def test_validate_sync_to(self):
for goodurl in ('http://1.1.1.1/v1/a/c/o', 'http://1.1.1.1:8080/a/c/o',
'http://2.2.2.2/a/c/o', 'https://1.1.1.1/v1/a/c/o'):
self.assertEquals(
utils.validate_sync_to(goodurl, ['1.1.1.1', '2.2.2.2']), None)
for badurl in ('http://1.1.1.1', 'httpq://1.1.1.1/v1/a/c/o',
'http://1.1.1.1/v1/a/c/o?query',
'http://1.1.1.1/v1/a/c/o#frag',
'http://1.1.1.1/v1/a/c/o?query#frag',
'http://1.1.1.1/v1/a/c/o?query=param',
'http://1.1.1.1/v1/a/c/o?query=param#frag',
'http://1.1.1.2/v1/a/c/o'):
self.assertNotEquals(
utils.validate_sync_to(badurl, ['1.1.1.1', '2.2.2.2']), None)
def POST(self, req):
"""Handle HTTP POST request."""
drive, part, account, container = split_and_validate_path(req, 4)
if 'x-timestamp' not in req.headers or \
not check_float(req.headers['x-timestamp']):
return HTTPBadRequest(body='Missing or bad timestamp',
request=req,
content_type='text/plain')
if 'x-container-sync-to' in req.headers:
err = validate_sync_to(req.headers['x-container-sync-to'],
self.allowed_sync_hosts)
if err:
return HTTPBadRequest(err)
if self.mount_check and not check_mount(self.root, drive):
return HTTPInsufficientStorage(drive=drive, request=req)
broker = self._get_container_broker(drive, part, account, container)
if broker.is_deleted():
return HTTPNotFound(request=req)
timestamp = normalize_timestamp(req.headers['x-timestamp'])
metadata = {}
metadata.update((key, (value, timestamp))
for key, value in req.headers.iteritems()
if key.lower() in self.save_headers
or is_sys_or_user_meta('container', key))
if metadata:
if 'X-Container-Sync-To' in metadata:
if 'X-Container-Sync-To' not in broker.metadata or \
metadata['X-Container-Sync-To'][0] != \
broker.metadata['X-Container-Sync-To'][0]:
broker.set_x_container_sync_points(-1, -1)
broker.update_metadata(metadata)
return HTTPNoContent(request=req)
def POST(self, req):
"""Handle HTTP POST request."""
drive, part, account, container = split_and_validate_path(req, 4)
req_timestamp = valid_timestamp(req)
if 'x-container-sync-to' in req.headers:
err, sync_to, realm, realm_key = validate_sync_to(
req.headers['x-container-sync-to'], self.allowed_sync_hosts,
self.realms_conf)
if err:
return HTTPBadRequest(err)
if self.mount_check and not check_mount(self.root, drive):
return HTTPInsufficientStorage(drive=drive, request=req)
broker = self._get_container_broker(drive, part, account, container)
if broker.is_deleted():
return HTTPNotFound(request=req)
metadata = {}
metadata.update(
(key, (value, req_timestamp.internal))
for key, value in req.headers.iteritems()
if key.lower() in self.save_headers or
is_sys_or_user_meta('container', key))
if metadata:
if 'X-Container-Sync-To' in metadata:
if 'X-Container-Sync-To' not in broker.metadata or \
metadata['X-Container-Sync-To'][0] != \
broker.metadata['X-Container-Sync-To'][0]:
broker.set_x_container_sync_points(-1, -1)
broker.update_metadata(metadata, validate_metadata=True)
return HTTPNoContent(request=req)
def POST(self, req):
"""Handle HTTP POST request."""
drive, part, account, container = split_and_validate_path(req, 4)
req_timestamp = valid_timestamp(req)
if 'x-container-sync-to' in req.headers:
err, sync_to, realm, realm_key = validate_sync_to(
req.headers['x-container-sync-to'], self.allowed_sync_hosts,
self.realms_conf)
if err:
return HTTPBadRequest(err)
if self.mount_check and not check_mount(self.root, drive):
return HTTPInsufficientStorage(drive=drive, request=req)
broker = self._get_container_broker(drive, part, account, container)
if broker.is_deleted():
return HTTPNotFound(request=req)
broker.update_put_timestamp(req_timestamp.internal)
metadata = {}
metadata.update((key, (value, req_timestamp.internal))
for key, value in req.headers.items()
if key.lower() in self.save_headers
or is_sys_or_user_meta('container', key))
if metadata:
if 'X-Container-Sync-To' in metadata:
if 'X-Container-Sync-To' not in broker.metadata or \
metadata['X-Container-Sync-To'][0] != \
broker.metadata['X-Container-Sync-To'][0]:
broker.set_x_container_sync_points(-1, -1)
broker.update_metadata(metadata, validate_metadata=True)
self._update_sync_store(broker, 'POST')
return HTTPNoContent(request=req)
def POST(self, req):
"""Handle HTTP POST request."""
