def purge_instance(instance_id, nb_to_keep):
instance = models.Instance.query.get(instance_id)
logger = get_instance_logger(instance)
logger.info('purge of backup directories for %s', instance.name)
instance_config = load_instance_config(instance.name)
backups = set(glob.glob('{}/*'.format(instance_config.backup_directory)))
logger.info('backups are: %s', backups)
# we add the realpath not to have problems with double / or stuff like that
loaded = set(
os.path.realpath(os.path.dirname(dataset.name)) for dataset in instance.last_datasets(nb_to_keep)
)
logger.info('loaded data are: %s', loaded)
running = set(os.path.realpath(os.path.dirname(dataset.name)) for dataset in instance.running_datasets())
logger.info('running bina are: %s', running)
to_remove = [os.path.join(instance_config.backup_directory, f) for f in backups - loaded - running]
missing = [l for l in loaded if l not in backups]
if missing:
logger.error(
"MISSING backup files! impossible to find %s in the backup dir, "
"we skip the purge, repair ASAP to fix the purge",
missing,
)
return
logger.info('we remove: %s', to_remove)
for path in to_remove:
shutil.rmtree(path)
def heartbeat():
"""
send a heartbeat to all kraken
"""
logging.info('ping krakens!!')
with kombu.Connection(
current_app.config['CELERY_BROKER_URL']) as connection:
instances = models.Instance.query_existing().all()
task = task_pb2.Task()
task.action = task_pb2.HEARTBEAT
for instance in instances:
try:
config = load_instance_config(instance.name)
exchange = kombu.Exchange(config.exchange,
'topic',
durable=True)
producer = connection.Producer(exchange=exchange)
producer.publish(task.SerializeToString(),
routing_key='{}.task.heartbeat'.format(
instance.name))
except Exception as e:
logging.error(
"Could not ping krakens for instance {i}: {e}".format(
i=instance, e=e))
def run(self, name_=""):
if name_:
instances = models.Instance.query.filter_by(name=name_).all()
else:
instances = models.Instance.query.all()
if not instances:
logging.getLogger(__name__).\
error("Unable to find any instance for name '{name}'"
.format(name=name_))
return
#TODO: create a real job
job_id = 1
instances_name = [instance.name for instance in instances]
for instance_name in instances_name:
instance_config = None
try:
instance_config = load_instance_config(instance_name)
except ValueError:
logging.getLogger(__name__).\
info("Unable to find instance " + instance_name)
continue
aggregate_places(instance_config, job_id)
job_id += 1
def update_data():
for instance in models.Instance.query_existing().all():
current_app.logger.debug("Update data of : {}".format(instance.name))
instance_config = load_instance_config(instance.name)
files = glob.glob(instance_config.source_directory + "/*")
if files:
import_data(files, instance, backup_file=True)
def build_data(instance):
job = models.Job()
job.instance = instance
job.state = 'pending'
instance_config = load_instance_config(instance.name)
models.db.session.add(job)
models.db.session.commit()
chain(ed2nav.si(instance_config, job.id, None), finish_job.si(job.id)).delay()
current_app.logger.info("Job build data of : %s queued"%instance.name)
def reload_kraken(instance_id):
instance = models.Instance.query.get(instance_id)
job = models.Job()
job.instance = instance
job.state = 'pending'
instance_config = load_instance_config(instance.name)
models.db.session.add(job)
models.db.session.commit()
chain(reload_data.si(instance_config, job.id), finish_job.si(job.id)).delay()
logging.info("Task reload kraken for instance {} queued".format(instance.name))
def reload_at(instance_id):
instance = models.Instance.query.get(instance_id)
job = models.Job()
job.instance = instance
job.state = 'pending'
instance_config = load_instance_config(instance.name)
models.db.session.add(job)
models.db.session.commit()
chain(nav2rt.si(instance_config, job.id),
