def __init__(self, **kwargs):
# create db environment
additional_flags = kwargs.get('flags', 0)
recovery_mode = kwargs.get('recover', 0)
if recovery_mode == 2:
additional_flags |= db.DB_RECOVER_FATAL
elif recovery_mode == 1:
additional_flags |= db.DB_RECOVER
if hasattr(db, 'DB_REGISTER'):
additional_flags |= db.DB_REGISTER
self._env = db.DBEnv()
# ability to override settings' dir for testing purposes
data_dir = kwargs.get('dir', self.dir)
self._env.set_data_dir(data_dir)
# ability to override settings' log_dir for testing purposes
log_dir = kwargs.get('log_dir', kwargs.get('dir', self.log_dir))
self._env.set_lg_dir(log_dir)
self._env.set_tx_max(self.txn_max)
if self.txn_timeout is not None:
if self.txn_timeout > 0:
self._env.set_timeout(self.txn_timeout, db.DB_SET_TXN_TIMEOUT)
else:
self._env.set_flags(db.DB_TXN_NOWAIT, 1)
if self.cache_size is not None:
self._env.set_cachesize(*self.cache_size)
if os.name != 'nt' and self.shm_key:
self._env.set_shm_key(self.shm_key)
additional_flags |= db.DB_SYSTEM_MEM
# replication settings
rep_config = settings['store'].get('rep_config', None)
init_rep = kwargs.get('init_rep', False)
if rep_config:
# in replicated environments use non-durable transactions
self._env.set_flags(db.DB_TXN_NOSYNC, 1)
additional_flags |= db.DB_INIT_REP
self._env.open(self.env_dir,
db.DB_THREAD | db.DB_INIT_MPOOL | db.DB_INIT_LOCK |
db.DB_INIT_LOG | db.DB_INIT_TXN | db.DB_CREATE |
additional_flags)
db_flags = db.DB_THREAD | db.DB_AUTO_COMMIT | db.DB_CREATE | \
db.DB_MULTIVERSION
db_mode = 0o660
if rep_config:
from porcupine.db.bsddb.replication import ReplicationService
# initialiaze replication service
self.replication_service = \
ReplicationService(self._env, rep_config)
if init_rep:
# check multiprocessing
is_multiprocess = services['main'].is_multiprocess or \
services['management'].is_multiprocess
if is_multiprocess and int(rep_config['priority']) > 0 \
and db.version() < (4, 8):
self._env.close()
self.__remove_env()
raise exceptions.ConfigurationError(
'Multiprocessing master candidates ' +
'require BerkeleyDB 4.8 or higher')
# start replication service
self.replication_service.start()
# wait for client start-up
timeout = time.time() + 20
while time.time() < timeout and \
not self.replication_service.is_master() and \
not self.replication_service.client_startup_done:
time.sleep(0.02)
timeout = time.time() + 20
while time.time() < timeout and \
not (os.path.exists(
os.path.join(self.dir, 'porcupine.db'))):
time.sleep(0.02)
else:
self.replication_service = None
# open items db
while True:
self._itemdb = db.DB(self._env)
self._itemdb.set_pagesize(2048)
try:
self._itemdb.open('porcupine.db',
'items',
dbtype=db.DB_BTREE,
mode=db_mode,
flags=db_flags)
except db.DBLockDeadlockError:
self._itemdb.close()
continue
break
# open documents db
while True:
self._docdb = db.DB(self._env)
try:
self._docdb.open('porcupine.db',
'docs',
dbtype=db.DB_HASH,
mode=db_mode,
flags=db_flags)
except db.DBLockDeadlockError:
self._docdb.close()
continue
break
# open indices
self._indices = {}
for name, unique, immutable in settings['store']['indices']:
self._indices[name] = DbIndex(self._env, self._itemdb,
name, unique, immutable, db_flags)
self._running = True
maintain = kwargs.get('maintain', False)
if maintain and self._maintenance_thread is None:
# start deadlock detector
self._maintenance_thread = Thread(target=self.__maintain,
name='DB maintenance thread')
self._maintenance_thread.start()
# start checkpoint thread
self._checkpoint_thread = Thread(target=self.__checkpoint,
name='DB checkpoint thread')
self._checkpoint_thread.start()
def __init__(self, **kwargs):
# create db environment
additional_flags = kwargs.get('flags', 0)
recovery_mode = kwargs.get('recover', 0)