drive, part, account, container = split_and_validate_path(req, 4)
if 'x-timestamp' not in req.headers or \
not check_float(req.headers['x-timestamp']):
return HTTPBadRequest(body='Missing or bad timestamp',
request=req, content_type='text/plain')
if 'x-container-sync-to' in req.headers:
err = validate_sync_to(req.headers['x-container-sync-to'],
self.allowed_sync_hosts)
if err:
return HTTPBadRequest(err)
if self.mount_check and not check_mount(self.root, drive):
return HTTPInsufficientStorage(drive=drive, request=req)
broker = self._get_container_broker(drive, part, account, container)
if broker.is_deleted():
return HTTPNotFound(request=req)
timestamp = normalize_timestamp(req.headers['x-timestamp'])
metadata = {}
metadata.update(
(key, (value, timestamp)) for key, value in req.headers.iteritems()
if key.lower() in self.save_headers or
key.lower().startswith('x-container-meta-'))
if metadata:
if 'X-Container-Sync-To' in metadata:
if 'X-Container-Sync-To' not in broker.metadata or \
metadata['X-Container-Sync-To'][0] != \
broker.metadata['X-Container-Sync-To'][0]:
broker.set_x_container_sync_points(-1, -1)
broker.update_metadata(metadata)
return HTTPNoContent(request=req)
def test_validate_sync_to(self):
for goodurl in ('http://1.1.1.1/v1/a/c/o',
'http://1.1.1.1:8080/a/c/o',
'http://2.2.2.2/a/c/o',
'https://1.1.1.1/v1/a/c/o'):
self.assertEquals(utils.validate_sync_to(goodurl,
['1.1.1.1', '2.2.2.2']),
None)
for badurl in ('http://1.1.1.1',
'httpq://1.1.1.1/v1/a/c/o',
'http://1.1.1.1/v1/a/c/o?query',
'http://1.1.1.1/v1/a/c/o#frag',
'http://1.1.1.1/v1/a/c/o?query#frag',
'http://1.1.1.1/v1/a/c/o?query=param',
'http://1.1.1.1/v1/a/c/o?query=param#frag',
'http://1.1.1.2/v1/a/c/o'):
self.assertNotEquals(utils.validate_sync_to(badurl,
['1.1.1.1', '2.2.2.2']),
None)
def PUT(self, req):
"""Handle HTTP PUT request."""
drive, part, account, container, obj = split_and_validate_path(
req, 4, 5, True)
if 'x-timestamp' not in req.headers or \
not check_float(req.headers['x-timestamp']):
return HTTPBadRequest(body='Missing timestamp', request=req,
content_type='text/plain')
if 'x-container-sync-to' in req.headers:
err, sync_to, realm, realm_key = validate_sync_to(
req.headers['x-container-sync-to'], self.allowed_sync_hosts,
self.realms_conf)
if err:
return HTTPBadRequest(err)
if self.mount_check and not check_mount(self.root, drive):
return HTTPInsufficientStorage(drive=drive, request=req)
timestamp = normalize_timestamp(req.headers['x-timestamp'])
broker = self._get_container_broker(drive, part, account, container)
if obj: # put container object
if account.startswith(self.auto_create_account_prefix) and \
not os.path.exists(broker.db_file):
try:
broker.initialize(timestamp)
except DatabaseAlreadyExists:
pass
if not os.path.exists(broker.db_file):
return HTTPNotFound()
broker.put_object(obj, timestamp, int(req.headers['x-size']),
req.headers['x-content-type'],
req.headers['x-etag'])
return HTTPCreated(request=req)
else: # put container
created = self._update_or_create(req, broker, timestamp)
metadata = {}
metadata.update(
(key, (value, timestamp))
for key, value in req.headers.iteritems()
if key.lower() in self.save_headers or
is_sys_or_user_meta('container', key))
if metadata:
if 'X-Container-Sync-To' in metadata:
if 'X-Container-Sync-To' not in broker.metadata or \
metadata['X-Container-Sync-To'][0] != \
broker.metadata['X-Container-Sync-To'][0]:
broker.set_x_container_sync_points(-1, -1)
broker.update_metadata(metadata)
resp = self.account_update(req, account, container, broker)
if resp:
return resp
if created:
return HTTPCreated(request=req)
else:
return HTTPAccepted(request=req)
def POST(self, req):
"""Handle HTTP POST request."""
drive, part, account, container = split_and_validate_path(req, 4)
req_timestamp = valid_timestamp(req)
if 'x-container-sync-to' in req.headers:
err, sync_to, realm, realm_key = validate_sync_to(
req.headers['x-container-sync-to'], self.allowed_sync_hosts,
self.realms_conf)
if err:
return HTTPBadRequest(err)
if not check_drive(self.root, drive, self.mount_check):
return HTTPInsufficientStorage(drive=drive, request=req)
broker = self._get_container_broker(drive, part, account, container)
if broker.is_deleted():
return HTTPNotFound(request=req)
broker.update_put_timestamp(req_timestamp.internal)
self._update_metadata(req, broker, req_timestamp, 'POST')
return HTTPNoContent(request=req)
def POST(self, req):
"""Handle HTTP POST request."""