reload_data.si(instance_config, job.id),
finish_job.si(job.id)).delay()
def scan_instances():
for instance_file in glob.glob(current_app.config['INSTANCES_DIR'] + '/*.ini'):
instance_name = os.path.basename(instance_file).replace('.ini', '')
instance = models.Instance.query.filter_by(name=instance_name).first()
if not instance:
current_app.logger.info('new instances detected: %s', instance_name)
instance = models.Instance(name=instance_name)
instance_config = load_instance_config(instance.name)
instance.is_free = instance_config.is_free
models.db.session.add(instance)
models.db.session.commit()
def build_all_data():
for instance in models.Instance.query.all():
job = models.Job()
job.instance = instance
job.state = 'pending'
instance_config = load_instance_config(instance.name)
models.db.session.add(job)
models.db.session.commit()
chain(ed2nav.si(instance_config, job.id),
nav2rt.si(instance_config, job.id)).delay()
current_app.logger.info("Job build data of : %s queued" %
instance.name)
def update_data():
for instance in models.Instance.query_existing().all():
current_app.logger.debug("Update data of : {}".format(instance.name))
instance_config = None
try:
instance_config = load_instance_config(instance.name)
except:
current_app.logger.exception("impossible to load instance configuration for %s", instance.name)
# Do not stop the task if only one instance is missing
continue
files = glob.glob(instance_config.source_directory + "/*")
if files:
import_data(files, instance, backup_file=True)
def send_to_mimir(instance, filename):
"""
:param instance: instance to receive the data
:param filename: file to inject towards mimir
- create a job with a data_set
- data injection towards mimir(stops2mimir, ntfs2mimir)
returns action list
"""
# This test is to avoid creating a new job if there is no action on mimir.
if not (instance.import_ntfs_in_mimir or instance.import_stops_in_mimir):
return []
actions = []
job = models.Job()
instance_config = load_instance_config(instance.name)
job.instance = instance
job.state = 'running'
dataset = models.DataSet()
dataset.family_type = 'mimir'
dataset.type = 'fusio'
# currently the name of a dataset is the path to it
dataset.name = filename
models.db.session.add(dataset)
job.data_sets.append(dataset)
models.db.session.add(job)
models.db.session.commit()
# Import ntfs in Mimir
if instance.import_ntfs_in_mimir:
actions.append(
ntfs2mimir.si(instance_config,
filename,
job.id,
dataset_uid=dataset.uid))
# Import stops in Mimir
# if we are loading pt data we might want to load the stops to autocomplete
if instance.import_stops_in_mimir and not instance.import_ntfs_in_mimir:
actions.append(
stops2mimir.si(instance_config,
filename,
job.id,
dataset_uid=dataset.uid))
actions.append(finish_job.si(job.id))
return actions
def _init(self):
instances = models.Instance.query.all()
self.connection = kombu.Connection(
current_app.config['CELERY_BROKER_URL'])
for instance in instances:
#initialize the last relaod at the minimum date possible
self.last_reload[instance.id] = datetime(1, 1, 1)
config = load_instance_config(instance.name)
exchange = kombu.Exchange(config.exchange, 'topic', durable=True)
for topic in config.rt_topics:
self.topics_to_instances[topic].append(instance)
queue = kombu.Queue(exchange=exchange,
durable=True,
routing_key=topic)
self.queues.append(queue)
:param files: files to import
:param instance: instance to receive the data
:param backup_file: If True the files are moved to a backup directory, else they are not moved
:param async: If True all jobs are run in background, else the jobs are run in sequence the function will only return when all of them are finish
:param reload: If True kraken would be reload at the end of the treatment
run the whole data import process:
- data import in bdd (fusio2ed, gtfs2ed, poi2ed, ...)