if recovery_mode == 2:
additional_flags |= db.DB_RECOVER_FATAL
elif recovery_mode == 1:
additional_flags |= db.DB_RECOVER
if hasattr(db, 'DB_REGISTER'):
additional_flags |= db.DB_REGISTER
self._env = db.DBEnv()
# ability to override settings' dir for testing purposes
data_dir = kwargs.get('dir', self.dir)
self._env.set_data_dir(data_dir)
# ability to override settings' log_dir for testing purposes
log_dir = kwargs.get('log_dir', kwargs.get('dir', self.log_dir))
self._env.set_lg_dir(log_dir)
self._env.set_tx_max(self.txn_max)
if self.txn_timeout is not None:
if self.txn_timeout > 0:
self._env.set_timeout(self.txn_timeout, db.DB_SET_TXN_TIMEOUT)
else:
self._env.set_flags(db.DB_TXN_NOWAIT, 1)
if self.cache_size is not None:
self._env.set_cachesize(*self.cache_size)
if os.name != 'nt' and self.shm_key:
self._env.set_shm_key(self.shm_key)
additional_flags |= db.DB_SYSTEM_MEM
# replication settings
rep_config = settings['store'].get('rep_config', None)
init_rep = kwargs.get('init_rep', False)
if rep_config:
# in replicated environments use non-durable transactions
self._env.set_flags(db.DB_TXN_NOSYNC, 1)
additional_flags |= db.DB_INIT_REP
self._env.open(
self.env_dir,
db.DB_THREAD | db.DB_INIT_MPOOL | db.DB_INIT_LOCK | db.DB_INIT_LOG
| db.DB_INIT_TXN | db.DB_CREATE | additional_flags)
db_flags = db.DB_THREAD | db.DB_AUTO_COMMIT | db.DB_CREATE | \
db.DB_MULTIVERSION
db_mode = 0o660
if rep_config:
from porcupine.db.bsddb.replication import ReplicationService
# initialiaze replication service
self.replication_service = \
ReplicationService(self._env, rep_config)
if init_rep:
# check multiprocessing
is_multiprocess = services['main'].is_multiprocess or \
services['management'].is_multiprocess
if is_multiprocess and int(rep_config['priority']) > 0 \
and db.version() < (4, 8):
self._env.close()
self.__remove_env()
raise exceptions.ConfigurationError(
'Multiprocessing master candidates ' +
'require BerkeleyDB 4.8 or higher')
# start replication service
self.replication_service.start()
# wait for client start-up
timeout = time.time() + 20
while time.time() < timeout and \
not self.replication_service.is_master() and \
not self.replication_service.client_startup_done:
time.sleep(0.02)
timeout = time.time() + 20
while time.time() < timeout and \
not (os.path.exists(
os.path.join(self.dir, 'porcupine.db'))):
time.sleep(0.02)
else:
self.replication_service = None
# open items db
while True:
self._itemdb = db.DB(self._env)
self._itemdb.set_pagesize(2048)
try:
self._itemdb.open('porcupine.db',
'items',
dbtype=db.DB_BTREE,
mode=db_mode,
flags=db_flags)
except db.DBLockDeadlockError:
self._itemdb.close()
continue
break
# open documents db
while True:
self._docdb = db.DB(self._env)
try:
self._docdb.open('porcupine.db',
'docs',
dbtype=db.DB_HASH,
mode=db_mode,
flags=db_flags)
except db.DBLockDeadlockError:
self._docdb.close()
continue
break
# open indices
self._indices = {}
for name, unique, immutable in settings['store']['indices']:
self._indices[name] = DbIndex(self._env, self._itemdb, name,
unique, immutable, db_flags)
self._running = True
maintain = kwargs.get('maintain', False)
if maintain and self._maintenance_thread is None:
# start deadlock detector
self._maintenance_thread = Thread(target=self.__maintain,
name='DB maintenance thread')
self._maintenance_thread.start()
# start checkpoint thread
self._checkpoint_thread = Thread(target=self.__checkpoint,
name='DB checkpoint thread')
self._checkpoint_thread.start()
class DB(object):
"Berkeley DB database interface"
# data dir
dir = os.path.abspath(settings['store']['bdb_data_dir'])
if dir[-1] != '/':
dir += '/'
# log_dir
log_dir = os.path.abspath(settings['store'].get('bdb_log_dir', dir))
if log_dir[-1] != '/':
log_dir += '/'
