class Checker(object):
def __init__(self,
namespace,
concurrency=50,
error_file=None,
rebuild_file=None,
check_xattr=True,
limit_listings=0,
request_attempts=1,
logger=None,
verbose=False,
check_hash=False,
min_time_in_error=0.0,
required_confirmations=0,
beanstalkd_addr=None,
beanstalkd_tube=BlobRebuilder.DEFAULT_BEANSTALKD_WORKER_TUBE,
cache_size=2**24,
**_kwargs):
self.pool = GreenPool(concurrency)
self.error_file = error_file
self.error_sender = None
self.check_xattr = bool(check_xattr)
self.check_hash = bool(check_hash)
self.logger = logger or get_logger(
{'namespace': namespace}, name='integrity', verbose=verbose)
# Optimisation for when we are only checking one object
# or one container.
# 0 -> do not limit
# 1 -> limit account listings (list of containers)
# 2 -> limit container listings (list of objects)
self.limit_listings = limit_listings
if self.error_file:
outfile = open(self.error_file, 'a')
self.error_writer = csv.writer(outfile, delimiter=' ')
self.rebuild_file = rebuild_file
if self.rebuild_file:
self.fd = open(self.rebuild_file, 'a')
self.rebuild_writer = csv.writer(self.fd, delimiter='|')
if beanstalkd_addr:
self.error_sender = BeanstalkdSender(beanstalkd_addr,
beanstalkd_tube, self.logger)
self.api = ObjectStorageApi(namespace,
logger=self.logger,
max_retries=request_attempts - 1,
request_attempts=request_attempts)
self.rdir_client = RdirClient({"namespace": namespace},
logger=self.logger)
self.accounts_checked = 0
self.containers_checked = 0
self.objects_checked = 0
self.chunks_checked = 0
self.account_not_found = 0
self.container_not_found = 0
self.object_not_found = 0
self.chunk_not_found = 0
self.account_exceptions = 0
self.container_exceptions = 0
self.object_exceptions = 0
self.chunk_exceptions = 0
self.list_cache = CacheDict(cache_size)
self.running_tasks = {}
self.running_lock = Semaphore(1)
self.result_queue = LightQueue(concurrency)
self.running = True
self.run_time = 0
# Set of targets which must be checked again, to confirm
# or deny the issues reported by previous passes.
self.delayed_targets = dict()
# Minimum time in error and number of confirmations of the error
# before triggering a reconstruction action.
self.min_time_in_error = min_time_in_error
self.required_confirmations = required_confirmations
def reset_stats(self):
self.accounts_checked = 0
self.containers_checked = 0
self.objects_checked = 0
self.chunks_checked = 0
self.account_not_found = 0
self.container_not_found = 0
self.object_not_found = 0
self.chunk_not_found = 0
self.account_exceptions = 0
self.container_exceptions = 0
self.object_exceptions = 0
self.chunk_exceptions = 0
def _spawn(self, func, target, *args, **kwargs):
"""
Spawn a task on the internal GreenPool.
Discards the task if the pool is no more running.
"""
if self.running:
return self.pool.spawn(func, target, *args, **kwargs)
self.logger.info("Discarding %s", target)
return None
def _spawn_n(self, func, target, *args, **kwargs):
"""
Spawn a task on the internal GreenPool, do not wait for the result.
Discards the task if the pool is no more running.
"""
if self.running:
return self.pool.spawn_n(func, target, *args, **kwargs)
self.logger.info("Discarding %s", target)
return None
def complete_target_from_chunk_metadata(self, target, xattr_meta):
"""
Complete a Target object from metadata found in chunk's extended
attributes. In case the "fullpath" is not available, try to read
legacy metadata, and maybe ask meta1 to resolve the CID into
account and container names.
"""
# pylint: disable=unbalanced-tuple-unpacking
try:
acct, ct, path, vers, content_id = \
decode_fullpath(xattr_meta['full_path'])
target.account = acct
target.container = ct
target.obj = path
target.content_id = content_id
target.version = vers
except KeyError:
# No fullpath header, try legacy headers
if 'content_path' in xattr_meta:
target.obj = xattr_meta['content_path']
if 'content_id' in xattr_meta:
target.content_id = xattr_meta['content_id']
if 'content_version' in xattr_meta:
target.version = xattr_meta['content_version']
cid = xattr_meta.get('container_id')
if cid:
try:
md = self.api.directory.show(cid=cid)
acct = md.get('account')
ct = md.get('name')
if acct:
target.account = acct
if ct:
target.container = ct
except Exception as err:
self.logger.warn(
"Failed to resolve CID %s into account "
"and container names: %s", cid, err)
def recover_and_complete_object_meta(self, target, chunk):
_, rawx_service, chunk_id = chunk.rsplit('/', 2)
# 1. Fetch chunk list from rdir (could be cached).
# Unfortunately we cannot seek for a chunk ID.
entries = [
x for x in self.rdir_client.chunk_fetch(rawx_service, limit=-1)
if x[2] == chunk_id
]
if not entries:
self.logger.warn('Chunk %s not found in rdir' % chunk_id)
return
elif len(entries) > 1:
self.logger.info('Chunk %s appears in %d objects', chunk_id,
len(entries))
# 2. Find content and container IDs
target.cid, target.content_id = entries[0][0:2]
meta = self.api.object_get_properties(None,
None,
None,
cid=target.cid,
content=target.content_id)
target.obj = meta['name']
target.version = meta['version']
target.account, target.container = self.api.resolve_cid(target.cid)
def send_result(self, target, errors=None, irreparable=False):
"""
Put an item in the result queue.
"""
# TODO(FVE): send to an external queue.
target.append_result(ItemResult(errors, irreparable))
self.result_queue.put(target)
def send_chunk_job(self, target, irreparable=False):
"""
Send a "content broken" event, to trigger the
reconstruction of the chunk.
"""
item = (self.api.namespace, target.cid, target.content_id,
target.chunk)
ev_dict = BlobRebuilder.task_event_from_item(item)
if irreparable:
ev_dict['data']['irreparable'] = irreparable
job = json.dumps(ev_dict)
self.error_sender.send_job(job)
self.error_sender.job_done() # Don't expect any response
def write_error(self, target, irreparable=False):
if not self.error_file:
return
error = list()
if irreparable:
error.append(IRREPARABLE_PREFIX)
error.append(target.account)
if target.container:
error.append(target.container)
if target.obj:
error.append(target.obj)
if target.chunk:
error.append(target.chunk)
self.error_writer.writerow(error)
def write_rebuilder_input(self, target, irreparable=False):
error = list()
if irreparable:
error.append(IRREPARABLE_PREFIX)
error.append(target.cid)
# FIXME(FVE): ensure we always resolve content_id,
# or pass object version along with object name.
error.append(target.content_id or target.obj)
error.append(target.chunk)
self.rebuild_writer.writerow(error)
def write_chunk_error(self, target, chunk=None, irreparable=False):
if chunk is not None:
target = target.copy()
target.chunk = chunk
self.write_error(target, irreparable=irreparable)
if self.rebuild_file:
self.write_rebuilder_input(target, irreparable=irreparable)
if self.error_sender:
self.send_chunk_job(target, irreparable=irreparable)
def _check_chunk_xattr(self, target, obj_meta, xattr_meta):
"""
Check coherency of chunk extended attributes with object metadata.
:returns: a list of errors
"""
errors = list()
# Composed position -> erasure coding
attr_prefix = 'meta' if '.' in obj_meta['pos'] else ''
attr_key = attr_prefix + 'chunk_size'
if str(obj_meta['size']) != xattr_meta.get(attr_key):
errors.append(
"'%s' xattr (%s) differs from size in meta2 (%s)" %
(attr_key, xattr_meta.get(attr_key), obj_meta['size']))
attr_key = attr_prefix + 'chunk_hash'
if obj_meta['hash'] != xattr_meta.get(attr_key):
errors.append(
"'%s' xattr (%s) differs from hash in meta2 (%s)" %
(attr_key, xattr_meta.get(attr_key), obj_meta['hash']))
return errors
def _check_chunk(self, target):
"""
Execute various checks on a chunk:
- does it appear in object's chunk list?
- is it reachable?
- are its extended attributes coherent?
:returns: the list of errors encountered,
and the chunk's owner object metadata.
