def atomic_lock(lock_dir, lock_name, lifetime=60):
"""A atomic, NFS safe implementation of a file lock.
Uses `flufl.lock` under the hood. Can be used as a context manager::
with atomic_lock(settings.TMP_PATH, 'extraction-1234'):
extract_xpi(...)
:return: `True` if the lock was attained, we are owning the lock,
`False` if there is an already existing lock.
"""
lock_name = lock_name + '.lock'
count = _lock_count.get(lock_name, 0)
log.debug('Acquiring lock %s, count is %d.' % (lock_name, count))
lock_name = os.path.join(lock_dir, lock_name)
lock = flufl.lock.Lock(lock_name, lifetime=timedelta(seconds=lifetime))
try:
# set `timeout=0` to avoid any process blocking but catch the
# TimeOutError raised instead.
lock.lock(timeout=timedelta(seconds=0))
except flufl.lock.AlreadyLockedError:
# This process already holds the lock
yield False
except flufl.lock.TimeOutError:
# Some other process holds the lock.
# Let's break the lock if it has expired. Unfortunately
# there's a bug in flufl.lock so let's do this manually.
# Bug: https://gitlab.com/warsaw/flufl.lock/merge_requests/1
release_time = lock._releasetime
max_release_time = release_time + flufl.lock._lockfile.CLOCK_SLOP
if (release_time != -1 and datetime.now() > max_release_time):
# Break the lock and try to aquire again
lock._break()
lock.lock(timeout=timedelta(seconds=0))
yield lock.is_locked
else:
# Already locked
yield False
else:
# Is usually `True` but just in case there were some weird `lifetime`
# values set we return the check if we really attained the lock.
yield lock.is_locked
if lock.is_locked:
log.debug('Releasing lock %s.' % lock.details[2])
lock.unlock()
def atomic_lock(lock_dir, lock_name, lifetime=60):
"""A atomic, NFS safe implementation of a file lock.
Uses `flufl.lock` under the hood. Can be used as a context manager::
with atomic_lock(settings.TMP_PATH, 'extraction-1234'):
extract_xpi(...)
:return: `True` if the lock was attained, we are owning the lock,
`False` if there is an already existing lock.
"""
lock_name = lock_name + '.lock'
count = _lock_count.get(lock_name, 0)
log.debug('Acquiring lock %s, count is %d.' % (lock_name, count))
lock_name = os.path.join(lock_dir, lock_name)
lock = flufl.lock.Lock(lock_name, lifetime=timedelta(seconds=lifetime))
try:
# set `timeout=0` to avoid any process blocking but catch the
# TimeOutError raised instead.
lock.lock(timeout=timedelta(seconds=0))
except flufl.lock.AlreadyLockedError:
# This process already holds the lock
yield False
except flufl.lock.TimeOutError:
# Some other process holds the lock.
# Let's break the lock if it has expired. Unfortunately
# there's a bug in flufl.lock so let's do this manually.
# Bug: https://gitlab.com/warsaw/flufl.lock/merge_requests/1
release_time = lock._releasetime
max_release_time = release_time + flufl.lock._lockfile.CLOCK_SLOP
if (release_time != -1 and datetime.now() > max_release_time):
# Break the lock and try to aquire again
lock._break()
lock.lock(timeout=timedelta(seconds=0))
yield lock.is_locked
else:
# Already locked
yield False
else:
# Is usually `True` but just in case there were some weird `lifetime`
# values set we return the check if we really attained the lock.
yield lock.is_locked
if lock.is_locked:
log.debug('Releasing lock %s.' % lock.details[2])
lock.unlock()