# environment files directory
env_dir = os.path.abspath(settings['store'].get('env_dir',
os.path.abspath(settings['global']['temp_folder'])))
if env_dir[-1] != '/':
env_dir += '/'
# cache size
cache_size = settings['store'].get('cache_size', None)
# maximum concurrent transactions
# due to snapshot isolation this should be kept high enough
txn_max = settings['store'].get('max_tx', 1000)
# transaction timeout
txn_timeout = settings['store'].get('tx_timeout', None)
# shared memory key
shm_key = settings['store'].get('shm_key', None)
# maintenance (deadlock detector) thread
_maintenance_thread = None
# checkpoint thread
_checkpoint_thread = None
# trickle thread
_trickle_thread = None
# log berkeleyDB version
logger.info('BerkeleyDB version is %s' %
'.'.join(str(x) for x in db.version()))
def __init__(self, **kwargs):
# create db environment
additional_flags = kwargs.get('flags', 0)
recovery_mode = kwargs.get('recover', 0)
if recovery_mode == 2:
additional_flags |= db.DB_RECOVER_FATAL
elif recovery_mode == 1:
additional_flags |= db.DB_RECOVER
if hasattr(db, 'DB_REGISTER'):
additional_flags |= db.DB_REGISTER
self._env = db.DBEnv()
# ability to override settings' dir for testing purposes
data_dir = kwargs.get('dir', self.dir)
self._env.set_data_dir(data_dir)
# ability to override settings' log_dir for testing purposes
log_dir = kwargs.get('log_dir', kwargs.get('dir', self.log_dir))
self._env.set_lg_dir(log_dir)
self._env.set_tx_max(self.txn_max)
if self.txn_timeout is not None:
if self.txn_timeout > 0:
self._env.set_timeout(self.txn_timeout, db.DB_SET_TXN_TIMEOUT)
else:
self._env.set_flags(db.DB_TXN_NOWAIT, 1)
if self.cache_size is not None:
self._env.set_cachesize(*self.cache_size)
if os.name != 'nt' and self.shm_key:
self._env.set_shm_key(self.shm_key)
additional_flags |= db.DB_SYSTEM_MEM
# replication settings
rep_config = settings['store'].get('rep_config', None)
init_rep = kwargs.get('init_rep', False)
if rep_config:
# in replicated environments use non-durable transactions
self._env.set_flags(db.DB_TXN_NOSYNC, 1)
additional_flags |= db.DB_INIT_REP
self._env.open(self.env_dir,
db.DB_THREAD | db.DB_INIT_MPOOL | db.DB_INIT_LOCK |
db.DB_INIT_LOG | db.DB_INIT_TXN | db.DB_CREATE |
additional_flags)
db_flags = db.DB_THREAD | db.DB_AUTO_COMMIT | db.DB_CREATE | \
db.DB_MULTIVERSION
db_mode = 0o660
if rep_config:
from porcupine.db.bsddb.replication import ReplicationService
# initialiaze replication service
self.replication_service = \
ReplicationService(self._env, rep_config)
if init_rep:
# check multiprocessing
is_multiprocess = services['main'].is_multiprocess or \
services['management'].is_multiprocess
if is_multiprocess and int(rep_config['priority']) > 0 \
and db.version() < (4, 8):
self._env.close()
self.__remove_env()
raise exceptions.ConfigurationError(
'Multiprocessing master candidates ' +
'require BerkeleyDB 4.8 or higher')
# start replication service
self.replication_service.start()
# wait for client start-up
timeout = time.time() + 20
while time.time() < timeout and \
not self.replication_service.is_master() and \
not self.replication_service.client_startup_done:
time.sleep(0.02)
timeout = time.time() + 20
while time.time() < timeout and \
not (os.path.exists(
os.path.join(self.dir, 'porcupine.db'))):
time.sleep(0.02)
else:
self.replication_service = None
# open items db
while True:
self._itemdb = db.DB(self._env)
self._itemdb.set_pagesize(2048)
try:
self._itemdb.open('porcupine.db',
'items',
dbtype=db.DB_BTREE,
mode=db_mode,
flags=db_flags)
except db.DBLockDeadlockError:
self._itemdb.close()
continue
break
# open documents db
while True:
self._docdb = db.DB(self._env)
try:
self._docdb.open('porcupine.db',
'docs',
dbtype=db.DB_HASH,
mode=db_mode,
flags=db_flags)
except db.DBLockDeadlockError:
self._docdb.close()
continue
break
# open indices
self._indices = {}
for name, unique, immutable in settings['store']['indices']:
self._indices[name] = DbIndex(self._env, self._itemdb,
name, unique, immutable, db_flags)
self._running = True
maintain = kwargs.get('maintain', False)
if maintain and self._maintenance_thread is None:
# start deadlock detector
self._maintenance_thread = Thread(target=self.__maintain,
name='DB maintenance thread')
self._maintenance_thread.start()
# start checkpoint thread
self._checkpoint_thread = Thread(target=self.__checkpoint,
name='DB checkpoint thread')
self._checkpoint_thread.start()
#if hasattr(self._env, 'memp_trickle'):
# # strart memp_trickle thread
# self._trickle_thread = Thread(target=self.__trickle,
# name='DB memp_trickle thread')
# self._trickle_thread.start()
def is_open(self):
return self._running
# item operations
def get_item(self, oid):
if type(oid) != bytes:
oid = oid.encode('utf-8')
try:
return self._indices['_id'].db.get(oid,
txn=context._trans and context._trans.txn)
except UnicodeEncodeError:
return None
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
if context._trans is not None:
context._trans.abort()
raise exceptions.DBRetryTransaction
def put_item(self, item):
try:
self._itemdb.put(
pack_value(item._pid) + b'_' + pack_value(item._id),
persist.dumps(item),
context._trans.txn)
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
context._trans.abort()
raise exceptions.DBRetryTransaction
except db.DBError as e:
if e.args[0] == _err_unsupported_index_type:
raise db.DBError('Unsupported indexed data type')
else:
raise
def delete_item(self, item):
try:
self._itemdb.delete(
pack_value(item._pid) + b'_' + pack_value(item._id),
context._trans.txn)
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
context._trans.abort()
raise exceptions.DBRetryTransaction
# containers
def get_children(self, container_id):
cursor = Cursor(self._itemdb, '_pid')
cursor.set_scope(container_id)
cursor.set_range(None, None)
return cursor
def get_child_by_name(self, container_id, name):
try:
return self._indices['displayName'].db.get(
pack_value(container_id) + b'_' + pack_value(name),
txn=context._trans and context._trans.txn)
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
if context._trans is not None:
context._trans.abort()
raise exceptions.DBRetryTransaction
# external attributes
def get_external(self, id):
try:
return self._docdb.get(id.encode('ascii'),
txn=context._trans and context._trans.txn)
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
if context._trans is not None:
context._trans.abort()
raise exceptions.DBRetryTransaction
def put_external(self, id, stream):
try:
self._docdb.put(id.encode('ascii'), stream, context._trans.txn)
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
context._trans.abort()
raise exceptions.DBRetryTransaction
def delete_external(self, id):
try:
self._docdb.delete(id.encode('ascii'), context._trans.txn)
except db.DBNotFoundError:
# virtual external due to elastic schema
pass
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
context._trans.abort()
raise exceptions.DBRetryTransaction
# indices
def get_cursor_list(self, conditions):
cur_list = []
for index, value in conditions:
cursor = Cursor(self._indices[index].db, self._indices[index].name)
if isinstance(value, (list, tuple)):
reversed = (len(value) == 3 and value[2])
cursor.set_range(value[0], value[1])
if reversed:
cursor.reverse()
else:
cursor.set(value)
cur_list.append(cursor)
return cur_list
def query(self, conditions):
cur_list = self.get_cursor_list(conditions)
if len(cur_list) == 1:
return cur_list[0]
else:
c_join = Join(self._itemdb, cur_list)
return c_join
def test_conditions(self, scope, conditions):