"""
chunk = target.chunk
errors = list()
obj_meta = None
xattr_meta = None
cached = self._get_cached_or_lock(chunk)
if cached is not None:
return cached + (True, )
self.logger.debug('Checking chunk "%s"', target)
try:
xattr_meta = self.api.blob_client.chunk_head(
chunk, xattr=self.check_xattr, check_hash=self.check_hash)
except exc.NotFound as err:
self.chunk_not_found += 1
errors.append('Not found: %s' % (err, ))
except exc.FaultyChunk as err:
self.chunk_exceptions += 1
errors.append('Faulty: %r' % (err, ))
except Exception as err:
self.chunk_exceptions += 1
errors.append('Check failed: %s' % (err, ))
if not target.obj:
if xattr_meta:
self.complete_target_from_chunk_metadata(target, xattr_meta)
else:
self.recover_and_complete_object_meta(target, chunk)
if target.obj:
obj_listing, obj_meta = self.check_obj(target.copy_object())
if chunk not in obj_listing:
errors.append('Missing from object listing')
db_meta = dict()
else:
db_meta = obj_listing[chunk]
if db_meta and xattr_meta and self.check_xattr:
errors.extend(
self._check_chunk_xattr(target, db_meta, xattr_meta))
self.list_cache[chunk] = errors, obj_meta
self._unlock(chunk)
# Do not send errors directly, let the caller do it.
# Indeed, it may want to check if the chunks can be repaired or not.
self.chunks_checked += 1
return errors, obj_meta, False
def check_chunk(self, target):
errors, _obj_meta, from_cache = self._check_chunk(target)
# If the result comes from the cache, we already reported it.
if not from_cache:
self.send_result(target, errors, target.irreparable)
return errors
def _check_metachunk(self, target, stg_met, pos, chunks, recurse=0):
"""
Check that a metachunk has the right number of chunks.
:returns: the list of errors
"""
required = stg_met.expected_chunks
errors = list()
chunk_results = list()
if len(chunks) < required:
missing_chunks = required - len(chunks)
if stg_met.ec:
subs = {x['num'] for x in chunks}
for sub in range(required):
if sub not in subs:
chkt = target.copy()
chkt.chunk = '%d.%d' % (pos, sub)
err = "Missing chunk at position %s" % chkt.chunk
chunk_results.append((chkt, [err], False))
errors.append(err)
else:
for _ in range(missing_chunks):
chkt = target.copy()
chkt.chunk = '%d.%d' % (pos, sub)
err = "Missing chunk at position %d" % pos
chunk_results.append((chkt, [err], False))
errors.append(err)
if recurse > 0:
for chunk in chunks:
tcopy = target.copy()
tcopy.chunk = chunk['url']
chunk_errors, _, from_cache = self._check_chunk(tcopy)
chunk_results.append((tcopy, chunk_errors, from_cache))
if chunk_errors:
errors.append("Unusable chunk %s at position %s" %
(chunk['url'], chunk['pos']))
irreparable = required - len(errors) < stg_met.min_chunks_to_read
if irreparable:
errors.append(
"Unavailable metachunk at position %s "
"(%d/%d chunks available, %d/%d required)" %
(pos, required - len(errors), stg_met.expected_chunks,
stg_met.min_chunks_to_read, stg_met.expected_chunks))
for tgt, errs, from_cache in chunk_results:
# If the result comes from the cache, we already reported it.
if not from_cache:
self.send_result(tgt, errs, irreparable)
# Since the "metachunk" is not an official item type,
# this method does not report errors itself. Errors will
# be reported as object errors.
return errors
def _check_obj_policy(self, target, obj_meta, chunks, recurse=0):
"""
Check that the list of chunks of an object matches
the object's storage policy.
:returns: the list of errors encountered
"""
stg_met = STORAGE_METHODS.load(obj_meta['chunk_method'])
chunks_by_pos = _sort_chunks(chunks, stg_met.ec)
tasks = list()
for pos, pchunks in iteritems(chunks_by_pos):
tasks.append((pos,
self._spawn(self._check_metachunk,
target.copy(),
stg_met,
pos,
pchunks,
recurse=recurse)))
errors = list()
for pos, task in tasks:
if not task and not self.running:
errors.append("Pos %d skipped: checker is exiting" % pos)
continue
try:
errors.extend(task.wait())
except Exception as err:
errors.append("Check failed: pos %d: %s" % (pos, err))
return errors
def check_obj_versions(self, target, versions, recurse=0):
"""
Run checks of all versions of the targeted object in parallel.
"""
tasks = list()
for ov in versions:
tcopy = target.copy_object()
tcopy.content_id = ov['id']
tcopy.version = str(ov['version'])
tasks.append((tcopy.version,
self._spawn(self.check_obj, tcopy, recurse=recurse)))
errors = list()
for version, task in tasks:
if not task and not self.running:
errors.append("Version %s skipped: checker is exiting" %
version)
continue
try:
task.wait()
except Exception as err:
errors.append("Check failed: version %s: %s" % (version, err))
if errors:
# Send a result with the target without version to tell
# we were not able to check all versions of the object.
self.send_result(target, errors)
def _load_obj_meta(self, target, errors):
"""
Load object metadata and chunks.
:param target: which object to check.
:param errors: list of errors that will be appended
in case any error occurs.
:returns: a tuple with object metadata and a list of chunks.
"""
try:
return self.api.object_locate(target.account,
target.container,
target.obj,
version=target.version,
properties=False)
except exc.NoSuchObject as err:
self.object_not_found += 1
errors.append('Not found: %s' % (err, ))
except Exception as err:
self.object_exceptions += 1
errors.append('Check failed: %s' % (err, ))
return None, []
def _get_cached_or_lock(self, lock_key):
# If something is running, wait for it
with self.running_lock:
event = self.running_tasks.get(lock_key)
if event:
event.wait()
event = None
# Maybe get a cached result
if lock_key in self.list_cache:
return self.list_cache[lock_key]
# No cached result, try to compute the thing ourselves
while True:
with self.running_lock:
# Another check while locked
if lock_key in self.list_cache:
return self.list_cache[lock_key]
# Still nothing cached
event = self.running_tasks.get(lock_key)
if event is None:
self.running_tasks[lock_key] = Event()
return None
event.wait()
def _unlock(self, lock_key):
with self.running_lock:
event = self.running_tasks[lock_key]
del self.running_tasks[lock_key]
event.send(True)
def check_obj(self, target, recurse=0):
"""
Check one object version.
If no version is specified, all versions of the object will be checked.
:returns: the result of the check of the most recent version,
or the one that is explicitly targeted.
"""
account = target.account
container = target.container
obj = target.obj
vers = target.version # can be None
cached = self._get_cached_or_lock((account, container, obj, vers))
if cached is not None:
return cached
self.logger.info('Checking object "%s"', target)
container_listing, _ = self.check_container(target.copy_container())
errors = list()
if obj not in container_listing:
errors.append('Missing from container listing')
# checksum = None
else:
versions = container_listing[obj]
if vers is None:
if target.content_id is None:
# No version specified, check all versions
self.check_obj_versions(target.copy_object(),
versions,
recurse=recurse)
# Now return the cached result of the most recent version
target.content_id = versions[0]['id']
target.version = str(versions[0]['version'])
res = self.check_obj(target, recurse=0)
self._unlock((account, container, obj, vers))
return res
else:
for ov in versions:
if ov['id'] == target.content_id:
vers = str(ov['version'])
target.version = vers
break
else:
errors.append('Missing from container listing')
# TODO check checksum match
# checksum = container_listing[obj]['hash']
pass
meta, chunks = self._load_obj_meta(target, errors)
chunk_listing = {c['url']: c for c in chunks}
if meta:
if target.content_id is None:
target.content_id = meta['id']
if target.version is None:
target.version = str(meta['version'])
self.list_cache[(account, container, obj, vers)] = \
(chunk_listing, meta)
self.objects_checked += 1
self._unlock((account, container, obj, vers))
# Skip the check if we could not locate the object
if meta:
errors.extend(
self._check_obj_policy(target, meta, chunks, recurse=recurse))
self.send_result(target, errors)
return chunk_listing, meta
def check_container(self, target, recurse=0):
account = target.account
container = target.container
cached = self._get_cached_or_lock((account, container))
if cached is not None:
return cached
self.logger.info('Checking container "%s"', target)
account_listing = self.check_account(target.copy_account())
errors = list()
if container not in account_listing:
errors.append('Missing from account listing')
marker = None
results = []
ct_meta = dict()
extra_args = dict()
if self.limit_listings > 1 and target.obj:
# When we are explicitly checking one object, start the listing
# where this object is supposed to be. Do not use a limit,
# but an end marker, in order to fetch all versions of the object.
extra_args['prefix'] = target.obj
extra_args['end_marker'] = target.obj + '\x00' # HACK
while True:
try:
resp = self.api.object_list(account,
container,
marker=marker,
versions=True,
**extra_args)
except exc.NoSuchContainer as err:
self.container_not_found += 1
errors.append('Not found: %s' % (err, ))
break
except Exception as err:
self.container_exceptions += 1
errors.append('Check failed: %s' % (err, ))
break
truncated = resp.get('truncated', False)
if truncated:
marker = resp['next_marker']
if resp['objects']:
# safeguard, probably useless
if not marker:
marker = resp['objects'][-1]['name']
results.extend(resp['objects'])
if not truncated or self.limit_listings > 1:
break
else:
ct_meta = resp
ct_meta.pop('objects')
break
container_listing = dict()
# Save all object versions, with the most recent first
for obj in results:
container_listing.setdefault(obj['name'], list()).append(obj)
for versions in container_listing.values():
versions.sort(key=lambda o: o['version'], reverse=True)
if self.limit_listings <= 1:
# We just listed the whole container, keep the result in a cache
self.containers_checked += 1
self.list_cache[(account, container)] = container_listing, ct_meta
self._unlock((account, container))
if recurse > 0:
for obj_vers in container_listing.values():
for obj in obj_vers:
tcopy = target.copy_object()
tcopy.obj = obj['name']
tcopy.content_id = obj['id']
tcopy.version = str(obj['version'])
self._spawn_n(self.check_obj, tcopy, recurse - 1)
self.send_result(target, errors)
return container_listing, ct_meta
def check_account(self, target, recurse=0):
account = target.account
cached = self._get_cached_or_lock(account)
if cached is not None:
return cached
self.logger.info('Checking account "%s"', target)
errors = list()
marker = None
results = []
extra_args = dict()
if self.limit_listings > 0 and target.container:
# When we are explicitly checking one container, start the listing
# where this container is supposed to be, and list only one
# container.
extra_args['prefix'] = target.container
extra_args['limit'] = 1
while True:
try:
resp = self.api.container_list(account,
marker=marker,
**extra_args)
except Exception as err:
self.account_exceptions += 1
errors.append('Check failed: %s' % (err, ))
break
if resp:
marker = resp[-1][0]
results.extend(resp)
if self.limit_listings > 0:
break
else:
break
containers = dict()
for container in results:
# Name, number of objects, number of bytes
containers[container[0]] = (container[1], container[2])
if self.limit_listings <= 0:
# We just listed the whole account, keep the result in a cache
self.accounts_checked += 1
self.list_cache[account] = containers
self._unlock(account)
if recurse > 0:
for container in containers:
tcopy = target.copy_account()
tcopy.container = container
self._spawn_n(self.check_container, tcopy, recurse - 1)
self.send_result(target, errors)
return containers
def check(self, target, recurse=0):
if target.type == 'chunk':
self._spawn_n(self.check_chunk, target)
elif target.type == 'object':
self._spawn_n(self.check_obj, target, recurse)
elif target.type == 'container':
self._spawn_n(self.check_container, target, recurse)
else:
self._spawn_n(self.check_account, target, recurse)
def check_all_accounts(self, recurse=0):
all_accounts = self.api.account_list()
for acct in all_accounts:
self.check(Target(acct), recurse=recurse)
def fetch_results(self, rate_limiter=None):
while self.running and not self.result_queue.empty():
res = self.result_queue.get(True)
yield res
# Rate limiting is done on the result queue for now.
# Someday we could implement a submission queue instead of
# letting each worker submit tasks to the pool, and do
# the rate limiting on this queue.
if rate_limiter is not None:
self.run_time = rate_limiter(self.run_time)
def merge_with_delayed_target(self, target):
"""
Merge the specified target with a delayed one.
:returns: the delayed target, if there is one, with an error log
including the errors of the new target. Return the new target
otherwise.
"""
tkey = repr(target)
prev_target = self.delayed_targets.get(tkey, target)
if prev_target is not target:
errors = dict(prev_target.error_log)
errors.update(target.error_log)
prev_target.error_log = sorted(errors.items())
return prev_target
def log_result(self, target):
"""
Log a check result, if it shows errors. Dispatch the errors to the
appropriate destinations (log files, queues, etc.).
"""
# The result may come from a new target, or from an old target
# we checked another time, or both.
target = self.merge_with_delayed_target(target)
if target.has_errors:
time_in_error, confirmations = target.time_in_error()
if (time_in_error < self.min_time_in_error
or confirmations < self.required_confirmations):
self.logger.info("Delaying check for %s, %d/%d confirmations",
target, confirmations,
self.required_confirmations)
self.delayed_targets[repr(target)] = target
else:
if target.type == 'chunk':
self.logger.info(
"Writing error for %s, %d/%d confirmations", target,
confirmations, self.required_confirmations)
self.write_chunk_error(target,
irreparable=target.irreparable)
else:
self.write_error(target, irreparable=target.irreparable)
self.delayed_targets.pop(repr(target), None)
self.logger.warn(
'%s:%s\n%s', target,
' irreparable' if target.irreparable else '',
target.latest_error_result().errors_to_str(err_format=' %s'))
def run(self, rate_limiter=None):
"""
Fetch results and write logs until all jobs have finished.
:returns: a generator yielding check results.
"""
while self.running and (self.pool.running() + self.pool.waiting()):
for result in self.fetch_results(rate_limiter):
self.log_result(result)
yield result
sleep(0.1)
if self.running:
self.pool.waitall()
# No rate limiting
for result in self.fetch_results():
self.log_result(result)
yield result
self.list_cache = CacheDict(self.list_cache.size)
def stop(self):
self.logger.info("Stopping")
self.running = False
def report(self):
success = True
def _report_stat(name, stat):
print("{0:18}: {1}".format(name, stat))
print()
print('Report')
_report_stat("Accounts checked", self.accounts_checked)
if self.account_not_found:
success = False
_report_stat("Missing accounts", self.account_not_found)
if self.account_exceptions:
success = False
_report_stat("Exceptions", self.account_exceptions)
print()
_report_stat("Containers checked", self.containers_checked)
if self.container_not_found:
success = False
_report_stat("Missing containers", self.container_not_found)
if self.container_exceptions:
success = False
_report_stat("Exceptions", self.container_exceptions)
print()
_report_stat("Objects checked", self.objects_checked)
if self.object_not_found:
success = False
_report_stat("Missing objects", self.object_not_found)
if self.object_exceptions:
success = False
_report_stat("Exceptions", self.object_exceptions)
print()
_report_stat("Chunks checked", self.chunks_checked)
if self.chunk_not_found:
success = False
_report_stat("Missing chunks", self.chunk_not_found)
if self.chunk_exceptions:
success = False
_report_stat("Exceptions", self.chunk_exceptions)
return success
class Checker(object):
def __init__(self,
namespace,
concurrency=50,
error_file=None,
rebuild_file=None,
full=True,
limit_listings=0,
request_attempts=1,
logger=None,
verbose=False,
integrity=False):
self.pool = GreenPool(concurrency)
self.error_file = error_file
self.full = bool(full)
self.integrity = bool(integrity)
# Optimisation for when we are only checking one object
# or one container.
# 0 -> do not limit
# 1 -> limit account listings (list of containers)
# 2 -> limit container listings (list of objects)
self.limit_listings = limit_listings
if self.error_file:
outfile = open(self.error_file, 'a')
self.error_writer = csv.writer(outfile, delimiter=' ')
self.rebuild_file = rebuild_file
if self.rebuild_file:
fd = open(self.rebuild_file, 'a')
self.rebuild_writer = csv.writer(fd, delimiter='|')
self.logger = logger or get_logger(
{'namespace': namespace}, name='integrity', verbose=verbose)
self.api = ObjectStorageApi(namespace,
logger=self.logger,
max_retries=request_attempts - 1,
request_attempts=request_attempts)
self.rdir_client = RdirClient({"namespace": namespace},
logger=self.logger)
self.accounts_checked = 0
self.containers_checked = 0
self.objects_checked = 0
self.chunks_checked = 0
self.account_not_found = 0
self.container_not_found = 0
self.object_not_found = 0
self.chunk_not_found = 0
self.account_exceptions = 0
self.container_exceptions = 0
self.object_exceptions = 0
self.chunk_exceptions = 0
self.list_cache = {}
self.running = {}
self.running_lock = Semaphore(1)
self.result_queue = Queue()
def complete_target_from_chunk_metadata(self, target, xattr_meta):
"""
Complete a Target object from metadata found in chunk's extended
attributes. In case the "fullpath" is not available, try to read
legacy metadata, and maybe ask meta1 to resolve the CID into
account and container names.
