openstack/swift
Code Issues Proposed changes

Added exception catchalls for auditors.

Browse Source 
Also, changed some os.listdir calls to a new
swift.common.utils.listdir call which returns an empty list for
non-existent paths.

Change-Id: I89964636899d39bc07b1ecf2688786ffca67bf17
This commit is contained in:
gholt 2011-12-15 08:06:39 +00:00
parent dfb9a9f0a3
commit 2eb0eb510b
4 changed files with 56 additions and 40 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
swift/account/auditor.py
View File

@ -44,21 +44,23 @@ class AccountAuditor(Daemon):
time.sleep(random() * self.interval)
while True:
begin = time.time()
all_locs = audit_location_generator(self.devices,
account_server.DATADIR,
mount_check=self.mount_check,
logger=self.logger)
for path, device, partition in all_locs:
self.account_audit(path)
if time.time() - reported >= 3600: # once an hour
self.logger.info(_('Since %(time)s: Account audits: '
'%(passed)s passed audit, %(failed)s failed audit'),
{'time': time.ctime(reported),
'passed': self.account_passes,
'failed': self.account_failures})
reported = time.time()
self.account_passes = 0
self.account_failures = 0
try:
all_locs = audit_location_generator(self.devices,
account_server.DATADIR, mount_check=self.mount_check,
logger=self.logger)
for path, device, partition in all_locs:
self.account_audit(path)
if time.time() - reported >= 3600: # once an hour
self.logger.info(_('Since %(time)s: Account audits: '
'%(passed)s passed audit, %(failed)s failed audit'),
{'time': time.ctime(reported),
'passed': self.account_passes,
'failed': self.account_failures})
reported = time.time()
self.account_passes = 0
self.account_failures = 0
except (Exception, Timeout):
self.logger.exception(_('ERROR auditing'))
elapsed = time.time() - begin
if elapsed < self.interval:
time.sleep(self.interval - elapsed)

25
swift/common/utils.py
View File

@ -731,11 +731,11 @@ def iter_devices_partitions(devices_dir, item_type):
:param item_type: One of 'accounts', 'containers', or 'objects'
:returns: Each iteration returns a tuple of (device, partition)
"""
devices = os.listdir(devices_dir)
devices = listdir(devices_dir)
shuffle(devices)
devices_partitions = []
for device in devices:
partitions = os.listdir(os.path.join(devices_dir, device, item_type))
partitions = listdir(os.path.join(devices_dir, device, item_type))
shuffle(partitions)
devices_partitions.append((device, iter(partitions)))
yielded = True
@ -757,7 +757,7 @@ def unlink_older_than(path, mtime):
:mtime: timestamp of oldest file to keep
"""
if os.path.exists(path):
for fname in os.listdir(path):
for fname in listdir(path):
fpath = os.path.join(path, fname)
try:
if os.path.getmtime(fpath) < mtime:
@ -929,7 +929,7 @@ def audit_location_generator(devices, datadir, mount_check=True, logger=None):
on devices
:param logger: a logger object
'''
device_dir = os.listdir(devices)
device_dir = listdir(devices)
# randomize devices in case of process restart before sweep completed
shuffle(device_dir)
for device in device_dir:
@ -942,22 +942,22 @@ def audit_location_generator(devices, datadir, mount_check=True, logger=None):
datadir_path = os.path.join(devices, device, datadir)
if not os.path.exists(datadir_path):
continue
partitions = os.listdir(datadir_path)
partitions = listdir(datadir_path)
for partition in partitions:
part_path = os.path.join(datadir_path, partition)
if not os.path.isdir(part_path):
continue
suffixes = os.listdir(part_path)
suffixes = listdir(part_path)
for suffix in suffixes:
suff_path = os.path.join(part_path, suffix)
if not os.path.isdir(suff_path):
continue
hashes = os.listdir(suff_path)
hashes = listdir(suff_path)
for hsh in hashes:
hash_path = os.path.join(suff_path, hsh)
if not os.path.isdir(hash_path):
continue
for fname in sorted(os.listdir(hash_path),
for fname in sorted(listdir(hash_path),
reverse=True):
path = os.path.join(hash_path, fname)
yield path, device, partition
@ -1106,3 +1106,12 @@ def dump_recon_cache(cache_key, cache_value, cache_file, lock_timeout=2):
except OSError, err:
if err.errno != errno.ENOENT:
raise
def listdir(path):
try:
return os.listdir(path)
except OSError, err:
if err.errno != errno.ENOENT:
raise
return []

34
swift/container/auditor.py
View File

@ -45,22 +45,24 @@ class ContainerAuditor(Daemon):
time.sleep(random() * self.interval)
while True:
begin = time.time()
all_locs = audit_location_generator(self.devices,
container_server.DATADIR,
mount_check=self.mount_check,
logger=self.logger)
for path, device, partition in all_locs:
self.container_audit(path)
if time.time() - reported >= 3600: # once an hour
self.logger.info(
_('Since %(time)s: Container audits: %(pass)s passed '
'audit, %(fail)s failed audit'),
{'time': time.ctime(reported),
'pass': self.container_passes,
'fail': self.container_failures})
reported = time.time()
self.container_passes = 0
self.container_failures = 0
try:
all_locs = audit_location_generator(self.devices,
container_server.DATADIR, mount_check=self.mount_check,
logger=self.logger)
for path, device, partition in all_locs:
self.container_audit(path)
if time.time() - reported >= 3600: # once an hour
self.logger.info(
_('Since %(time)s: Container audits: %(pass)s '
'passed audit, %(fail)s failed audit'),
{'time': time.ctime(reported),
'pass': self.container_passes,
'fail': self.container_failures})
reported = time.time()
self.container_passes = 0
self.container_failures = 0
except (Exception, Timeout):
self.logger.exception(_('ERROR auditing'))
elapsed = time.time() - begin
if elapsed < self.interval:
time.sleep(self.interval - elapsed)

5
swift/obj/auditor.py
View File

@ -204,7 +204,10 @@ class ObjectAuditor(Daemon):
if parent:
kwargs['zero_byte_fps'] = zbo_fps or self.conf_zero_byte_fps
while True:
self.run_once(**kwargs)
try:
self.run_once(**kwargs)
except (Exception, Timeout):
self.logger.exception(_('ERROR auditing'))
self._sleep()
def run_once(self, *args, **kwargs):