cur_list = self.get_cursor_list(conditions)
if len(cur_list) == 1:
cursor = cur_list[0]
else:
cursor = Join(self._itemdb, cur_list)
cursor.set_scope(scope)
iterator = iter(cursor)
try:
result = bool(next(iterator))
except StopIteration:
result = False
cursor.close()
return result
# transactions
def get_transaction(self, **kwargs):
nosync = kwargs.get('nosync', False)
snapshot = kwargs.get('snapshot', False)
return Transaction(self._env, nosync, snapshot)
def __remove_env(self):
files = glob.glob(self.env_dir + '__db.*')
for file in files:
try:
os.remove(file)
except OSError:
pass
# administrative
def __remove_files(self):
# environment files
self.__remove_env()
# log files
files = glob.glob(self.log_dir + 'log.*')
for file in files:
os.remove(file)
# database file
os.remove(self.dir + 'porcupine.db')
# index file
os.remove(self.dir + 'porcupine.idx')
def truncate(self):
# older versions of bsddb do not support truncate
if hasattr(self._itemdb, 'truncate'):
self._itemdb.truncate()
self._docdb.truncate()
else:
# close database
self.close()
# remove old database files
self.__remove_files()
# open db
self.__init__()
def backup(self, output_file):
# force checkpoint
self._env.txn_checkpoint(0, 0, db.DB_FORCE)
logs = self._env.log_archive(db.DB_ARCH_LOG)
backfiles = [self.dir + 'porcupine.db', self.dir + 'porcupine.idx'] + \
[self.log_dir + log.decode() for log in logs]
# compact backup....
backup = BackupFile(output_file)
backup.add_files(backfiles)
def restore(self, bset):
self.__remove_files()
backup = BackupFile(bset)
backup.extract(self.dir, self.log_dir)
def shrink(self):
logs = self._env.log_archive()
for log in logs:
os.remove(self.log_dir + log)
return len(logs)
def __maintain(self):
"deadlock detection thread"
while self._running:
time.sleep(0.01)
# deadlock detection
try:
aborted = self._env.lock_detect(db.DB_LOCK_YOUNGEST)
if aborted:
logger.critical(
"Deadlock: Aborted %d deadlocked transaction(s)"
% aborted)
except db.DBError:
pass
def __trickle(self):
"memp_trickle thread"
while self._running:
self._env.memp_trickle(95)
time.sleep(120)
def __checkpoint(self):
"checkpoint thread"
while self._running:
if self.replication_service is None \
or self.replication_service.is_master():
# checkpoint every 512KB written
self._env.txn_checkpoint(512, 0)
time.sleep(16)
#stats = self._env.txn_stat()
#print('txns: %d' % stats['nactive'])
#print('max txns: %d' % stats['maxnactive'])
#print()
#stats = self._env.lock_stat()
#print('Lockers: %d' % stats['nlockers'])
#print('Max Lockers: %d' % stats['maxnlockers'])
#print('Lockers wait: %d' % stats['lockers_wait'])
#print()
#print('Locks: %d' % stats['nlocks'])
#print('Max Locks: %d' % stats['maxnlocks'])
#print('Locks wait: %d' % stats['lock_wait'])
#print('Locks no-wait: %d' % stats['lock_nowait'])
#print()
#print('Lock objects: %d' % stats['nobjects'])
#print('Max objects: %d' % stats['maxnobjects'])
#print('Objects wait: %d' % stats['objs_wait'])
#print()
#print('Requested: %d' % stats['nrequests'])
#print('Released: %d' % stats['nreleases'])
#print('-' * 80)
def close(self, **kwargs):
if self._running:
self._running = False
# join threads
if self._maintenance_thread is not None:
self._maintenance_thread.join()
if self._checkpoint_thread is not None:
self._checkpoint_thread.join()
if self._trickle_thread is not None:
self._trickle_thread.join()
self._itemdb.close()
self._docdb.close()
# close indexes
[index.close() for index in self._indices.values()]
self._env.close()
# clean-up environment files
if (self._maintenance_thread is not None
or kwargs.get('clear_env', False)):
self.__remove_env()
class DB(object):
"Berkeley DB database interface"
# data dir
dir = os.path.abspath(settings['store']['bdb_data_dir'])
if dir[-1] != '/':
dir += '/'