"""
# pylint: disable=unbalanced-tuple-unpacking
try:
acct, ct, path, vers, content_id = \
decode_fullpath(xattr_meta['full_path'])
target.account = acct
target.container = ct
target.obj = path
target.content_id = content_id
target.version = vers
except KeyError:
# No fullpath header, try legacy headers
if 'content_path' in xattr_meta:
target.obj = xattr_meta['content_path']
if 'content_id' in xattr_meta:
target.content_id = xattr_meta['content_id']
if 'content_version' in xattr_meta:
target.version = xattr_meta['content_version']
cid = xattr_meta.get('container_id')
if cid:
try:
md = self.api.directory.show(cid=cid)
acct = md.get('account')
ct = md.get('name')
if acct:
target.account = acct
if ct:
target.container = ct
except Exception as err:
self.logger.warn(
"Failed to resolve CID %s into account "
"and container names: %s", cid, err)
def recover_and_complete_object_meta(self, target, chunk):
_, rawx_service, chunk_id = chunk.rsplit('/', 2)
# 1. Fetch chunk list from rdir (could be cached).
# Unfortunately we cannot seek for a chunk ID.
entries = [
x for x in self.rdir_client.chunk_fetch(rawx_service, limit=-1)
if x[2] == chunk_id
]
if not entries:
self.logger.warn('Chunk %s not found in rdir' % chunk_id)
return
elif len(entries) > 1:
self.logger.info('Chunk %s appears in %d objects', chunk_id,
len(entries))
# 2. Find content and container IDs
target.cid, target.content_id = entries[0][0:2]
meta = self.api.object_get_properties(None,
None,
None,
cid=target.cid,
content=target.content_id)
target.obj = meta['name']
target.version = meta['version']
target.account, target.container = self.api.resolve_cid(target.cid)
def send_result(self, target, errors=None):
"""
Put an item in the result queue.
"""
# TODO(FVE): send to an external queue.
self.result_queue.put(ItemResult(target, errors))
def write_error(self, target, irreparable=False):
if not self.error_file:
return
error = list()
if irreparable:
error.append('#IRREPARABLE')
error.append(target.account)
if target.container:
error.append(target.container)
if target.obj:
error.append(target.obj)
if target.chunk:
error.append(target.chunk)
self.error_writer.writerow(error)
def write_rebuilder_input(self, target, irreparable=False):
# FIXME(FVE): cid can be computed from account and container names
if target.cid:
cid = target.cid
else:
ct_meta = self.list_cache[(target.account, target.container)][1]
try:
cid = ct_meta['system']['sys.name'].split('.', 1)[0]
except KeyError:
cid = ct_meta['properties']['sys.name'].split('.', 1)[0]
error = list()
if irreparable:
error.append('#IRREPARABLE')
error.append(cid)
# FIXME(FVE): ensure we always resolve content_id,
# or pass object version along with object name.
error.append(target.content_id or target.obj)
error.append(target.chunk)
self.rebuild_writer.writerow(error)
def write_chunk_error(self, target, chunk=None, irreparable=False):
if chunk is not None:
target = target.copy()
target.chunk = chunk
self.write_error(target, irreparable=irreparable)
if self.rebuild_file:
self.write_rebuilder_input(target, irreparable=irreparable)
def _check_chunk_xattr(self, target, obj_meta, xattr_meta):
"""
Check coherency of chunk extended attributes with object metadata.
:returns: a list of errors
"""
errors = list()
# Composed position -> erasure coding
attr_prefix = 'meta' if '.' in obj_meta['pos'] else ''
attr_key = attr_prefix + 'chunk_size'
if str(obj_meta['size']) != xattr_meta.get(attr_key):
errors.append(
"'%s' xattr (%s) differs from size in meta2 (%s)" %
(attr_key, xattr_meta.get(attr_key), obj_meta['size']))
attr_key = attr_prefix + 'chunk_hash'
if obj_meta['hash'] != xattr_meta.get(attr_key):
errors.append(
"'%s' xattr (%s) differs from hash in meta2 (%s)" %
(attr_key, xattr_meta.get(attr_key), obj_meta['hash']))
return errors
def _check_chunk(self, target):
"""
Execute various checks on a chunk:
- does it appear in object's chunk list?
- is it reachable?
- are its extended attributes coherent?
:returns: the list of errors encountered,
and the chunk's owner object metadata.
"""
chunk = target.chunk
errors = list()
obj_meta = None
xattr_meta = None
try:
xattr_meta = self.api.blob_client.chunk_head(
chunk, xattr=self.full, check_hash=self.integrity)
except exc.NotFound as err:
self.chunk_not_found += 1
errors.append('Not found: %s' % (err, ))
except exc.FaultyChunk as err:
self.chunk_exceptions += 1
errors.append('Faulty: %r' % (err, ))
except Exception as err:
self.chunk_exceptions += 1
errors.append('Check failed: %s' % (err, ))
if not target.obj:
if xattr_meta:
self.complete_target_from_chunk_metadata(target, xattr_meta)
else:
self.recover_and_complete_object_meta(target, chunk)
if target.obj:
obj_listing, obj_meta = self.check_obj(target.copy_object())
if chunk not in obj_listing:
errors.append('Missing from object listing')
db_meta = dict()
else:
db_meta = obj_listing[chunk]
if db_meta and xattr_meta and self.full:
errors.extend(
self._check_chunk_xattr(target, db_meta, xattr_meta))
self.send_result(target, errors)
self.chunks_checked += 1
return errors, obj_meta
def check_chunk(self, target):
errors, _obj_meta = self._check_chunk(target)
return errors
def _check_metachunk(self, target, stg_met, pos, chunks, recurse=0):
"""
Check that a metachunk has the right number of chunks.
:returns: the list of errors
"""
required = stg_met.expected_chunks
errors = list()
if len(chunks) < required:
missing_chunks = required - len(chunks)
if stg_met.ec:
subs = {x['num'] for x in chunks}
for sub in range(required):
if sub not in subs:
errors.append("Missing chunk at position %d.%d" %
(pos, sub))
else:
for _ in range(missing_chunks):
errors.append("Missing chunk at position %d" % pos)
if recurse > 0:
for chunk in chunks:
tcopy = target.copy()
tcopy.chunk = chunk['url']
chunk_errors, _ = self._check_chunk(tcopy)
if chunk_errors:
# The errors have already been reported by _check_chunk,
# but we must count this chunk among the unusable chunks
# of the current metachunk.
errors.append("Unusable chunk %s at position %s" %
(chunk['url'], chunk['pos']))
irreparable = required - len(errors) < stg_met.min_chunks_to_read
if irreparable:
errors.append(
"Unavailable metachunk at position %s (%d/%d chunks)" %
(pos, required - len(errors), stg_met.expected_chunks))
# Since the "metachunk" is not an official item type,
# this method does not report errors itself. Errors will
# be reported as object errors.
return errors
def _check_obj_policy(self, target, obj_meta, chunks, recurse=0):
"""
Check that the list of chunks of an object matches
the object's storage policy.
:returns: the list of errors encountered
"""
stg_met = STORAGE_METHODS.load(obj_meta['chunk_method'])
chunks_by_pos = _sort_chunks(chunks, stg_met.ec)
tasks = list()
for pos, pchunks in chunks_by_pos.iteritems():
tasks.append((pos,
self.pool.spawn(self._check_metachunk,
target.copy(),
stg_met,
pos,
pchunks,
recurse=recurse)))
errors = list()
for pos, task in tasks:
try:
errors.extend(task.wait())
except Exception as err:
errors.append("Check failed: pos %d: %s" % (pos, err))
return errors
def check_obj_versions(self, target, versions, recurse=0):
"""
Run checks of all versions of the targeted object in parallel.
"""
tasks = list()
for ov in versions:
tcopy = target.copy_object()
tcopy.content_id = ov['id']
tcopy.version = str(ov['version'])
tasks.append((tcopy.version,
self.pool.spawn(self.check_obj,
tcopy,
recurse=recurse)))
errors = list()
for version, task in tasks:
try:
task.wait()
except Exception as err:
errors.append("Check failed: version %s: %s" % (version, err))
if errors:
# Send a result with the target without version to tell
# we were not able to check all versions of the object.
self.send_result(target, errors)
def _load_obj_meta(self, target, errors):
"""
Load object metadata and chunks.
