_store.list_processes(
visibility=Visibility.PUBLIC)
]
service = WorkerService(processes_wps,
is_worker=is_worker,
settings=settings)
except Exception as ex:
LOGGER.exception(
"Error occurred during PyWPS Service and/or Processes setup.")
raise OWSNoApplicableCode(
f"Failed setup of PyWPS Service and/or Processes. Error [{ex!r}]")
return service
def _validate_process(self):
try:
registry = get_registry()
store = get_db(registry).get_store(StoreProcesses)
process = store.fetch_by_id(self.process) # raise if not found
except (ProcessNotAccessible, ProcessNotFound):
raise PackageNotFound("Cannot find '{}' package for process '{}'".format(PROCESS_BUILTIN, self.process))
if process.type != PROCESS_BUILTIN:
raise PackageExecutionError("Invalid package is not of type '{}'".format(PROCESS_BUILTIN))
def get_db(container=None, reset_connection=False):
# type: (Optional[Union[AnyRegistryContainer, AnySettingsContainer]], bool) -> MongoDatabase
"""
Obtains the database connection from configured application settings.
If :paramref:`reset_connection` is ``True``, the :paramref:`container` must be the application :class:`Registry` or
any container that can retrieve it to accomplish reference reset. Otherwise, any settings container can be provided.
.. note::
It is preferable to provide a registry reference to reuse any available connection whenever possible.
Giving application settings will require establishing a new connection.
"""
if not reset_connection and isinstance(container, Request):
db = getattr(container, "db", None)
if isinstance(db, MongoDatabase):
return db
registry = get_registry(container, nothrow=True)
if not reset_connection and registry and isinstance(
getattr(registry, "db", None), MongoDatabase):
return registry.db
database = MongoDatabase(container)
if reset_connection:
registry.db = database
return database
def execute_process(task, job_id, wps_url, headers=None):
# type: (Task, UUID, str, Optional[HeadersType]) -> StatusType
"""
Celery task that executes the WPS process job monitoring as status updates (local and remote).
"""
from weaver.wps.service import get_pywps_service
LOGGER.debug("Job execute process called.")
task_process = get_celery_process()
rss_start = task_process.memory_info().rss
registry = get_registry(
None) # local thread, whether locally or dispatched celery
settings = get_settings(registry)
db = get_db(
registry, reset_connection=True
) # reset the connection because we are in a forked celery process
store = db.get_store(StoreJobs)
job = store.fetch_by_id(job_id)
job.started = now()
job.status = Status.STARTED # will be mapped to 'RUNNING'
job.status_message = f"Job {Status.STARTED}." # will preserve detail of STARTED vs RUNNING
job.save_log(message=job.status_message)
task_logger = get_task_logger(__name__)
job.save_log(logger=task_logger, message="Job task setup initiated.")
load_pywps_config(settings)
job.progress = JobProgress.SETUP
job.task_id = task.request.id
job.save_log(logger=task_logger, message="Job task setup completed.")
job = store.update_job(job)
# Flag to keep track if job is running in background (remote-WPS, CWL app, etc.).
# If terminate signal is sent to worker task via API dismiss request while still running in background,
# the raised exception within the task will switch the job to Status.FAILED, but this will not raise an
# exception here. Since the task execution 'succeeds' without raising, it skips directly to the last 'finally'.
# Patch it back to Status.DISMISSED in this case.
task_terminated = True
try:
job.progress = JobProgress.DESCRIBE
job.save_log(logger=task_logger,
message=f"Employed WPS URL: [{wps_url!s}]",
level=logging.DEBUG)
job.save_log(
logger=task_logger,
message=f"Execute WPS request for process [{job.process!s}]")
wps_process = fetch_wps_process(job, wps_url, headers, settings)
# prepare inputs
job.progress = JobProgress.GET_INPUTS
job.save_log(logger=task_logger,
message="Fetching job input definitions.")
wps_inputs = parse_wps_inputs(wps_process, job)
# prepare outputs
job.progress = JobProgress.GET_OUTPUTS
job.save_log(logger=task_logger,
message="Fetching job output definitions.")
wps_outputs = [(o.identifier, o.dataType == WPS_COMPLEX_DATA)
for o in wps_process.processOutputs]
# if process refers to a remote WPS provider, pass it down to avoid unnecessary re-fetch request
if job.is_local:
process = None # already got all the information needed pre-loaded in PyWPS service
else:
service = Service(name=job.service, url=wps_url)
process = Process.from_ows(wps_process, service, settings)
job.progress = JobProgress.EXECUTE_REQUEST
job.save_log(logger=task_logger,
message="Starting job process execution.")
job.save_log(
logger=task_logger,
message=
"Following updates could take a while until the Application Package answers..."
