def filter_missing():
futures = {}
cutoff = time.time() - 60*60
for filepath, filestats in gcs.list_files(gs_bucket):
# If the file was uploaded in the last hour, ignore it.
if filestats.st_ctime >= cutoff:
continue
# This must match the logic in model.entry_key(). Since this request
# will in practice touch every item, do not use memcache since it'll
# mess it up by loading every items in it.
# TODO(maruel): Batch requests to use get_multi_async() similar to
# datastore_utils.page_queries().
future = model.entry_key_from_id(filepath).get_async(
use_cache=False, use_memcache=False)
futures[future] = filepath
if len(futures) > 20:
future = ndb.Future.wait_any(futures)
filepath = futures.pop(future)
if future.get_result():
continue
yield filepath
while futures:
future = ndb.Future.wait_any(futures)
filepath = futures.pop(future)
if future.get_result():
continue
yield filepath
def filter_missing():
futures = {}
cutoff = time.time() - 60 * 60
for filepath, filestats in gcs.list_files(gs_bucket):
# If the file was uploaded in the last hour, ignore it.
if filestats.st_ctime >= cutoff:
continue
# This must match the logic in model.get_entry_key(). Since this request
# will in practice touch every item, do not use memcache since it'll
# mess it up by loading every items in it.
# TODO(maruel): Batch requests to use get_multi_async() similar to
# datastore_utils.page_queries().
future = model.entry_key_from_id(filepath).get_async(
use_cache=False, use_memcache=False)
futures[future] = filepath
if len(futures) > 20:
future = ndb.Future.wait_any(futures)
filepath = futures.pop(future)
if future.get_result():
continue
yield filepath
while futures:
future = ndb.Future.wait_any(futures)
filepath = futures.pop(future)
if future.get_result():
continue
yield filepath
def post(self):
logging.info('Deleting ContentEntry')
incremental_delete(model.ContentEntry.query().iter(keys_only=True),
ndb.delete_multi_async)
gs_bucket = config.settings().gs_bucket
logging.info('Deleting GS bucket %s', gs_bucket)
incremental_delete(
(i[0] for i in gcs.list_files(gs_bucket)),
lambda filenames: gcs.delete_files(gs_bucket, filenames))
logging.info('Flushing memcache')
# High priority (.isolated files) are cached explicitly. Make sure ghosts
# are zapped too.
memcache.flush_all()
logging.info('Finally done!')
def post(self):
logging.info('Deleting ContentEntry')
incremental_delete(
model.ContentEntry.query().iter(keys_only=True),
ndb.delete_multi_async)
gs_bucket = config.settings().gs_bucket
logging.info('Deleting GS bucket %s', gs_bucket)
incremental_delete(
(i[0] for i in gcs.list_files(gs_bucket)),
lambda filenames: gcs.delete_files(gs_bucket, filenames))
logging.info('Flushing memcache')
# High priority (.isolated files) are cached explicitly. Make sure ghosts
# are zapped too.
memcache.flush_all()
logging.info('Finally done!')
def _yield_orphan_gcs_files(gs_bucket):
"""Iterates over the whole GCS bucket for unreferenced files.
Finds files in GCS that are not referred to by a ContentEntry.
Only return files at least 1 day old to reduce the risk of failure.
Yields:
path of unreferenced files in the bucket
"""
good = 0
orphaned = 0
size_good = 0
size_orphaned = 0
# pylint: disable=too-many-nested-blocks
try:
futures = {}
cutoff = time.time() - 24 * 60 * 60
# https://cloud.google.com/appengine/docs/standard/python/googlecloudstorageclient/gcsfilestat_class
for filepath, filestats in gcs.list_files(gs_bucket):
# If the file was uploaded in the last hour, ignore it.
if filestats.st_ctime >= cutoff:
continue
# This must match the logic in model.get_entry_key(). Since this request
# will touch every ContentEntry, do not use memcache since it'll
# overflow it up by loading every items in it.
try:
# TODO(maruel): Handle non-ascii files, in practice they cannot be
# digests so they must be deleted anyway.
key = model.entry_key_from_id(str(filepath))
except AssertionError:
# It's not even a valid entry.
orphaned += 1
size_orphaned += filestats.st_size
yield filepath
continue
futures[key.get_async(use_memcache=False)] = (filepath, filestats)
if len(futures) > 100:
ndb.Future.wait_any(futures)
tmp = {}
# pylint: disable=redefined-outer-name
for f, (filepath, filestats) in futures.items():
if f.done():
if not f.get_result():
# Orphaned, delete.
orphaned += 1
size_orphaned += filestats.st_size
yield filepath
else:
good += 1
size_good += filestats.st_size
else:
tmp[f] = (filepath, filestats)
futures = tmp
while futures:
ndb.Future.wait_any(futures)
tmp = {}
for f, (filepath, filestats) in futures.items():
if f.done():
if not f.get_result():
# Orphaned, delete.
orphaned += 1
size_orphaned += filestats.st_size
yield filepath
else:
good += 1
size_good += filestats.st_size
else:
# Not done yet.
tmp[f] = (filepath, filestats)
futures = tmp
finally:
size_good_tb = size_good / 1024. / 1024. / 1024. / 1024.
size_orphaned_gb = size_orphaned / 1024. / 1024. / 1024.
logging.info(
'Found:\n'
'- %d good GCS files; %d bytes (%.1fTiB)\n'
'- %d orphaned files; %d bytes (%.1fGiB)', good, size_good,
size_good_tb, orphaned, size_orphaned, size_orphaned_gb)