:param target: which object to check.
:param errors: list of errors that will be appended
in case any error occurs.
:returns: a tuple with object metadata and a list of chunks.
"""
try:
return self.api.object_locate(target.account,
target.container,
target.obj,
version=target.version,
properties=False)
except exc.NoSuchObject as err:
self.object_not_found += 1
errors.append('Not found: %s' % (err, ))
except Exception as err:
self.object_exceptions += 1
errors.append('Check failed: %s' % (err, ))
return None, []
def _get_cached_or_lock(self, lock_key):
# If something is running, wait for it
with self.running_lock:
event = self.running.get(lock_key)
if event:
event.wait()
event = None
# Maybe get a cached result
if lock_key in self.list_cache:
return self.list_cache[lock_key]
# No cached result, try to compute the thing ourselves
while True:
with self.running_lock:
# Another check while locked
if lock_key in self.list_cache:
return self.list_cache[lock_key]
# Still nothing cached
event = self.running.get(lock_key)
if event is None:
self.running[lock_key] = Event()
return None
event.wait()
def _unlock(self, lock_key):
with self.running_lock:
event = self.running[lock_key]
del self.running[lock_key]
event.send(True)
def check_obj(self, target, recurse=0):
"""
Check one object version.
If no version is specified, all versions of the object will be checked.
:returns: the result of the check of the most recent version,
or the one that is explicitly targeted.
"""
account = target.account
container = target.container
obj = target.obj
vers = target.version # can be None
cached = self._get_cached_or_lock((account, container, obj, vers))
if cached:
return cached
self.logger.info('Checking object "%s"', target)
container_listing, _ = self.check_container(target.copy_container())
errors = list()
if obj not in container_listing:
errors.append('Missing from container listing')
# checksum = None
else:
versions = container_listing[obj]
if vers is None:
if target.content_id is None:
# No version specified, check all versions
self.check_obj_versions(target.copy_object(),
versions,
recurse=recurse)
# Now return the cached result of the most recent version
target.content_id = versions[0]['id']
target.version = str(versions[0]['version'])
res = self.check_obj(target, recurse=0)
self._unlock((account, container, obj, vers))
return res
else:
for ov in versions:
if ov['id'] == target.content_id:
vers = str(ov['version'])
target.version = vers
break
else:
errors.append('Missing from container listing')
# TODO check checksum match
# checksum = container_listing[obj]['hash']
pass
meta, chunks = self._load_obj_meta(target, errors)
chunk_listing = {c['url']: c for c in chunks}
if meta:
if target.content_id is None:
target.content_id = meta['id']
if target.version is None:
target.version = str(meta['version'])
self.list_cache[(account, container, obj, vers)] = \
(chunk_listing, meta)
self.objects_checked += 1
self._unlock((account, container, obj, vers))
# Skip the check if we could not locate the object
if meta:
errors.extend(
self._check_obj_policy(target, meta, chunks, recurse=recurse))
self.send_result(target, errors)
return chunk_listing, meta
def check_container(self, target, recurse=0):
account = target.account
container = target.container
cached = self._get_cached_or_lock((account, container))
if cached:
return cached
self.logger.info('Checking container "%s"', target)
account_listing = self.check_account(target.copy_account())
errors = list()
if container not in account_listing:
errors.append('Missing from account listing')
marker = None
results = []
ct_meta = dict()
extra_args = dict()
if self.limit_listings > 1 and target.obj:
# When we are explicitly checking one object, start the listing
# where this object is supposed to be. Do not use a limit,
# but an end marker, in order to fetch all versions of the object.
extra_args['prefix'] = target.obj
extra_args['end_marker'] = target.obj + '\x00' # HACK
while True:
try:
resp = self.api.object_list(account,
container,
marker=marker,
versions=True,
**extra_args)
except exc.NoSuchContainer as err:
self.container_not_found += 1
errors.append('Not found: %s' % (err, ))
break
except Exception as err:
self.container_exceptions += 1
errors.append('Check failed: %s' % (err, ))
break
if resp.get('truncated', False):
marker = resp['next_marker']
if resp['objects']:
# safeguard, probably useless
if not marker:
marker = resp['objects'][-1]['name']
results.extend(resp['objects'])
if self.limit_listings > 1:
break
else:
ct_meta = resp
ct_meta.pop('objects')
break
container_listing = dict()
# Save all object versions, with the most recent first
for obj in results:
container_listing.setdefault(obj['name'], list()).append(obj)
for versions in container_listing.values():
versions.sort(key=lambda o: o['version'], reverse=True)
if self.limit_listings <= 1:
# We just listed the whole container, keep the result in a cache
self.containers_checked += 1
self.list_cache[(account, container)] = container_listing, ct_meta
self._unlock((account, container))
if recurse > 0:
for obj_vers in container_listing.values():
for obj in obj_vers:
tcopy = target.copy_object()
tcopy.obj = obj['name']
tcopy.content_id = obj['id']
tcopy.version = str(obj['version'])
self.pool.spawn_n(self.check_obj, tcopy, recurse - 1)
self.send_result(target, errors)
return container_listing, ct_meta
def check_account(self, target, recurse=0):
account = target.account
cached = self._get_cached_or_lock(account)
if cached:
return cached
self.logger.info('Checking account "%s"', target)
errors = list()
marker = None
results = []
extra_args = dict()
if self.limit_listings > 0 and target.container:
# When we are explicitly checking one container, start the listing
# where this container is supposed to be, and list only one
# container.
extra_args['prefix'] = target.container
extra_args['limit'] = 1
while True:
try:
resp = self.api.container_list(account,
marker=marker,
**extra_args)
except Exception as err:
self.account_exceptions += 1
errors.append('Check failed: %s' % (err, ))
break
if resp:
marker = resp[-1][0]
results.extend(resp)
if self.limit_listings > 0:
break
else:
break
containers = dict()
for container in results:
# Name, number of objects, number of bytes
containers[container[0]] = (container[1], container[2])
if self.limit_listings <= 0:
# We just listed the whole account, keep the result in a cache
self.accounts_checked += 1
self.list_cache[account] = containers
self._unlock(account)
if recurse > 0:
for container in containers:
tcopy = target.copy_account()
tcopy.container = container
self.pool.spawn_n(self.check_container, tcopy, recurse - 1)
self.send_result(target, errors)
return containers
def check(self, target, recurse=0):
if target.type == 'chunk':
self.pool.spawn_n(self.check_chunk, target)
elif target.type == 'object':
self.pool.spawn_n(self.check_obj, target, recurse)
elif target.type == 'container':
self.pool.spawn_n(self.check_container, target, recurse)
else:
self.pool.spawn_n(self.check_account, target, recurse)
def check_all_accounts(self, recurse=0):
all_accounts = self.api.account_list()
for acct in all_accounts:
self.check(Target(acct), recurse=recurse)
def fetch_results(self):
while not self.result_queue.empty():
res = self.result_queue.get(True)
yield res
def log_result(self, result):
if result.errors:
if result.target.type == 'chunk':
# FIXME(FVE): check error criticity
# and set the irreparable flag.
self.write_chunk_error(result.target)
else:
self.write_error(result.target)
self.logger.warn('%s:\n%s', result.target,
result.errors_to_str(err_format=' %s'))
def run(self):
"""
Fetch results and write logs until all jobs have finished.
:returns: a generator yielding check results.