- export bdd to nav file
- update the jormungandr db with the new data for the instance
- reload the krakens
"""
actions = []
job = models.Job()
instance_config = load_instance_config(instance.name)
job.instance = instance
job.state = 'pending'
task = {
'gtfs': gtfs2ed,
'fusio': fusio2ed,
'osm': osm2ed,
'geopal': geopal2ed,
'fare': fare2ed,
'poi': poi2ed,
'synonym': synonym2ed,
'shape': shape2ed,
}
for _file in files:
filename = None
def import_data(
files, instance, backup_file, asynchronous=True, reload=True, custom_output_dir=None, skip_mimir=False
):
"""
import the data contains in the list of 'files' in the 'instance'
:param files: files to import
:param instance: instance to receive the data
:param backup_file: If True the files are moved to a backup directory, else they are not moved
:param asynchronous: If True all jobs are run in background, else the jobs are run in sequence the function
will only return when all of them are finish
:param reload: If True kraken would be reload at the end of the treatment
:param custom_output_dir: subdirectory for the nav file created. If not given, the instance default one is taken
:param skip_mimir: skip importing data into mimir
run the whole data import process:
- data import in bdd (fusio2ed, gtfs2ed, poi2ed, ...)
- export bdd to nav file
- update the jormungandr db with the new data for the instance
- reload the krakens
"""
actions = []
job = models.Job()
instance_config = load_instance_config(instance.name)
job.instance = instance
job.state = 'running'
task = {
'gtfs': gtfs2ed,
'fusio': fusio2ed,
'osm': osm2ed,
'geopal': geopal2ed,
'fare': fare2ed,
'poi': poi2ed,
'synonym': synonym2ed,
'shape': shape2ed,
}
for _file in files:
filename = None
dataset = models.DataSet()
# NOTE: for the moment we do not use the path to load the data here
# but we'll need to refactor this to take it into account
try:
dataset.type, _ = utils.type_of_data(_file)
dataset.family_type = utils.family_of_data(dataset.type)
except Exception:
if backup_file:
move_to_backupdirectory(_file, instance_config.backup_directory)
current_app.logger.debug(
"Corrupted source file : {} moved to {}".format(_file, instance_config.backup_directory)
)
continue
if dataset.type in task:
if backup_file:
filename = move_to_backupdirectory(_file, instance_config.backup_directory)
else:
filename = _file
actions.append(task[dataset.type].si(instance_config, filename, dataset_uid=dataset.uid))
else:
# unknown type, we skip it
current_app.logger.debug("unknown file type: {} for file {}".format(dataset.type, _file))
continue
# currently the name of a dataset is the path to it
dataset.name = filename
models.db.session.add(dataset)
job.data_sets.append(dataset)
if actions:
models.db.session.add(job)
models.db.session.commit()
# We pass the job id to each tasks, but job need to be commited for having an id
for action in actions:
action.kwargs['job_id'] = job.id
# Create binary file (New .nav.lz4)
binarisation = [ed2nav.si(instance_config, job.id, custom_output_dir)]
actions.append(chain(*binarisation))
# Reload kraken with new data after binarisation (New .nav.lz4)
if reload:
actions.append(reload_data.si(instance_config, job.id))
if not skip_mimir:
for dataset in job.data_sets:
actions.extend(send_to_mimir(instance, dataset.name, dataset.family_type))
else:
current_app.logger.info("skipping mimir import")
actions.append(finish_job.si(job.id))
if asynchronous:
return chain(*actions).delay()
else:
# all job are run in sequence and import_data will only return when all the jobs are finish
return chain(*actions).apply()
def import_data(
files,
instance,
backup_file,
asynchronous=True,
reload=True,
custom_output_dir=None,
skip_mimir=False,
skip_2ed=False,
):
"""
import the data contains in the list of 'files' in the 'instance'
:param files: files to import
:param instance: instance to receive the data
:param backup_file: If True the files are moved to a backup directory, else they are not moved
:param asynchronous: If True all jobs are run in background, else the jobs are run in sequence the function
will only return when all of them are finish
:param reload: If True kraken would be reload at the end of the treatment
:param custom_output_dir: subdirectory for the nav file created. If not given, the instance default one is taken
:param skip_mimir: skip importing data into mimir
:param skip_2ed: skip inserting last_load_dataset files into ed database
run the whole data import process:
- data import in bdd (fusio2ed, gtfs2ed, poi2ed, ...)