drive, part, account, container = split_and_validate_path(req, 4)
req_timestamp = valid_timestamp(req)
if 'x-container-sync-to' in req.headers:
err, sync_to, realm, realm_key = validate_sync_to(
req.headers['x-container-sync-to'], self.allowed_sync_hosts,
self.realms_conf)
if err:
return HTTPBadRequest(err)
try:
check_drive(self.root, drive, self.mount_check)
except ValueError:
return HTTPInsufficientStorage(drive=drive, request=req)
if not self.check_free_space(drive):
return HTTPInsufficientStorage(drive=drive, request=req)
broker = self._get_container_broker(drive, part, account, container)
if broker.is_deleted():
return HTTPNotFound(request=req)
broker.update_put_timestamp(req_timestamp.internal)
self._update_metadata(req, broker, req_timestamp, 'POST')
return HTTPNoContent(request=req)
def container_sync(self, path):
"""
Checks the given path for a container database, determines if syncing
is turned on for that database and, if so, sends any updates to the
other container.
:param path: the path to a container db
"""
broker = None
try:
broker = ContainerBroker(path, logger=self.logger)
# The path we pass to the ContainerBroker is a real path of
# a container DB. If we get here, however, it means that this
# path is linked from the sync_containers dir. In rare cases
# of race or processes failures the link can be stale and
# the get_info below will raise a DB doesn't exist exception
# In this case we remove the stale link and raise an error
# since in most cases the db should be there.
try:
info = broker.get_info()
except DatabaseConnectionError as db_err:
if str(db_err).endswith("DB doesn't exist"):
self.sync_store.remove_synced_container(broker)
raise
x, nodes = self.container_ring.get_nodes(info['account'],
info['container'])
for ordinal, node in enumerate(nodes):
if is_local_device(self._myips, self._myport, node['ip'],
node['port']):
break
else:
return
if not broker.is_deleted():
sync_to = None
user_key = None
sync_point1 = info['x_container_sync_point1']
sync_point2 = info['x_container_sync_point2']
for key, (value, timestamp) in broker.metadata.items():
if key.lower() == 'x-container-sync-to':
sync_to = value
elif key.lower() == 'x-container-sync-key':
user_key = value
if not sync_to or not user_key:
self.container_skips += 1
self.logger.increment('skips')
return
err, sync_to, realm, realm_key = validate_sync_to(
sync_to, self.allowed_sync_hosts, self.realms_conf)
if err:
self.logger.info(
_('ERROR %(db_file)s: %(validate_sync_to_err)s'), {
'db_file': str(broker),
'validate_sync_to_err': err
})
self.container_failures += 1
self.logger.increment('failures')
return
start_at = time()
stop_at = start_at + self.container_time
next_sync_point = None
sync_stage_time = start_at
try:
while time() < stop_at and sync_point2 < sync_point1:
rows = broker.get_items_since(sync_point2, 1)
if not rows:
break
row = rows[0]
if row['ROWID'] > sync_point1:
break
# This node will only initially sync out one third
# of the objects (if 3 replicas, 1/4 if 4, etc.)
# and will skip problematic rows as needed in case of
# faults.
# This section will attempt to sync previously skipped
# rows in case the previous attempts by any of the
# nodes didn't succeed.
if not self.container_sync_row(row, sync_to, user_key,
broker, info, realm,
realm_key):
if not next_sync_point:
next_sync_point = sync_point2
sync_point2 = row['ROWID']
broker.set_x_container_sync_points(None, sync_point2)
if next_sync_point:
broker.set_x_container_sync_points(
None, next_sync_point)
else:
next_sync_point = sync_point2
sync_stage_time = time()
while sync_stage_time < stop_at:
rows = broker.get_items_since(sync_point1, 1)
if not rows:
break
row = rows[0]
key = hash_path(info['account'],
info['container'],
row['name'],
raw_digest=True)
# This node will only initially sync out one third of
# the objects (if 3 replicas, 1/4 if 4, etc.).
# It'll come back around to the section above
# and attempt to sync previously skipped rows in case
# the other nodes didn't succeed or in case it failed
# to do so the first time.
if unpack_from('>I', key)[0] % \
len(nodes) == ordinal:
self.container_sync_row(row, sync_to, user_key,
broker, info, realm,
realm_key)
sync_point1 = row['ROWID']
broker.set_x_container_sync_points(sync_point1, None)
sync_stage_time = time()
self.container_syncs += 1
self.logger.increment('syncs')
finally:
self.container_report(start_at, sync_stage_time,
sync_point1, next_sync_point, info,
broker.get_max_row())
except (Exception, Timeout):
self.container_failures += 1
self.logger.increment('failures')
self.logger.exception(_('ERROR Syncing %s'),
broker if broker else path)
def PUT(self, req):
"""Handle HTTP PUT request."""