# log_dir
log_dir = os.path.abspath(settings['store'].get('bdb_log_dir', dir))
if log_dir[-1] != '/':
log_dir += '/'
# environment files directory
env_dir = os.path.abspath(settings['store'].get(
'env_dir', os.path.abspath(settings['global']['temp_folder'])))
if env_dir[-1] != '/':
env_dir += '/'
# cache size
cache_size = settings['store'].get('cache_size', None)
# maximum concurrent transactions
# due to snapshot isolation this should be kept high enough
txn_max = settings['store'].get('max_tx', 1000)
# transaction timeout
txn_timeout = settings['store'].get('tx_timeout', None)
# shared memory key
shm_key = settings['store'].get('shm_key', None)
# maintenance (deadlock detector) thread
_maintenance_thread = None
# checkpoint thread
_checkpoint_thread = None
# trickle thread
_trickle_thread = None
# log berkeleyDB version
logger.info('BerkeleyDB version is %s' %
'.'.join(str(x) for x in db.version()))
def __init__(self, **kwargs):
# create db environment
additional_flags = kwargs.get('flags', 0)
recovery_mode = kwargs.get('recover', 0)
if recovery_mode == 2:
additional_flags |= db.DB_RECOVER_FATAL
elif recovery_mode == 1:
additional_flags |= db.DB_RECOVER
if hasattr(db, 'DB_REGISTER'):
additional_flags |= db.DB_REGISTER
self._env = db.DBEnv()
# ability to override settings' dir for testing purposes
data_dir = kwargs.get('dir', self.dir)
self._env.set_data_dir(data_dir)
# ability to override settings' log_dir for testing purposes
log_dir = kwargs.get('log_dir', kwargs.get('dir', self.log_dir))
self._env.set_lg_dir(log_dir)
self._env.set_tx_max(self.txn_max)
if self.txn_timeout is not None:
if self.txn_timeout > 0:
self._env.set_timeout(self.txn_timeout, db.DB_SET_TXN_TIMEOUT)
else:
self._env.set_flags(db.DB_TXN_NOWAIT, 1)
if self.cache_size is not None:
self._env.set_cachesize(*self.cache_size)
if os.name != 'nt' and self.shm_key:
self._env.set_shm_key(self.shm_key)
additional_flags |= db.DB_SYSTEM_MEM
# replication settings
rep_config = settings['store'].get('rep_config', None)
init_rep = kwargs.get('init_rep', False)
if rep_config:
# in replicated environments use non-durable transactions
self._env.set_flags(db.DB_TXN_NOSYNC, 1)
additional_flags |= db.DB_INIT_REP
self._env.open(
self.env_dir,
db.DB_THREAD | db.DB_INIT_MPOOL | db.DB_INIT_LOCK | db.DB_INIT_LOG
| db.DB_INIT_TXN | db.DB_CREATE | additional_flags)
db_flags = db.DB_THREAD | db.DB_AUTO_COMMIT | db.DB_CREATE | \
db.DB_MULTIVERSION
db_mode = 0o660
if rep_config:
from porcupine.db.bsddb.replication import ReplicationService
# initialiaze replication service
self.replication_service = \
ReplicationService(self._env, rep_config)
if init_rep:
# check multiprocessing
is_multiprocess = services['main'].is_multiprocess or \
services['management'].is_multiprocess
if is_multiprocess and int(rep_config['priority']) > 0 \
and db.version() < (4, 8):
self._env.close()
self.__remove_env()
raise exceptions.ConfigurationError(
'Multiprocessing master candidates ' +
'require BerkeleyDB 4.8 or higher')
# start replication service
self.replication_service.start()
# wait for client start-up
timeout = time.time() + 20
while time.time() < timeout and \
not self.replication_service.is_master() and \
not self.replication_service.client_startup_done:
time.sleep(0.02)
timeout = time.time() + 20
while time.time() < timeout and \
not (os.path.exists(
os.path.join(self.dir, 'porcupine.db'))):
time.sleep(0.02)
else:
self.replication_service = None
# open items db
while True:
self._itemdb = db.DB(self._env)
self._itemdb.set_pagesize(2048)
try:
self._itemdb.open('porcupine.db',
'items',
dbtype=db.DB_BTREE,
mode=db_mode,
flags=db_flags)
except db.DBLockDeadlockError:
self._itemdb.close()
continue
break
# open documents db
while True:
self._docdb = db.DB(self._env)
try:
self._docdb.open('porcupine.db',
'docs',
dbtype=db.DB_HASH,
mode=db_mode,
flags=db_flags)
except db.DBLockDeadlockError:
self._docdb.close()
continue
break
# open indices
self._indices = {}
for name, unique, immutable in settings['store']['indices']:
self._indices[name] = DbIndex(self._env, self._itemdb, name,
unique, immutable, db_flags)
self._running = True
maintain = kwargs.get('maintain', False)
if maintain and self._maintenance_thread is None:
# start deadlock detector
self._maintenance_thread = Thread(target=self.__maintain,
name='DB maintenance thread')
self._maintenance_thread.start()
# start checkpoint thread
self._checkpoint_thread = Thread(target=self.__checkpoint,
name='DB checkpoint thread')
self._checkpoint_thread.start()
#if hasattr(self._env, 'memp_trickle'):
# # strart memp_trickle thread
# self._trickle_thread = Thread(target=self.__trickle,
# name='DB memp_trickle thread')
# self._trickle_thread.start()
def is_open(self):
return self._running
# item operations
def get_item(self, oid):
if type(oid) != bytes:
oid = oid.encode('utf-8')
try:
return self._indices['_id'].db.get(oid,
txn=context._trans
and context._trans.txn)
except UnicodeEncodeError:
return None
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
if context._trans is not None:
context._trans.abort()
raise exceptions.DBRetryTransaction
def put_item(self, item):
try:
self._itemdb.put(
pack_value(item._pid) + b'_' + pack_value(item._id),
persist.dumps(item), context._trans.txn)
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
context._trans.abort()
raise exceptions.DBRetryTransaction
except db.DBError as e:
if e.args[0] == _err_unsupported_index_type:
raise db.DBError('Unsupported indexed data type')
else:
raise
def delete_item(self, item):
try:
self._itemdb.delete(
pack_value(item._pid) + b'_' + pack_value(item._id),
context._trans.txn)
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
context._trans.abort()
raise exceptions.DBRetryTransaction
# containers
def get_children(self, container_id):
cursor = Cursor(self._itemdb, '_pid')
cursor.set_scope(container_id)
cursor.set_range(None, None)
return cursor
def get_child_by_name(self, container_id, name):
try:
return self._indices['displayName'].db.get(
pack_value(container_id) + b'_' + pack_value(name),
txn=context._trans and context._trans.txn)
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
if context._trans is not None:
context._trans.abort()
raise exceptions.DBRetryTransaction
# external attributes
def get_external(self, id):
try:
return self._docdb.get(id.encode('ascii'),
txn=context._trans and context._trans.txn)
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
if context._trans is not None:
context._trans.abort()
raise exceptions.DBRetryTransaction
def put_external(self, id, stream):
try:
self._docdb.put(id.encode('ascii'), stream, context._trans.txn)
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
context._trans.abort()
raise exceptions.DBRetryTransaction
def delete_external(self, id):
try:
self._docdb.delete(id.encode('ascii'), context._trans.txn)
except db.DBNotFoundError:
# virtual external due to elastic schema
pass
except (db.DBLockDeadlockError, db.DBLockNotGrantedError):
context._trans.abort()
raise exceptions.DBRetryTransaction
# indices
def get_cursor_list(self, conditions):
cur_list = []
for index, value in conditions:
cursor = Cursor(self._indices[index].db, self._indices[index].name)
if isinstance(value, (list, tuple)):
reversed = (len(value) == 3 and value[2])
cursor.set_range(value[0], value[1])
if reversed:
cursor.reverse()
else:
cursor.set(value)
cur_list.append(cursor)
return cur_list
def query(self, conditions):
cur_list = self.get_cursor_list(conditions)