- export bdd to nav file
- update the jormungandr db with the new data for the instance
- reload the krakens
"""
actions = []
job = models.Job()
instance_config = load_instance_config(instance.name)
job.instance = instance
job.state = 'running'
task = {
'gtfs': gtfs2ed,
'fusio': fusio2ed,
'osm': osm2ed,
'geopal': geopal2ed,
'fare': fare2ed,
'poi': poi2ed,
'synonym': synonym2ed,
'shape': shape2ed,
}
def process_ed2nav():
models.db.session.add(job)
models.db.session.commit()
# We pass the job id to each tasks, but job need to be commited for having an id
for action in actions:
action.kwargs['job_id'] = job.id
# Create binary file (New .nav.lz4)
binarisation = [ed2nav.si(instance_config, job.id, custom_output_dir)]
actions.append(chain(*binarisation))
# Reload kraken with new data after binarisation (New .nav.lz4)
if reload:
actions.append(reload_data.si(instance_config, job.id))
if not skip_mimir:
for dataset in job.data_sets:
actions.extend(
send_to_mimir(instance, dataset.name, dataset.family_type))
else:
current_app.logger.info("skipping mimir import")
actions.append(finish_job.si(job.id))
# We should delete old backup directories related to this instance
actions.append(
purge_instance.si(
instance.id,
current_app.config['DATASET_MAX_BACKUPS_TO_KEEP']))
if asynchronous:
return chain(*actions).delay()
else:
# all job are run in sequence and import_data will only return when all the jobs are finish
return chain(*actions).apply()
if skip_2ed:
# For skip_2ed, skip inserting last_load_dataset files into ed database
return process_ed2nav()
for _file in files:
filename = None
dataset = models.DataSet()
# NOTE: for the moment we do not use the path to load the data here
# but we'll need to refactor this to take it into account
try:
dataset.type, _ = utils.type_of_data(_file)
dataset.family_type = utils.family_of_data(dataset.type)
except Exception:
if backup_file:
move_to_backupdirectory(_file,
instance_config.backup_directory)
current_app.logger.debug(
"Corrupted source file : {} moved to {}".format(
_file, instance_config.backup_directory))
continue
if dataset.type in task:
if backup_file:
filename = move_to_backupdirectory(
_file,
instance_config.backup_directory,
manage_sp_char=True)
else:
filename = _file
has_pt_planner_loki = (
hasattr(instance, 'pt_planners_configurations')
and "loki" in instance.pt_planners_configurations)
if has_pt_planner_loki:
loki_data_source = instance.pt_planners_configurations.get(
'loki', {}).get('data_source')
if loki_data_source is not None:
if loki_data_source == "minio":
if dataset.type == "fusio":
actions.append(
fusio2s3.si(instance_config,
filename,
dataset_uid=dataset.uid))
if dataset.type == "gtfs":
actions.append(
gtfs2s3.si(instance_config,
filename,
dataset_uid=dataset.uid))
elif loki_data_source == "local" and dataset.type in [
"fusio", "gtfs"
]:
zip_file = zip_if_needed(filename)
dest = os.path.join(
os.path.dirname(instance_config.target_file),
"ntfs.zip")
shutil.copy(zip_file, dest)
else:
current_app.logger.debug(
"unknown loki data_source '{}' for coverage '{}'".