drive, part, account, container, obj = get_obj_name_and_placement(req)
req_timestamp = valid_timestamp(req)
if 'x-container-sync-to' in req.headers:
err, sync_to, realm, realm_key = validate_sync_to(
req.headers['x-container-sync-to'], self.allowed_sync_hosts,
self.realms_conf)
if err:
return HTTPBadRequest(err)
try:
check_drive(self.root, drive, self.mount_check)
except ValueError:
return HTTPInsufficientStorage(drive=drive, request=req)
if not self.check_free_space(drive):
return HTTPInsufficientStorage(drive=drive, request=req)
requested_policy_index = self.get_and_validate_policy_index(req)
broker = self._get_container_broker(drive, part, account, container)
if obj: # put container object
# obj put expects the policy_index header, default is for
# legacy support during upgrade.
obj_policy_index = requested_policy_index or 0
self._maybe_autocreate(broker, req_timestamp, account,
obj_policy_index)
# redirect if a shard exists for this object name
response = self._redirect_to_shard(req, broker, obj)
if response:
return response
broker.put_object(
obj, req_timestamp.internal, int(req.headers['x-size']),
wsgi_to_str(req.headers['x-content-type']),
wsgi_to_str(req.headers['x-etag']), 0, obj_policy_index,
wsgi_to_str(req.headers.get('x-content-type-timestamp')),
wsgi_to_str(req.headers.get('x-meta-timestamp')))
return HTTPCreated(request=req)
record_type = req.headers.get('x-backend-record-type', '').lower()
if record_type == RECORD_TYPE_SHARD:
try:
# validate incoming data...
shard_ranges = [
ShardRange.from_dict(sr) for sr in json.loads(req.body)
]
except (ValueError, KeyError, TypeError) as err:
return HTTPBadRequest('Invalid body: %r' % err)
created = self._maybe_autocreate(broker, req_timestamp, account,
requested_policy_index)
self._update_metadata(req, broker, req_timestamp, 'PUT')
if shard_ranges:
# TODO: consider writing the shard ranges into the pending
# file, but if so ensure an all-or-none semantic for the write
broker.merge_shard_ranges(shard_ranges)
else: # put container
if requested_policy_index is None:
# use the default index sent by the proxy if available
new_container_policy = req.headers.get(
'X-Backend-Storage-Policy-Default', int(POLICIES.default))
else:
new_container_policy = requested_policy_index
created = self._update_or_create(req, broker,
req_timestamp.internal,
new_container_policy,
requested_policy_index)
self._update_metadata(req, broker, req_timestamp, 'PUT')
resp = self.account_update(req, account, container, broker)
if resp:
return resp
if created:
return HTTPCreated(request=req,
headers={
'x-backend-storage-policy-index':
broker.storage_policy_index
})
else:
return HTTPAccepted(request=req,
headers={
'x-backend-storage-policy-index':
broker.storage_policy_index
})
def PUT(self, req):
"""Handle HTTP PUT request."""
drive, part, account, container, obj = split_and_validate_path(
req, 4, 5, True)
req_timestamp = valid_timestamp(req)
if 'x-container-sync-to' in req.headers:
err, sync_to, realm, realm_key = validate_sync_to(
req.headers['x-container-sync-to'], self.allowed_sync_hosts,
self.realms_conf)
if err:
return HTTPBadRequest(err)
try:
check_drive(self.root, drive, self.mount_check)
except ValueError:
return HTTPInsufficientStorage(drive=drive, request=req)
if not self.check_free_space(drive):
return HTTPInsufficientStorage(drive=drive, request=req)
requested_policy_index = self.get_and_validate_policy_index(req)
broker = self._get_container_broker(drive, part, account, container)
if obj: # put container object
# obj put expects the policy_index header, default is for
# legacy support during upgrade.
obj_policy_index = requested_policy_index or 0
self._maybe_autocreate(broker, req_timestamp, account,
obj_policy_index)
# redirect if a shard exists for this object name
response = self._redirect_to_shard(req, broker, obj)
if response:
return response
broker.put_object(obj, req_timestamp.internal,
int(req.headers['x-size']),
wsgi_to_str(req.headers['x-content-type']),
wsgi_to_str(req.headers['x-etag']), 0,
obj_policy_index,
wsgi_to_str(req.headers.get(
'x-content-type-timestamp')),
wsgi_to_str(req.headers.get('x-meta-timestamp')))
return HTTPCreated(request=req)
record_type = req.headers.get('x-backend-record-type', '').lower()
if record_type == RECORD_TYPE_SHARD:
try:
# validate incoming data...
shard_ranges = [ShardRange.from_dict(sr)
for sr in json.loads(req.body)]
except (ValueError, KeyError, TypeError) as err:
return HTTPBadRequest('Invalid body: %r' % err)
created = self._maybe_autocreate(broker, req_timestamp, account,
requested_policy_index)
self._update_metadata(req, broker, req_timestamp, 'PUT')
if shard_ranges:
# TODO: consider writing the shard ranges into the pending
# file, but if so ensure an all-or-none semantic for the write
broker.merge_shard_ranges(shard_ranges)
else: # put container
if requested_policy_index is None:
# use the default index sent by the proxy if available
new_container_policy = req.headers.get(
'X-Backend-Storage-Policy-Default', int(POLICIES.default))
else:
new_container_policy = requested_policy_index
created = self._update_or_create(req, broker,
req_timestamp.internal,
new_container_policy,
requested_policy_index)
self._update_metadata(req, broker, req_timestamp, 'PUT')
resp = self.account_update(req, account, container, broker)
if resp:
return resp
if created:
return HTTPCreated(request=req,
headers={'x-backend-storage-policy-index':
broker.storage_policy_index})
else:
return HTTPAccepted(request=req,
headers={'x-backend-storage-policy-index':
broker.storage_policy_index})
def PUT(self, req):
"""Handle HTTP PUT request."""