)
wps_worker = get_pywps_service(environ=settings, is_worker=True)
execution = wps_worker.execute_job(job,
wps_inputs=wps_inputs,
wps_outputs=wps_outputs,
remote_process=process,
headers=headers)
if not execution.process and execution.errors:
raise execution.errors[0]
# adjust status location
wps_status_path = get_wps_local_status_location(
execution.statusLocation, settings)
job.progress = JobProgress.EXECUTE_STATUS_LOCATION
LOGGER.debug("WPS status location that will be queried: [%s]",
wps_status_path)
if not wps_status_path.startswith("http") and not os.path.isfile(
wps_status_path):
LOGGER.warning(
"WPS status location not resolved to local path: [%s]",
wps_status_path)
job.save_log(
logger=task_logger,
level=logging.DEBUG,
message=f"Updated job status location: [{wps_status_path}].")
job.status = Status.RUNNING
job.status_message = execution.statusMessage or f"{job!s} initiation done."
job.status_location = wps_status_path
job.request = execution.request
job.response = execution.response
job.progress = JobProgress.EXECUTE_MONITOR_START
job.save_log(logger=task_logger,
message="Starting monitoring of job execution.")
job = store.update_job(job)
max_retries = 5
num_retries = 0
run_step = 0
while execution.isNotComplete() or run_step == 0:
if num_retries >= max_retries:
job.save_log(errors=execution.errors, logger=task_logger)
job = store.update_job(job)
raise Exception(
f"Could not read status document after {max_retries} retries. Giving up."
)
try:
# NOTE:
# Don't actually log anything here until process is completed (success or fail) so that underlying
# WPS execution logs can be inserted within the current job log and appear continuously.
# Only update internal job fields in case they get referenced elsewhere.
progress_min = JobProgress.EXECUTE_MONITOR_LOOP
progress_max = JobProgress.EXECUTE_MONITOR_DONE
job.progress = progress_min
run_delay = wait_secs(run_step)
execution = check_wps_status(location=wps_status_path,
settings=settings,
sleep_secs=run_delay)
job_msg = (execution.statusMessage or "").strip()
job.response = execution.response
job.status = map_status(execution.getStatus())
job_status_msg = job_msg or "n/a"
job_percent = execution.percentCompleted
job.status_message = f"Job execution monitoring (progress: {job_percent}%, status: {job_status_msg})."
if execution.isComplete():
msg_progress = f" (status: {job_msg})" if job_msg else ""
if execution.isSucceded():
wps_package.retrieve_package_job_log(
execution, job, progress_min, progress_max)
job.status = map_status(Status.SUCCEEDED)
job.status_message = f"Job succeeded{msg_progress}."
job.progress = progress_max
job.save_log(logger=task_logger)
job_results = [
ows2json_output_data(output, process, settings)
for output in execution.processOutputs
]
job.results = make_results_relative(
job_results, settings)
else:
task_logger.debug("Job failed.")
wps_package.retrieve_package_job_log(
execution, job, progress_min, progress_max)
job.status_message = f"Job failed{msg_progress}."
job.progress = progress_max
job.save_log(errors=execution.errors,
logger=task_logger)
task_logger.debug(
"Mapping Job references with generated WPS locations.")
map_locations(job, settings)
job = store.update_job(job)
except Exception as exc:
num_retries += 1
task_logger.debug("Exception raised: %s", repr(exc))
job.status_message = f"Could not read status XML document for {job!s}. Trying again..."
job.save_log(errors=execution.errors, logger=task_logger)
job = store.update_job(job)
sleep(1)
else:
num_retries = 0
run_step += 1
finally:
task_terminated = False # reached only if WPS execution completed (worker not terminated beforehand)
job = store.update_job(job)
except Exception as exc:
# if 'execute_job' finishes quickly before even reaching the 'monitoring loop'
# consider WPS execution produced an error (therefore Celery worker not terminated)
task_terminated = False
LOGGER.exception("Failed running [%s]", job)
LOGGER.debug("Failed job [%s] raised an exception.", job, exc_info=exc)
# note: don't update the progress here to preserve last one that was set
job.status = map_status(Status.FAILED)
job.status_message = f"Failed to run {job!s}."
errors = f"{fully_qualified_name(exc)}: {exc!s}"
job.save_log(errors=errors, logger=task_logger)
job = store.update_job(job)
finally:
# if task worker terminated, local 'job' is out of date compared to remote/background runner last update
job = store.fetch_by_id(job.id)
if task_terminated and map_status(job.status) == Status.FAILED:
job.status = Status.DISMISSED
task_success = map_status(
job.status) not in JOB_STATUS_CATEGORIES[StatusCategory.FAILED]
collect_statistics(task_process, settings, job, rss_start)
if task_success:
job.progress = JobProgress.EXECUTE_MONITOR_END
job.status_message = f"Job {job.status}."
job.save_log(logger=task_logger)
if task_success:
job.progress = JobProgress.NOTIFY
send_job_complete_notification_email(job, task_logger, settings)
if job.status not in JOB_STATUS_CATEGORIES[StatusCategory.FINISHED]:
job.status = Status.SUCCEEDED
job.status_message = f"Job {job.status}."
job.mark_finished()
if task_success:
job.progress = JobProgress.DONE
job.save_log(logger=task_logger, message="Job task complete.")
job = store.update_job(job)
return job.status