"""
while self.pool.running() + self.pool.waiting():
for result in self.fetch_results():
self.log_result(result)
yield result
sleep(0.1)
self.pool.waitall()
for result in self.fetch_results():
self.log_result(result)
yield result
def report(self):
success = True
def _report_stat(name, stat):
print("{0:18}: {1}".format(name, stat))
print()
print('Report')
_report_stat("Accounts checked", self.accounts_checked)
if self.account_not_found:
success = False
_report_stat("Missing accounts", self.account_not_found)
if self.account_exceptions:
success = False
_report_stat("Exceptions", self.account_exceptions)
print()
_report_stat("Containers checked", self.containers_checked)
if self.container_not_found:
success = False
_report_stat("Missing containers", self.container_not_found)
if self.container_exceptions:
success = False
_report_stat("Exceptions", self.container_exceptions)
print()
_report_stat("Objects checked", self.objects_checked)
if self.object_not_found:
success = False
_report_stat("Missing objects", self.object_not_found)
if self.object_exceptions:
success = False
_report_stat("Exceptions", self.object_exceptions)
print()
_report_stat("Chunks checked", self.chunks_checked)
if self.chunk_not_found:
success = False
_report_stat("Missing chunks", self.chunk_not_found)
if self.chunk_exceptions:
success = False
_report_stat("Exceptions", self.chunk_exceptions)
return success
class TestContainerDownload(BaseTestCase):
def setUp(self):
super(TestContainerDownload, self).setUp()
# FIXME: should we use direct API from BaseTestCase
# or still container.client ?
self.conn = ObjectStorageApi(self.ns)
self._streaming = 'http://' + self.get_service_url('container')[2]
self._cnt = random_container()
self._uri = self.make_uri('dump')
self._data = {}
self.conn.container_create(self.account, self._cnt)
self.raw = ""
self._slo = []
def make_uri(self, action, account=None, container=None):
account = account or self.account
container = container or self._cnt
return '%s/v1.0/container/%s?acct=%s&ref=%s' % (
self._streaming, action, account, container)
def tearDown(self):
for name in self._data:
self.conn.object_delete(self.account, self._cnt, name)
self.conn.container_delete(self.account, self._cnt)
super(TestContainerDownload, self).tearDown()
def _create_data(self,
name=gen_names,
metadata=None,
size=513,
append=False):
for idx, _name in itertools.islice(name(), 5):
mime = random.choice(MIMETYPE)
if append and size > 0:
data = gen_data(size / 2 * idx)
entry = {'data': data, 'meta': None, 'mime': mime}
self.conn.object_create(self.account,
self._cnt,
obj_name=_name,
data=data,
mime_type=mime)
data = gen_data(size / 2 * idx)
self.conn.object_create(self.account,
self._cnt,
obj_name=_name,
data=data,
mime_type=mime,
append=True)
entry['data'] += data
else:
data = gen_data(size * idx)
entry = {'data': data, 'meta': None, 'mime': mime}
self.conn.object_create(self.account,
self._cnt,
obj_name=_name,
data=data,
mime_type=mime)
if metadata:
entry['meta'] = {}
for _ in xrange(10):
key, val = metadata()
entry['meta'][key] = val
self.conn.object_update(self.account, self._cnt, _name,
entry['meta'])
self._data[_name] = entry
def _create_s3_slo(self, name=gen_names, metadata=None):
# create a fake S3 bucket with a SLO object
chunksize = 10000
parts = 5
res = []
full_data = ""
self.conn.container_create(self.account, self._cnt + '+segments')
_name = "toto"
etag = rand_str(50)
part_number = 1
for size in [chunksize] * parts + [444]:
data = gen_data(size)
res.append({
'bytes':
size,
'content_type':
'application/octect-stream',
'hash':
md5(data).hexdigest().upper(),
'last_modified':
'2017-06-21T12:42:47.000000',
'name':
'/%s+segments/%s/%s/%d' % (self._cnt, _name, etag, part_number)
})
self.conn.object_create(self.account,
"%s+segments" % self._cnt,
obj_name='%s/%s/%d' %
(_name, etag, part_number),
data=data)
full_data += data
part_number += 1
self._data[_name] = {
'data': full_data,
'meta': {
'x-static-large-object': 'true',
'x-object-sysmeta-slo-etag': etag,
'x-object-sysmeta-slo-size': str(len(full_data))
}
}
self._slo.append(_name)
data = json.dumps(res)
self.conn.object_create(self.account,
self._cnt,
obj_name=_name,
data=data)
self.conn.object_update(self.account, self._cnt, _name,
self._data[_name]['meta'])
def _check_tar(self, data):
raw = BytesIO(data)
tar = tarfile.open(fileobj=raw, ignore_zeros=True)
info = self._data.keys()
for entry in tar.getnames():
if entry == CONTAINER_MANIFEST:
# skip special entry
continue
self.assertIn(entry, info)
tmp = tar.extractfile(entry)
self.assertEqual(self._data[entry]['data'], tmp.read())
info.remove(entry)
self.assertEqual(info, [])
return tar
def _check_container(self, cnt):
ret = self.conn.object_list(account=self.account, container=cnt)
names = self._data.keys()
for obj in ret['objects']:
name = obj['name']
self.assertIn(name, self._data)
self.assertEqual(obj['size'], len(self._data[name]['data']))
_, data = self.conn.object_fetch(self.account, cnt, name)
raw = "".join(data)
self.assertEqual(
md5(raw).hexdigest(),
md5(self._data[name]['data']).hexdigest())
meta = self.conn.object_get_properties(self.account, cnt, name)
self.assertEqual(meta['properties'], self._data[name]['meta'])
names.remove(name)
self.assertEqual(len(names), 0)
def _simple_download(self,
name=gen_names,
metadata=None,
size=513,
append=False):
self._create_data(name=name,
metadata=metadata,
size=size,
append=append)
ret = requests.get(self._uri)
self.assertGreater(len(ret.content), 0)
self.assertEqual(ret.status_code, 200)
self.raw = ret.content
return self._check_tar(ret.content)
def _check_metadata(self, tar):
for entry in tar.getnames():
if entry == CONTAINER_MANIFEST:
# skip special entry
continue
headers = tar.getmember(entry).pax_headers
keys = headers.keys()[:]
for key, val in self._data[entry]['meta'].items():
key = u"SCHILY.xattr.user." + key.decode('utf-8')
self.assertIn(key, headers)
self.assertEqual(val.decode('utf-8'), headers[key])
keys.remove(key)
#
self.assertEqual(self._data[entry]['mime'], headers['mime_type'])
keys.remove('mime_type')
#
self.assertEqual(keys, [])
def test_missing_container(self):
ret = requests.get(self._streaming + '/' + random_container("ms-"))
self.assertEqual(ret.status_code, 404)
def test_invalid_url(self):
ret = requests.get(self._streaming)
self.assertEqual(ret.status_code, 404)
ret = requests.head(self._streaming + '/' + random_container('inv') +
'/' + random_container('inv'))
self.assertEqual(ret.status_code, 404)
def test_download_empty_container(self):
ret = requests.get(self._uri)
self.assertEqual(ret.status_code, 204)
def test_simple_download(self):
self._simple_download()
def test_check_head(self):
self._create_data()
get = requests.get(self._uri)
head = requests.head(self._uri)
self.assertEqual(get.headers['content-length'],
head.headers['content-length'])
def test_download_per_range(self):
self._create_data()
org = requests.get(self._uri)
data = []
for idx in xrange(0, int(org.headers['content-length']), 512):
ret = requests.get(
self._uri, headers={'Range': 'bytes=%d-%d' % (idx, idx + 511)})
self.assertEqual(ret.status_code, 206)
self.assertEqual(len(ret.content), 512)
self.assertEqual(ret.content, org.content[idx:idx + 512])
data.append(ret.content)
data = "".join(data)
self.assertGreater(len(data), 0)
self.assertEqual(md5(data).hexdigest(), md5(org.content).hexdigest())
def test_invalid_range(self):
self._create_data()
ranges = ((-512, 511), (512, 0), (1, 3), (98888, 99999))
for start, end in ranges:
ret = requests.get(self._uri,
headers={'Range': 'bytes=%d-%d' % (start, end)})
self.assertEqual(
ret.status_code, 416,
"Invalid error code for range %d-%d" % (start, end))
ret = requests.get(self._uri,
headers={'Range': 'bytes=0-511, 512-1023'})
self.assertEqual(ret.status_code, 416)
def test_file_metadata(self):
tar = self._simple_download(metadata=gen_metadata)
self._check_metadata(tar)
def test_container_metadata(self):
key, val = gen_metadata()
ret = self.conn.container_update(self.account, self._cnt, {key: val})
ret = self.conn.container_show(self.account, self._cnt)
ret = requests.get(self._uri)
self.assertEqual(ret.status_code, 200)
raw = BytesIO(ret.content)
tar = tarfile.open(fileobj=raw, ignore_zeros=True)
self.assertIn(CONTAINER_PROPERTIES, tar.getnames())