try:
drive, part, account, container, obj = split_path(
unquote(req.path), 4, 5, True)
except ValueError, err:
return HTTPBadRequest(body=str(err),
content_type='text/plain',
request=req)
if 'x-timestamp' not in req.headers or \
not check_float(req.headers['x-timestamp']):
return HTTPBadRequest(body='Missing timestamp',
request=req,
content_type='text/plain')
if 'x-container-sync-to' in req.headers:
err = validate_sync_to(req.headers['x-container-sync-to'],
self.allowed_sync_hosts)
if err:
return HTTPBadRequest(err)
if self.mount_check and not check_mount(self.root, drive):
return Response(status='507 %s is not mounted' % drive)
timestamp = normalize_timestamp(req.headers['x-timestamp'])
broker = self._get_container_broker(drive, part, account, container)
if obj: # put container object
if account.startswith(self.auto_create_account_prefix) and \
not os.path.exists(broker.db_file):
broker.initialize(timestamp)
if not os.path.exists(broker.db_file):
return HTTPNotFound()
broker.put_object(obj, timestamp, int(req.headers['x-size']),
req.headers['x-content-type'],
req.headers['x-etag'])
def PUT(self, req):
"""Handle HTTP PUT request."""
drive, part, account, container, obj = split_and_validate_path(
req, 4, 5, True)
req_timestamp = valid_timestamp(req)
if 'x-container-sync-to' in req.headers:
err, sync_to, realm, realm_key = validate_sync_to(
req.headers['x-container-sync-to'], self.allowed_sync_hosts,
self.realms_conf)
if err:
return HTTPBadRequest(err)
if self.mount_check and not check_mount(self.root, drive):
return HTTPInsufficientStorage(drive=drive, request=req)
requested_policy_index = self.get_and_validate_policy_index(req)
broker = self._get_container_broker(drive, part, account, container)
if obj: # put container object
# obj put expects the policy_index header, default is for
# legacy support during upgrade.
obj_policy_index = requested_policy_index or 0
if account.startswith(self.auto_create_account_prefix) and \
not os.path.exists(broker.db_file):
try:
broker.initialize(req_timestamp.internal, obj_policy_index)
except DatabaseAlreadyExists:
pass
if not os.path.exists(broker.db_file):
return HTTPNotFound()
broker.put_object(obj, req_timestamp.internal,
int(req.headers['x-size']),
req.headers['x-content-type'],
req.headers['x-etag'], 0, obj_policy_index,
req.headers.get('x-content-type-timestamp'),
req.headers.get('x-meta-timestamp'))
return HTTPCreated(request=req)
else: # put container
if requested_policy_index is None:
# use the default index sent by the proxy if available
new_container_policy = req.headers.get(
'X-Backend-Storage-Policy-Default', int(POLICIES.default))
else:
new_container_policy = requested_policy_index
created = self._update_or_create(req, broker,
req_timestamp.internal,
new_container_policy,
requested_policy_index)
metadata = {}
metadata.update((key, (value, req_timestamp.internal))
for key, value in req.headers.items()
if key.lower() in self.save_headers
or is_sys_or_user_meta('container', key))
if 'X-Container-Sync-To' in metadata:
if 'X-Container-Sync-To' not in broker.metadata or \
metadata['X-Container-Sync-To'][0] != \
broker.metadata['X-Container-Sync-To'][0]:
broker.set_x_container_sync_points(-1, -1)
broker.update_metadata(metadata, validate_metadata=True)
if metadata:
self._update_sync_store(broker, 'PUT')
resp = self.account_update(req, account, container, broker)
if resp:
return resp
if created:
return HTTPCreated(request=req,
headers={
'x-backend-storage-policy-index':
broker.storage_policy_index
})
else:
return HTTPAccepted(request=req,
headers={
'x-backend-storage-policy-index':
broker.storage_policy_index
})
def container_sync(self, path):
"""
Checks the given path for a container database, determines if syncing
is turned on for that database and, if so, sends any updates to the
other container.
:param path: the path to a container db
"""
try:
if not path.endswith(".db"):
return
broker = ContainerBroker(path)
info = broker.get_info()
x, nodes = self.container_ring.get_nodes(info["account"], info["container"])
for ordinal, node in enumerate(nodes):
if node["ip"] in self._myips and node["port"] == self._myport:
break
else:
return
if not broker.is_deleted():
sync_to = None
sync_key = None
sync_point1 = info["x_container_sync_point1"]
sync_point2 = info["x_container_sync_point2"]
for key, (value, timestamp) in broker.metadata.iteritems():
if key.lower() == "x-container-sync-to":
sync_to = value
elif key.lower() == "x-container-sync-key":
sync_key = value
if not sync_to or not sync_key:
self.container_skips += 1
return
sync_to = sync_to.rstrip("/")
err = validate_sync_to(sync_to, self.allowed_sync_hosts)
if err:
self.logger.info(
_("ERROR %(db_file)s: %(validate_sync_to_err)s"),
{"db_file": broker.db_file, "validate_sync_to_err": err},
)
self.container_failures += 1
return
stop_at = time() + self.container_time
while time() < stop_at and sync_point2 < sync_point1:
rows = broker.get_items_since(sync_point2, 1)
if not rows:
break
row = rows[0]
if row["ROWID"] >= sync_point1:
break
key = hash_path(info["account"], info["container"], row["name"], raw_digest=True)