format(loki_data_source, instance.name))
actions.append(task[dataset.type].si(instance_config,
filename,
dataset_uid=dataset.uid))
else:
# unknown type, we skip it
current_app.logger.debug(
"unknown file type: {} for file {}".format(
dataset.type, _file))
continue
# currently the name of a dataset is the path to it
dataset.name = filename
dataset.state = "pending"
models.db.session.add(dataset)
job.data_sets.append(dataset)
if actions:
return process_ed2nav()
def update_data():
for instance in models.Instance.query.all():
current_app.logger.debug("Update data of : %s" % instance.name)
instance_config = load_instance_config(instance.name)
files = glob.glob(instance_config.source_directory + "/*")
actions = []
job = models.Job()
job.instance = instance
job.state = 'pending'
for _file in files:
dataset = models.DataSet()
filename = None
dataset.type = type_of_data(_file)
if dataset.type == 'gtfs':
filename = move_to_backupdirectory(
_file, instance_config.backup_directory)
actions.append(gtfs2ed.si(instance_config, filename))
elif dataset.type == 'fusio':
filename = move_to_backupdirectory(
_file, instance_config.backup_directory)
actions.append(fusio2ed.si(instance_config, filename))
elif dataset.type == 'osm':
filename = move_to_backupdirectory(
_file, instance_config.backup_directory)
actions.append(osm2ed.si(instance_config, filename))
elif dataset.type == 'geopal':
filename = move_to_backupdirectory(
_file, instance_config.backup_directory)
actions.append(geopal2ed.si(instance_config, filename))
elif dataset.type == 'fare':
filename = move_to_backupdirectory(
_file, instance_config.backup_directory)
actions.append(fare2ed.si(instance_config, filename))
elif dataset.type == 'poi':
filename = move_to_backupdirectory(
_file, instance_config.backup_directory)
actions.append(poi2ed.si(instance_config, filename))
elif dataset.type == 'synonym':
filename = move_to_backupdirectory(
_file, instance_config.backup_directory)
actions.append(synonym2ed.si(instance_config, filename))
else:
#unknown type, we skip it
continue
#currently the name of a dataset is the path to it
dataset.name = filename
models.db.session.add(dataset)
job.data_sets.append(dataset)
if actions:
models.db.session.add(job)
models.db.session.commit()
for action in actions:
action.kwargs['job_id'] = job.id
#We pass the job id to each tasks, but job need to be commited for
#having an id
binarisation = [
ed2nav.si(instance_config, job.id),
nav2rt.si(instance_config, job.id)
]
aggregate = aggregate_places.si(instance_config, job.id)
#We pass the job id to each tasks, but job need to be commited for
#having an id
actions.append(group(chain(*binarisation), aggregate))
actions.append(reload_data.si(instance_config, job.id))
actions.append(finish_job.si(job.id))
chain(*actions).delay()
def bounding_shape(instance_name, shape_path):
""" Set the bounding shape to a custom value """
instance_conf = load_instance_config(instance_name)
load_bounding_shape(instance_name, instance_conf, shape_path)
def import_data(files, instance, backup_file):
"""
import the data contains in the list of 'files' in the 'instance'
:param files: files to import
:param instance: instance to receive the data
:param backup_file: If True the files are moved to a backup directory, else they are not moved
run the whole data import process:
- data import in bdd (fusio2ed, gtfs2ed, poi2ed, ...)
- export bdd to nav file
- update the jormungandr db with the new data for the instance
- reload the krakens
"""
actions = []
job = models.Job()
instance_config = load_instance_config(instance.name)
job.instance = instance
job.state = 'pending'
task = {
'gtfs': gtfs2ed,
'fusio': fusio2ed,
'osm': osm2ed,
'geopal': geopal2ed,
'fare': fare2ed,
'poi': poi2ed,
'synonym': synonym2ed,
}
for _file in files:
filename = None
dataset = models.DataSet()
dataset.type = type_of_data(_file)
if dataset.type in task:
if backup_file:
filename = move_to_backupdirectory(
_file, instance_config.backup_directory)
else:
filename = _file
actions.append(task[dataset.type].si(instance_config, filename))
else:
#unknown type, we skip it
current_app.logger.debug("unknwn file type: {} for file {}".format(
dataset.type, _file))
continue
#currently the name of a dataset is the path to it
dataset.name = filename
models.db.session.add(dataset)
job.data_sets.append(dataset)
if actions:
models.db.session.add(job)
models.db.session.commit()
for action in actions:
action.kwargs['job_id'] = job.id
#We pass the job id to each tasks, but job need to be commited for
#having an id
binarisation = [
ed2nav.si(instance_config, job.id),
nav2rt.si(instance_config, job.id)
]
aggregate = aggregate_places.si(instance_config, job.id)
#We pass the job id to each tasks, but job need to be commited for
#having an id
actions.append(group(chain(*binarisation), aggregate))
actions.append(reload_data.si(instance_config, job.id))
actions.append(finish_job.si(job.id))
chain(*actions).delay()