if len(cur_list) == 1:
return cur_list[0]
else:
c_join = Join(self._itemdb, cur_list)
return c_join
def test_conditions(self, scope, conditions):
cur_list = self.get_cursor_list(conditions)
if len(cur_list) == 1:
cursor = cur_list[0]
else:
cursor = Join(self._itemdb, cur_list)
cursor.set_scope(scope)
iterator = iter(cursor)
try:
result = bool(next(iterator))
except StopIteration:
result = False
cursor.close()
return result
# transactions
def get_transaction(self, **kwargs):
nosync = kwargs.get('nosync', False)
snapshot = kwargs.get('snapshot', False)
return Transaction(self._env, nosync, snapshot)
def __remove_env(self):
files = glob.glob(self.env_dir + '__db.*')
for file in files:
try:
os.remove(file)
except OSError:
pass
# administrative
def __remove_files(self):
# environment files
self.__remove_env()
# log files
files = glob.glob(self.log_dir + 'log.*')
for file in files:
os.remove(file)
# database file
os.remove(self.dir + 'porcupine.db')
# index file
os.remove(self.dir + 'porcupine.idx')
def truncate(self):
# older versions of bsddb do not support truncate
if hasattr(self._itemdb, 'truncate'):
self._itemdb.truncate()
self._docdb.truncate()
else:
# close database
self.close()
# remove old database files
self.__remove_files()
# open db
self.__init__()
def backup(self, output_file):
# force checkpoint
self._env.txn_checkpoint(0, 0, db.DB_FORCE)
logs = self._env.log_archive(db.DB_ARCH_LOG)
backfiles = [self.dir + 'porcupine.db', self.dir + 'porcupine.idx'] + \
[self.log_dir + log.decode() for log in logs]
# compact backup....
backup = BackupFile(output_file)
backup.add_files(backfiles)
def restore(self, bset):
self.__remove_files()
backup = BackupFile(bset)
backup.extract(self.dir, self.log_dir)
def shrink(self):
logs = self._env.log_archive()
for log in logs:
os.remove(self.log_dir + log)
return len(logs)
def __maintain(self):
"deadlock detection thread"
while self._running:
time.sleep(0.01)
# deadlock detection
try:
aborted = self._env.lock_detect(db.DB_LOCK_YOUNGEST)
if aborted:
logger.critical(
"Deadlock: Aborted %d deadlocked transaction(s)" %
aborted)
except db.DBError:
pass
def __trickle(self):
"memp_trickle thread"
while self._running:
self._env.memp_trickle(95)
time.sleep(120)
def __checkpoint(self):
"checkpoint thread"
while self._running:
if self.replication_service is None \
or self.replication_service.is_master():
# checkpoint every 512KB written
self._env.txn_checkpoint(512, 0)
time.sleep(16)
#stats = self._env.txn_stat()
#print('txns: %d' % stats['nactive'])
#print('max txns: %d' % stats['maxnactive'])
#print()
#stats = self._env.lock_stat()
#print('Lockers: %d' % stats['nlockers'])
#print('Max Lockers: %d' % stats['maxnlockers'])
#print('Lockers wait: %d' % stats['lockers_wait'])
#print()
#print('Locks: %d' % stats['nlocks'])
#print('Max Locks: %d' % stats['maxnlocks'])
#print('Locks wait: %d' % stats['lock_wait'])
#print('Locks no-wait: %d' % stats['lock_nowait'])
#print()
#print('Lock objects: %d' % stats['nobjects'])
#print('Max objects: %d' % stats['maxnobjects'])
#print('Objects wait: %d' % stats['objs_wait'])
#print()
#print('Requested: %d' % stats['nrequests'])
#print('Released: %d' % stats['nreleases'])
#print('-' * 80)
def close(self, **kwargs):
if self._running:
self._running = False
# join threads
if self._maintenance_thread is not None:
self._maintenance_thread.join()
if self._checkpoint_thread is not None:
self._checkpoint_thread.join()
if self._trickle_thread is not None:
self._trickle_thread.join()
self._itemdb.close()
self._docdb.close()
# close indexes
[index.close() for index in self._indices.values()]
self._env.close()
# clean-up environment files
if (self._maintenance_thread is not None
or kwargs.get('clear_env', False)):
self.__remove_env()