# This node will only intially sync out one third of the
# objects (if 3 replicas, 1/4 if 4, etc.). This section
# will attempt to sync previously skipped rows in case the
# other nodes didn't succeed.
if unpack_from(">I", key)[0] % len(nodes) != ordinal:
if not self.container_sync_row(row, sync_to, sync_key, broker, info):
return
sync_point2 = row["ROWID"]
broker.set_x_container_sync_points(None, sync_point2)
while time() < stop_at:
rows = broker.get_items_since(sync_point1, 1)
if not rows:
break
row = rows[0]
key = hash_path(info["account"], info["container"], row["name"], raw_digest=True)
# This node will only intially sync out one third of the
# objects (if 3 replicas, 1/4 if 4, etc.). It'll come back
# around to the section above and attempt to sync
# previously skipped rows in case the other nodes didn't
# succeed.
if unpack_from(">I", key)[0] % len(nodes) == ordinal:
if not self.container_sync_row(row, sync_to, sync_key, broker, info):
return
sync_point1 = row["ROWID"]
broker.set_x_container_sync_points(sync_point1, None)
self.container_syncs += 1
except (Exception, Timeout), err:
self.container_failures += 1
self.logger.exception(_("ERROR Syncing %s"), (broker.db_file))
def container_sync(self, path):
"""
Checks the given path for a container database, determines if syncing
is turned on for that database and, if so, sends any updates to the
other container.
:param path: the path to a container db
"""
broker = None
try:
broker = ContainerBroker(path)
# The path we pass to the ContainerBroker is a real path of
# a container DB. If we get here, however, it means that this
# path is linked from the sync_containers dir. In rare cases
# of race or processes failures the link can be stale and
# the get_info below will raise a DB doesn't exist exception
# In this case we remove the stale link and raise an error
# since in most cases the db should be there.
try:
info = broker.get_info()
except DatabaseConnectionError as db_err:
if str(db_err).endswith("DB doesn't exist"):
self.sync_store.remove_synced_container(broker)
raise
x, nodes = self.container_ring.get_nodes(info['account'],
info['container'])
for ordinal, node in enumerate(nodes):
if is_local_device(self._myips, self._myport,
node['ip'], node['port']):
break
else:
return
if not broker.is_deleted():
sync_to = None
user_key = None
sync_point1 = info['x_container_sync_point1']
sync_point2 = info['x_container_sync_point2']
for key, (value, timestamp) in broker.metadata.items():
if key.lower() == 'x-container-sync-to':
sync_to = value
elif key.lower() == 'x-container-sync-key':
user_key = value
if not sync_to or not user_key:
self.container_skips += 1
self.logger.increment('skips')
return
err, sync_to, realm, realm_key = validate_sync_to(
sync_to, self.allowed_sync_hosts, self.realms_conf)
if err:
self.logger.info(
_('ERROR %(db_file)s: %(validate_sync_to_err)s'),
{'db_file': str(broker),
'validate_sync_to_err': err})
self.container_failures += 1
self.logger.increment('failures')
return
stop_at = time() + self.container_time
next_sync_point = None
while time() < stop_at and sync_point2 < sync_point1:
rows = broker.get_items_since(sync_point2, 1)
if not rows:
break
row = rows[0]
if row['ROWID'] > sync_point1:
break
key = hash_path(info['account'], info['container'],
row['name'], raw_digest=True)
# This node will only initially sync out one third of the
# objects (if 3 replicas, 1/4 if 4, etc.) and will skip
# problematic rows as needed in case of faults.
# This section will attempt to sync previously skipped
# rows in case the previous attempts by any of the nodes
# didn't succeed.
if not self.container_sync_row(
row, sync_to, user_key, broker, info, realm,
realm_key):
if not next_sync_point:
next_sync_point = sync_point2
sync_point2 = row['ROWID']
broker.set_x_container_sync_points(None, sync_point2)
if next_sync_point:
broker.set_x_container_sync_points(None, next_sync_point)
while time() < stop_at:
rows = broker.get_items_since(sync_point1, 1)
if not rows:
break
row = rows[0]
key = hash_path(info['account'], info['container'],
row['name'], raw_digest=True)
# This node will only initially sync out one third of the
# objects (if 3 replicas, 1/4 if 4, etc.). It'll come back
# around to the section above and attempt to sync
# previously skipped rows in case the other nodes didn't
# succeed or in case it failed to do so the first time.
if unpack_from('>I', key)[0] % \
len(nodes) == ordinal:
self.container_sync_row(
row, sync_to, user_key, broker, info, realm,
realm_key)
sync_point1 = row['ROWID']
broker.set_x_container_sync_points(sync_point1, None)
self.container_syncs += 1
self.logger.increment('syncs')
except (Exception, Timeout):
self.container_failures += 1
self.logger.increment('failures')
self.logger.exception(_('ERROR Syncing %s'),
broker if broker else path)
def PUT(self, req):
"""Handle HTTP PUT request."""