data = json.load(tar.extractfile(CONTAINER_PROPERTIES))
self.assertIn(key, data)
self.assertEqual(val, data[key])
def test_charset_file(self):
self._simple_download(name=gen_charset_names)
@unittest.skip("wip")
def test_byte_metadata(self):
tar = self._simple_download(metadata=gen_byte_metadata)
self._check_metadata(tar)
def test_charset_metadata(self):
tar = self._simple_download(metadata=gen_charset_metadata)
self._check_metadata(tar)
@attr('s3')
def test_s3_simple_download(self):
self._create_s3_slo()
ret = requests.get(self._uri)
self.assertGreater(len(ret.content), 0)
self.assertEqual(ret.status_code, 200)
self.raw = ret.content
raw = BytesIO(ret.content)
tar = tarfile.open(fileobj=raw, ignore_zeros=True)
info = self._data.keys()
for entry in tar.getnames():
if entry == CONTAINER_MANIFEST:
# skip special entry
continue
self.assertIn(entry, info)
tmp = tar.extractfile(entry)
self.assertEqual(self._data[entry]['data'], tmp.read())
info.remove(entry)
self.assertEqual(len(info), 0)
return tar
@attr('s3')
def test_s3_range_download(self):
self._create_s3_slo()
org = requests.get(self._uri)
self.assertEqual(org.status_code, 200)
data = []
for idx in xrange(0, int(org.headers['content-length']), 512):
ret = requests.get(
self._uri, headers={'Range': 'bytes=%d-%d' % (idx, idx + 511)})
self.assertEqual(ret.status_code, 206)
self.assertEqual(len(ret.content), 512)
self.assertEqual(ret.content, org.content[idx:idx + 512])
data.append(ret.content)
data = "".join(data)
self.assertGreater(len(data), 0)
self.assertEqual(md5(data).hexdigest(), md5(org.content).hexdigest())
@attr('s3')
def test_s3_check_slo_metadata_download(self):
self._create_s3_slo()
org = requests.get(self.make_uri('dump'))
self.assertEqual(org.status_code, 200)
cnt = rand_str(20)
res = requests.put(self.make_uri('restore', container=cnt),
data=org.content)
self.assertEqual(org.status_code, 200)
res = self.conn.object_get_properties(self.account, cnt, self._slo[0])
props = res['properties']
self.assertNotIn('x-static-large-object', props)
self.assertNotIn('x-object-sysmeta-slo-size', props)
self.assertNotIn('x-object-sysmeta-slo-etag', props)
@attr('simple')
def test_simple_restore(self):
self._create_data(metadata=gen_metadata)
org = requests.get(self.make_uri('dump'))
cnt = rand_str(20)
res = requests.put(self.make_uri('restore', container=cnt),
data=org.content)
self.assertEqual(res.status_code, 201)
self._check_container(cnt)
@attr('restore')
def test_multipart_restore(self):
self._create_data(metadata=gen_metadata, size=1025 * 1024)
org = requests.get(self.make_uri('dump'))
cnt = rand_str(20)
size = 1014 * 1024
parts = [
org.content[x:x + size] for x in xrange(0, len(org.content), size)
]
uri = self.make_uri('restore', container=cnt)
start = 0
for part in parts:
hdrs = {'Range': 'bytes=%d-%d' % (start, start + len(part) - 1)}
res = requests.put(uri, data=part, headers=hdrs)
start += len(part)
self.assertIn(res.status_code, [201, 206])
self._check_container(cnt)
@attr('restore')
def test_multipart_invalid_restore(self):
self._create_data(metadata=gen_metadata, size=1025 * 1024)
org = requests.get(self.make_uri('dump'))
cnt = rand_str(20)
uri = self.make_uri('restore', container=cnt)
size = 1014 * 1024
parts = [
org.content[x:x + size] for x in xrange(0, len(org.content), size)
]
start = 0
for part in parts:
hdrs = {'Range': 'bytes=%d-%d' % (start, start + len(part) - 1)}
res = requests.put(uri, data=part, headers=hdrs)
self.assertIn(res.status_code, [201, 206])
start += len(part)
# only unfinished restoration expose X-Consumed-Size
if res.status_code == 206:
res = requests.head(uri)
self.assertEqual(int(res.headers['X-Consumed-Size']), start)
inv = requests.put(uri, data=part, headers=hdrs)
self.assertEqual(inv.status_code, 422)
if res.status_code == 206:
res = requests.head(uri)
self.assertEqual(int(res.headers['X-Consumed-Size']), start)
uri = self.make_uri('restore', container=rand_str(20))
hdrs = {'Range': 'bytes=%d-%d' % (size, size + len(parts[1]) - 1)}
res = requests.put(uri, data=part, headers=hdrs)
self.assertEqual(res.status_code, 422)
self._check_container(cnt)
@attr('concurrency')
def test_multipart_concurrency(self):
self._create_data(metadata=gen_metadata, size=1025 * 1024)
org = requests.get(self.make_uri('dump'))
cnt = rand_str(20)
uri = self.make_uri('restore', container=cnt)
size = divmod(len(org.content) / 3, 512)[0] * 512
parts = [
org.content[x:x + size] for x in xrange(0, len(org.content), size)
]
start = 0
class StreamWithContentLength(Thread):
"""Thread to send data with delays to restore API"""
def __init__(self, data, headers):
self._count = 0
self._data = data
self._hdrs = headers
super(StreamWithContentLength, self).__init__()
def __len__(self):
return len(self._data)
def read(self, *args):
if self._count < len(self._data):
time.sleep(0.5)
data = self._data[self._count:self._count + size / 3]
self._count += len(data)
return data
return ""
def run(self):
self._ret = requests.put(uri, data=self, headers=self._hdrs)
for idx, part in enumerate(parts):
hdrs = {'Range': 'bytes=%d-%d' % (start, start + len(part) - 1)}
if idx == 0:
res = requests.put(uri, data=part, headers=hdrs)
self.assertIn(res.status_code, [201, 206])
else:
# launch Thread and simulate slow bandwidth
thr = StreamWithContentLength(part, hdrs)
thr.start()
# send data on same range
time.sleep(0.5)
res = requests.put(uri, data=part, headers=hdrs)
self.assertEqual(res.status_code, 422)
thr.join()
self.assertIn(thr._ret.status_code, [201, 206])
start += len(part)
self._check_container(cnt)
@attr('disconnected')
def test_broken_connectivity(self):
self._create_data(metadata=gen_metadata, size=1025 * 1024)
org = requests.get(self.make_uri('dump'))
cnt = rand_str(20)
class FakeStream(object):
"""Send data and simulate a connectivity issue"""
def __init__(self, data, size):
self._count = 0
self._data = data
self._size = size
def __len__(self):
return len(self._data)
def read(self, *args):
if self._count < self._size:
data = self._data[self._count:self._count + size / 3]
self._count += len(data)
return data
if self._count == len(self._data):
return ""
raise Exception("break connection")
def wait_lock():
"""When the lock is gone, return current consumed size"""
nb = 0
while True:
time.sleep(0.1)
req = requests.head(uri)
if (req.status_code == 200 and req.headers.get(
'X-Upload-In-Progress', '1') == '0'):
print("Tried before lock free", nb)
print("Got consumed-size", req.headers['X-Consumed-Size'])
return int(req.headers['X-Consumed-Size'])
nb += 1
self.assertLess(nb, 10)
uri = self.make_uri('restore', container=cnt)
block = 1000 * 512
start = 0
cut = False
while True:
if start:
start = wait_lock()
stop = min(len(org.content), start + block)
hdrs = {'Range': 'bytes=%d-%d' % (start, stop - 1)}
size = stop - start
if cut:
size = block / 2
cut = not cut
try:
ret = requests.put(uri,
headers=hdrs,
data=FakeStream(org.content[start:stop],
size))
except Exception:
pass
else:
self.assertIn(
ret.status_code, (201, 206),
"Unexpected %d HTTP response: %s" %
(ret.status_code, ret.content))
start += size
if ret.status_code == 201:
break
result = requests.get(self.make_uri('dump', container=cnt))
self._check_tar(result.content)
@attr('rawtar')
def test_rawtar(self):
"""Create a normal tar archive and restore it"""
raw = BytesIO()
tarfile = TarFile(mode='w', fileobj=raw)
testdata = rand_str(20) * 5000
inf = TarInfo("simpletar")
fileraw = BytesIO()
fileraw.write(testdata)
inf.size = len(testdata)
fileraw.seek(0)
tarfile.addfile(inf, fileobj=fileraw)
tarfile.close()
raw.seek(0)
data = raw.read()
cnt = rand_str(20)
ret = requests.put(self.make_uri("restore", container=cnt), data=data)
self.assertEqual(ret.status_code, 201)
meta, stream = self.conn.object_fetch(self.account, cnt, "simpletar")
self.assertEqual(
md5("".join(stream)).hexdigest(),
md5(testdata).hexdigest())
@attr('invalid')
def test_checksums(self):
"""Check restore operation with invalid tar"""
tar = self._simple_download(append=True)
manifest = json.load(tar.extractfile(CONTAINER_MANIFEST),
object_pairs_hook=OrderedDict)
# => add random bytes inside each file (either header and data)
for entry in manifest:
if entry['name'] == CONTAINER_MANIFEST:
# CONTAINER_MANIFEST does not have checksum at this time
continue
