def __init__(self, **kw):
self.recipe = kw.pop('recipe_obj')
self.sous_chef_path = kw.pop('sous_chef_path')
self.sous_chef_kwargs = kw
self.kw_prefix = settings.MERLYNNE_KWARGS_PREFIX
self.kw_ttl = settings.MERLYNNE_KWARGS_TTL
self.result_ttl = settings.MERLYNNE_RESULTS_TTL
self.q = queues.get('recipe')
def __init__(self, **kw):
self.recipe = kw.pop('recipe_obj')
self.sous_chef_path = kw.pop('sous_chef_path')
self.sous_chef_kwargs = kw
self.kw_prefix = settings.MERLYNNE_KWARGS_PREFIX
self.kw_ttl = settings.MERLYNNE_KWARGS_TTL
self.result_ttl = settings.MERLYNNE_RESULTS_TTL
self.passthrough = kw.get('passthrough', False)
self.q = queues.get('recipe')
def bulkload(data, **kw):
"""
Bulk Load any data.
"""
kw['src'] = kw.pop('q_src', kw.pop('src', None))
if not kw['src']:
raise ValueError('Missing src.')
job_id = gen_uuid()
# set queue defaults
qkw = dict(
queued=kw.pop('queued', True),
job_id=job_id,
timeout=kw.pop('q_timeout', 1000),
serializer=kw.pop('q_serializer', 'json'),
result_ttl=kw.pop('q_result_ttl', 60),
kwargs_ttl=kw.pop('q_kwargs_ttl', 120),
name=kw.pop('q_name', 'bulk'),
max_workers=kw.pop('q_max_workers', MAX_WORKERS),
job_key_fmt=kw.pop('q_job_key', 'rq:{src}:bulk:'.format(**kw)+"{}"),
chunk_size=kw.pop('q_chunk_size', MAX_CHUNK_SIZE)
)
kw.update({'queued': qkw.get('queued', True)})
# if this is not a queued job, just run ingest.
if not qkw.get('queued'):
return ingest.source(data, **kw)
q = queues.get(qkw.pop('name', 'bulk'))
# store the data + kwargs in redis temporarily
# this makes the enqueuing process much, much more
# efficient by allowing us to only pass a single key
# into the queue rather than a massive dump of data
# however it also means that all kwargs must be
# json serializable
job_key = qkw['job_key_fmt'].format(job_id)
job = {'data': data, 'kw': kw}
if qkw['serializer'] == 'json':
job = obj_to_json(job)
elif qkw['serializer'] == 'pickle':
job = obj_to_pickle(job)
rds.set(job_key, job, ex=qkw['kwargs_ttl'])
q.enqueue(bulkworker, job_id, **qkw)
return job_id
def bulkload(data, **kw):
"""
Bulk Load any data.
"""
kw['src'] = kw.pop('q_src', kw.pop('src', None))
if not kw['src']:
raise ValueError('Missing src.')
job_id = gen_uuid()
# set queue defaults
qkw = dict(queued=kw.pop('queued', True),
job_id=job_id,
timeout=kw.pop('q_timeout', 1000),
serializer=kw.pop('q_serializer', 'json'),
result_ttl=kw.pop('q_result_ttl', 60),
kwargs_ttl=kw.pop('q_kwargs_ttl', 120),
name=kw.pop('q_name', 'bulk'),
max_workers=kw.pop('q_max_workers', MAX_WORKERS),
job_key_fmt=kw.pop('q_job_key',
'rq:{src}:bulk:'.format(**kw) + "{}"),
chunk_size=kw.pop('q_chunk_size', MAX_CHUNK_SIZE))
kw.update({'queued': qkw.get('queued', True)})
# if this is not a queued job, just run ingest.
if not qkw.get('queued'):
return ingest.source(data, **kw)
q = queues.get(qkw.pop('name', 'bulk'))
# store the data + kwargs in redis temporarily
# this makes the enqueuing process much, much more
# efficient by allowing us to only pass a single key
# into the queue rather than a massive dump of data
# however it also means that all kwargs must be
# json serializable
job_key = qkw['job_key_fmt'].format(job_id)
job = {'data': data, 'kw': kw}
if qkw['serializer'] == 'json':
job = obj_to_json(job)
elif qkw['serializer'] == 'pickle':
job = obj_to_pickle(job)
rds.set(job_key, job, ex=qkw['kwargs_ttl'])
q.enqueue(bulkworker, job_id, **qkw)
return job_id
def get_status(user, job_id):
"""
Get the status of a queued job.