drive, part, account, container, obj = split_and_validate_path(
req, 4, 5, True)
req_timestamp = valid_timestamp(req)
if 'x-container-sync-to' in req.headers:
err, sync_to, realm, realm_key = validate_sync_to(
req.headers['x-container-sync-to'], self.allowed_sync_hosts,
self.realms_conf)
if err:
return HTTPBadRequest(err)
if self.mount_check and not check_mount(self.root, drive):
return HTTPInsufficientStorage(drive=drive, request=req)
requested_policy_index = self.get_and_validate_policy_index(req)
broker = self._get_container_broker(drive, part, account, container)
if obj: # put container object
# obj put expects the policy_index header, default is for
# legacy support during upgrade.
obj_policy_index = requested_policy_index or 0
if account.startswith(self.auto_create_account_prefix) and \
not os.path.exists(broker.db_file):
try:
broker.initialize(req_timestamp.internal, obj_policy_index)
except DatabaseAlreadyExists:
pass
if not os.path.exists(broker.db_file):
return HTTPNotFound()
broker.put_object(obj, req_timestamp.internal,
int(req.headers['x-size']),
req.headers['x-content-type'],
req.headers['x-etag'], 0,
obj_policy_index)
return HTTPCreated(request=req)
else: # put container
if requested_policy_index is None:
# use the default index sent by the proxy if available
new_container_policy = req.headers.get(
'X-Backend-Storage-Policy-Default', int(POLICIES.default))
else:
new_container_policy = requested_policy_index
created = self._update_or_create(req, broker,
req_timestamp.internal,
new_container_policy,
requested_policy_index)
metadata = {}
metadata.update(
(key, (value, req_timestamp.internal))
for key, value in req.headers.iteritems()
if key.lower() in self.save_headers or
is_sys_or_user_meta('container', key))
if 'X-Container-Sync-To' in metadata:
if 'X-Container-Sync-To' not in broker.metadata or \
metadata['X-Container-Sync-To'][0] != \
broker.metadata['X-Container-Sync-To'][0]:
broker.set_x_container_sync_points(-1, -1)
broker.update_metadata(metadata, validate_metadata=True)
resp = self.account_update(req, account, container, broker)
if resp:
return resp
if created:
return HTTPCreated(request=req)
else:
return HTTPAccepted(request=req)
def container_sync(self, path):
"""
Checks the given path for a container database, determines if syncing
is turned on for that database and, if so, sends any updates to the
other container.
:param path: the path to a container db
"""
broker = None
try:
broker = ContainerBroker(path)
info = broker.get_info()
x, nodes = self.container_ring.get_nodes(info['account'],
info['container'])
for ordinal, node in enumerate(nodes):
if node['ip'] in self._myips and node['port'] == self._myport:
break
else:
return
if not broker.is_deleted():
sync_to = None
sync_key = None
sync_point1 = info['x_container_sync_point1']
sync_point2 = info['x_container_sync_point2']
for key, (value, timestamp) in broker.metadata.iteritems():
if key.lower() == 'x-container-sync-to':
sync_to = value
elif key.lower() == 'x-container-sync-key':
sync_key = value
if not sync_to or not sync_key:
self.container_skips += 1
self.logger.increment('skips')
return
sync_to = sync_to.rstrip('/')
err = validate_sync_to(sync_to, self.allowed_sync_hosts)
if err:
self.logger.info(
_('ERROR %(db_file)s: %(validate_sync_to_err)s'),
{'db_file': broker.db_file,
'validate_sync_to_err': err})
self.container_failures += 1
self.logger.increment('failures')
return
stop_at = time() + self.container_time
next_sync_point = None
while time() < stop_at and sync_point2 < sync_point1:
rows = broker.get_items_since(sync_point2, 1)
if not rows:
break
row = rows[0]
if row['ROWID'] > sync_point1:
break
key = hash_path(info['account'], info['container'],
row['name'], raw_digest=True)
# This node will only initially sync out one third of the
# objects (if 3 replicas, 1/4 if 4, etc.) and will skip
# problematic rows as needed in case of faults.
# This section will attempt to sync previously skipped
# rows in case the previous attempts by any of the nodes
# didn't succeed.
if not self.container_sync_row(row, sync_to, sync_key,
broker, info):
if not next_sync_point:
next_sync_point = sync_point2
sync_point2 = row['ROWID']
broker.set_x_container_sync_points(None, sync_point2)
if next_sync_point:
broker.set_x_container_sync_points(None, next_sync_point)
while time() < stop_at:
rows = broker.get_items_since(sync_point1, 1)
if not rows:
break
row = rows[0]
key = hash_path(info['account'], info['container'],
row['name'], raw_digest=True)
# This node will only initially sync out one third of the
# objects (if 3 replicas, 1/4 if 4, etc.). It'll come back
# around to the section above and attempt to sync
# previously skipped rows in case the other nodes didn't
# succeed or in case it failed to do so the first time.
if unpack_from('>I', key)[0] % \
len(nodes) == ordinal:
self.container_sync_row(row, sync_to, sync_key,
broker, info)
sync_point1 = row['ROWID']
broker.set_x_container_sync_points(sync_point1, None)
self.container_syncs += 1
self.logger.increment('syncs')
except (Exception, Timeout) as err:
self.container_failures += 1
self.logger.increment('failures')
self.logger.exception(_('ERROR Syncing %s'),
broker.db_file if broker else path)
"""Handle HTTP PUT request."""
start_time = time.time()
try:
drive, part, account, container, obj = split_path(
unquote(req.path), 4, 5, True)
validate_device_partition(drive, part)
except ValueError, err:
return jresponse('-1', 'bad request', req,400)
if 'x-timestamp' not in req.headers or \
not check_float(req.headers['x-timestamp']):
return jresponse('-1', 'bad request', req,400)
if 'x-container-sync-to' in req.headers:
err = validate_sync_to(req.headers['x-container-sync-to'],
self.allowed_sync_hosts)
if err:
return jresponse('-1', 'bad request', req,400)
if self.mount_check and not check_mount(self.root, drive):
return jresponse('-1', 'insufficient storage', req,507)
timestamp = normalize_timestamp(req.headers['x-timestamp'])
broker = self._get_container_broker(drive, part, account, container)
if obj: # put container object
# print '00000000000000000000000000' + ' '+ req.path
return jresponse('0', '', req,201)
else: # put container
if True:
created = broker.is_deleted()
# print '00000000000000000000000004' +' '+ req.path
def PUT(self, req):
"""Handle HTTP PUT request."""
#如果包含object信息，根据相关信息创建一个object
#如果不包object，则更新container元数据信息，并调用account_update通知account-server更新状态
drive, part, account, container, obj = split_and_validate_path(
req, 4, 5, True)
req_timestamp = valid_timestamp(req)
if 'x-container-sync-to' in req.headers:
err, sync_to, realm, realm_key = validate_sync_to(
req.headers['x-container-sync-to'], self.allowed_sync_hosts,
self.realms_conf)
if err:
return HTTPBadRequest(err)
if self.mount_check and not check_mount(self.root, drive):
return HTTPInsufficientStorage(drive=drive, request=req)
requested_policy_index = self.get_and_validate_policy_index(req)
broker = self._get_container_broker(drive, part, account, container)
#如果存在对象，仅更新container数据库中对象的元数据
if obj: # put container object
# obj put expects the policy_index header, default is for
# legacy support during upgrade.
obj_policy_index = requested_policy_index or 0
if account.startswith(self.auto_create_account_prefix) and \
not os.path.exists(broker.db_file):
try:
#初始化container的元数据数据库
broker.initialize(req_timestamp.internal, obj_policy_index)
except DatabaseAlreadyExists:
pass
if not os.path.exists(broker.db_file):
return HTTPNotFound()
#将对象信息存放在container的元数据数据库，对象文件何时存放？？？
broker.put_object(obj, req_timestamp.internal,
int(req.headers['x-size']),
req.headers['x-content-type'],
req.headers['x-etag'], 0,
obj_policy_index)
return HTTPCreated(request=req)
#如果不包含object信息，则根据request。head中的key-value更新container_statu数据库的元数据信息
# 并调用account_update通知account-server更新状态
# 如果不包含object信息，实现更新container的数据库中的元数据信息
#如果不包含对象，则
else: # put container
if requested_policy_index is None:
# use the default index sent by the proxy if available
new_container_policy = req.headers.get(
'X-Backend-Storage-Policy-Default', int(POLICIES.default))
else:
new_container_policy = requested_policy_index
# 如果数据库不存在，则创建；否则，仅更新数据库的时间戳
created = self._update_or_create(req, broker,
req_timestamp.internal,
new_container_policy,
requested_policy_index)
#根据请求生成元数据信息
metadata = {}
metadata.update(
(key, (value, req_timestamp.internal))
for key, value in req.headers.items()
if key.lower() in self.save_headers or
is_sys_or_user_meta('container', key))
if 'X-Container-Sync-To' in metadata:
if 'X-Container-Sync-To' not in broker.metadata or \
metadata['X-Container-Sync-To'][0] != \
broker.metadata['X-Container-Sync-To'][0]:
broker.set_x_container_sync_points(-1, -1)
#将container的元数据写入数据库中
broker.update_metadata(metadata, validate_metadata=True)
# 根据最新的container信息更新account服务；
# 用于在对container做删除/修改操作时通知其所属account做同步修改；
# 主要部分就是向account所在server_ip发送PUT请求，URL格式为：
# PUThttp://{account_ip}:{account_port}/{account_device}/{account_partition}/{account}/{container}
resp = self.account_update(req, account, container, broker)
if resp:
return resp
if created:
return HTTPCreated(request=req,
headers={'x-backend-storage-policy-index':
broker.storage_policy_index})
else:
return HTTPAccepted(request=req,
headers={'x-backend-storage-policy-index':
broker.storage_policy_index})