"""
# parse args.
queue = request.args.get('queue')
if not queue:
raise RequestError(
'You must pass in the queue name to fetch a job\'s status')
if not queue in queues:
raise RequestError('"{}" is not a valid queue.'.format(queue))
q = queues.get(queue)
job = q.fetch_job(job_id)
if not job:
raise RequestError('A job with ID {} does not exist'.format(job_id))
# fetch metadata about this job
# from the session
# parse args.
started = request.args.get('started')
orig_url = request.args.get('orig_url')
if started:
started = dates.parse_iso(started)
# format return value
ret = {
'job_id': job_id,
'queue': queue,
'status': None,
'started': started,
'orig_url': orig_url
}
# determine time since start
if started:
ret['time_since_start'] = (dates.now() - started).seconds
# determine status
if job.is_queued:
ret['status'] = 'queued'
if job.is_started:
ret['status'] = 'running'
if job.is_failed:
ret['status'] = 'error'
ret['message'] = "An unknown error occurred."
if job.is_finished:
rv = job.return_value
# job will return true if successful
if rv is True:
ret['status'] = 'success'
# job will return an error if unsuccessful
else:
ret['status'] = 'error'
ret['message'] = rv.message
return jsonify(ret)
def get_status(user, job_id):
"""
Get the status of a queued job.
"""
# parse args.
queue = request.args.get('queue')
if not queue:
raise RequestError(
'You must pass in the queue name to fetch a job\'s status')
if not queue in queues:
raise RequestError(
'"{}" is not a valid queue.'
.format(queue))
q = queues.get(queue)
job = q.fetch_job(job_id)
if not job:
raise RequestError(
'A job with ID {} does not exist'
.format(job_id))
# fetch metadata about this job
# from the session
# parse args.
started = request.args.get('started')
orig_url = request.args.get('orig_url')
if started:
started = dates.parse_iso(started)
# format return value
ret = {
'job_id': job_id,
'queue': queue,
'status': None,
'started': started,
'orig_url': orig_url
}
# determine time since start
if started:
ret['time_since_start'] = (dates.now() - started).seconds
# determine status
if job.is_queued:
ret['status'] = 'queued'
if job.is_started:
ret['status'] = 'running'
if job.is_failed:
ret['status'] = 'error'
ret['message'] = "An unknown error occurred."
if job.is_finished:
rv = job.return_value
# job will return true if successful
if rv is True:
ret['status'] = 'success'
# job will return an error if unsuccessful
else:
ret['status'] = 'error'
ret['message'] = str(rv.message)
return jsonify(ret)
class BulkLoader(object):
__module__ = 'newslynx.tasks.bulk'
returns = None # either "model" or "query"
timeout = 1000 # seconds
result_ttl = 60 # seconds
kwargs_ttl = 1000 # in case there is a backup in the queue
max_workers = 7
concurrent = True
kwargs_key = 'rq:kwargs:{}'
q = queues.get('bulk')
redis = rds
def load_one(self, item, **kw):
"""
The method to overwrite.
"""
raise NotImplemented
def _load_one(self, item, **kw):
"""
A wrapper which will catch errors
and bubble them up
"""
try:
return self.load_one(item, **kw)
except Exception as e:
return Exception(e.message)
def _handle_errors(self, errors):
if not isinstance(errors, list):
errors = [errors]
return RequestError('There was an error while bulk uploading: '
'{}'.format(errors[0].message))
def load_all(self, kwargs_key):
"""
Do the work.
"""
start = time.time()
try:
# create a session specific to this task
session = gen_session()
# get the inputs from redis
kwargs = self.redis.get(kwargs_key)
if not kwargs:
raise InternalServerError(
'An unexpected error occurred while processing bulk upload.'
)
kwargs = pickle_to_obj(kwargs)
data = kwargs.get('data')
kw = kwargs.get('kw')
# delete them
self.redis.delete(kwargs_key)
outputs = []
errors = []
fx = partial(self._load_one, **kw)
if self.concurrent:
pool = Pool(min([len(data), self.max_workers]))
for res in pool.imap_unordered(fx, data):
if isinstance(res, Exception):
errors.append(res)
else:
outputs.append(res)
else:
for item in data:
res = fx(item)
if isinstance(res, Exception):
errors.append(res)
else:
outputs.append(res)
# return errors
if len(errors):
self._handle_errors(errors)
# add objects and execute
if self.returns == 'model':
for o in outputs:
if o is not None:
try:
session.add(o)
session.commit(o)
except Exception as e:
self._handle_errors(e)
# union all queries
elif self.returns == 'query':
for query in outputs:
if query is not None:
try:
session.execute(query)
except Exception as e:
self._handle_errors(e)
try:
session.commit()
except Exception as e:
session.rollback()
session.remove()
self._handle_errors(e)
# return true if everything worked.
session.close()
return True
except JobTimeoutException:
end = time.time()
return InternalServerError(
'Bulk loading timed out after {} seconds'.format(end - start))
def run(self, data, **kw):
# store the data + kwargs in redis temporarily
# this makes the enqueuing process much, much more
# efficient by allowing us to only pass a single key
# into the queue rather than a massive dump of data
# however it also means that all kwargs must be
# json serializable
job_id = gen_uuid()
kwargs_key = self.kwargs_key.format(job_id)
kwargs = {'data': data, 'kw': kw}
self.redis.set(kwargs_key, obj_to_pickle(kwargs), ex=self.kwargs_ttl)
# send the job to the task queue
self.q.enqueue(self.load_all,
kwargs_key,
job_id=job_id,
timeout=self.timeout,
result_ttl=self.result_ttl)
return job_id