inv = self.raw
# Test with tar entry
# checksum tar doesn't work very well with SCHILY attributes
# so only apply changes on regular block entry
idx = entry['start_block'] * BLOCKSIZE \
+ random.randint(0, BLOCKSIZE)
# + random.randint(0, entry['hdr_blocks'] * BLOCKSIZE)
while self.raw[idx] == inv[idx]:
inv = inv[:idx] + chr(random.randint(0, 255)) + inv[idx + 1:]
cnt = rand_str(20)
res = requests.put(self.make_uri('restore', container=cnt),
data=inv)
self.assertEqual(res.status_code, 400)
# skip emty file
if entry['size'] == 0:
continue
# Test with data blocks
inv = self.raw
idx = (entry['start_block'] + entry['hdr_blocks']) * BLOCKSIZE \
+ random.randint(0, entry['size'] - 1)
while self.raw[idx] == inv[idx]:
inv = inv[:idx] + chr(random.randint(0, 255)) + inv[idx + 1:]
cnt = rand_str(20)
res = requests.put(self.make_uri('restore', container=cnt),
data=inv)
self.assertEqual(res.status_code, 400)
class TestObjectStorageAPI(BaseTestCase):
def setUp(self):
super(TestObjectStorageAPI, self).setUp()
self.api = ObjectStorageApi(self.ns, endpoint=self.uri)
self.created = list()
def tearDown(self):
super(TestObjectStorageAPI, self).tearDown()
for ct, name in self.created:
try:
self.api.object_delete(self.account, ct, name)
except Exception:
logging.exception("Failed to delete %s/%s/%s//%s", self.ns,
self.account, ct, name)
def _create(self, name, metadata=None):
return self.api.container_create(self.account,
name,
properties=metadata)
def _delete(self, name):
self.api.container_delete(self.account, name)
def _clean(self, name, clear=False):
if clear:
# must clean properties before
self.api.container_del_properties(self.account, name, [])
self._delete(name)
def _get_properties(self, name, properties=None):
return self.api.container_get_properties(self.account,
name,
properties=properties)
def _set_properties(self, name, properties=None):
return self.api.container_set_properties(self.account,
name,
properties=properties)
def test_container_show(self):
# container_show on unknown container
name = random_str(32)
self.assertRaises(exc.NoSuchContainer, self.api.container_show,
self.account, name)
self._create(name)
# container_show on existing container
res = self.api.container_show(self.account, name)
self.assertIsNot(res['properties'], None)
self._delete(name)
# container_show on deleted container
self.assertRaises(exc.NoSuchContainer, self.api.container_show,
self.account, name)
def test_container_create(self):
name = random_str(32)
res = self._create(name)
self.assertEqual(res, True)
# second create
res = self._create(name)
self.assertEqual(res, False)
# clean
self._delete(name)
def test_create_properties(self):
name = random_str(32)
metadata = {
random_str(32): random_str(32),
random_str(32): random_str(32),
}
res = self._create(name, metadata)
self.assertEqual(res, True)
data = self._get_properties(name)
self.assertEqual(data['properties'], metadata)
# clean
self._clean(name, True)
def test_container_delete(self):
name = random_str(32)
# container_delete on unknown container
self.assertRaises(exc.NoSuchContainer, self.api.container_delete,
self.account, name)
res = self._create(name)
self.assertEqual(res, True)
# container_delete on existing container
self._delete(name)
# verify deleted
self.assertRaises(exc.NoSuchContainer, self.api.container_show,
self.account, name)
# second delete
self.assertRaises(exc.NoSuchContainer, self.api.container_delete,
self.account, name)
# verify deleted
self.assertRaises(exc.NoSuchContainer, self.api.container_show,
self.account, name)
def test_container_get_properties(self):
name = random_str(32)
# container_get_properties on unknown container
self.assertRaises(exc.NoSuchContainer,
self.api.container_get_properties, self.account,
name)
res = self._create(name)
self.assertEqual(res, True)
# container_get_properties on existing container
data = self.api.container_get_properties(self.account, name)
self.assertEqual(data['properties'], {})
self.assertIsNot(data['system'], None)
self.assertIn("sys.user.name", data['system'])
# container_get_properties
metadata = {
random_str(32): random_str(32),
random_str(32): random_str(32),
}
self._set_properties(name, metadata)
data = self.api.container_get_properties(self.account, name)
self.assertEqual(data['properties'], metadata)
# clean
self._clean(name, True)
# container_get_properties on deleted container
self.assertRaises(exc.NoSuchContainer,
self.api.container_get_properties, self.account,
name)
def test_container_get_properties_filtered(self):
self.skipTest("Server side properties filtering not implemented")
name = random_str(32)
res = self._create(name)
self.assertEqual(res, True)
# container_get_properties on existing container
data = self.api.container_get_properties(self.account, name)
self.assertEqual(data['properties'], {})
# container_get_properties
metadata = {
random_str(32): random_str(32),
random_str(32): random_str(32),
}
self._set_properties(name, metadata)
# container_get_properties specify key
key = metadata.keys().pop(0)
data = self.api.container_get_properties(self.account, name, [key])
self.assertEqual({key: metadata[key]}, data['properties'])
# clean
self._clean(name, True)
def test_container_set_properties(self):
name = random_str(32)
metadata = {
random_str(32): random_str(32),
random_str(32): random_str(32),
}
# container_set_properties on unknown container
self.assertRaises(exc.NoSuchContainer,
self.api.container_set_properties, self.account,
name, metadata)
res = self._create(name)
self.assertEqual(res, True)
# container_set_properties on existing container
self.api.container_set_properties(self.account, name, metadata)
data = self._get_properties(name)
self.assertEqual(data['properties'], metadata)
# container_set_properties
key = random_str(32)
value = random_str(32)
metadata2 = {key: value}
self._set_properties(name, metadata2)
metadata.update(metadata2)
data = self._get_properties(name)
self.assertEqual(data['properties'], metadata)
# container_set_properties overwrite key
key = metadata.keys().pop(0)
value = random_str(32)
metadata3 = {key: value}
metadata.update(metadata3)
self.api.container_set_properties(self.account, name, metadata3)
data = self._get_properties(name)
self.assertEqual(data['properties'], metadata)
# clean
self._clean(name, True)
# container_set_properties on deleted container
self.assertRaises(exc.NoSuchContainer,
self.api.container_set_properties, self.account,
name, metadata)
def test_del_properties(self):
name = random_str(32)
metadata = {
random_str(32): random_str(32),
random_str(32): random_str(32),
}
# container_del_properties on unknown container
self.assertRaises(exc.NoSuchContainer,
self.api.container_del_properties, self.account,
name, [])
res = self._create(name, metadata)
self.assertEqual(res, True)
key = metadata.keys().pop()
del metadata[key]
# container_del_properties on existing container
self.api.container_del_properties(self.account, name, [key])
data = self._get_properties(name)
self.assertNotIn(key, data['properties'])
key = random_str(32)
# We do not check if a property exists before deleting it
# self.assertRaises(
# exc.NoSuchContainer, self.api.container_del_properties,
# self.account, name, [key])
self.api.container_del_properties(self.account, name, [key])
data = self._get_properties(name)
self.assertEqual(data['properties'], metadata)
# clean
self._clean(name, True)
# container_del_properties on deleted container
self.assertRaises(exc.NoSuchContainer,
self.api.container_del_properties, self.account,
name, metadata.keys())
def test_object_create_mime_type(self):
name = random_str(32)
self.api.object_create(self.account,
name,
data="data",
obj_name=name,
mime_type='text/custom')
meta, _ = self.api.object_locate(self.account, name, name)
self.assertEqual(meta['mime_type'], 'text/custom')
def _upload_data(self, name):
chunksize = int(self.conf["chunk_size"])
size = int(chunksize * 12)
data = random_data(int(size))
self.api.object_create(self.account, name, obj_name=name, data=data)
self.created.append((name, name))
_, chunks = self.api.object_locate(self.account, name, name)
logging.debug("Chunks: %s", chunks)
return sort_chunks(chunks, False), data
def _fetch_range(self, name, range_):
if not isinstance(range_[0], tuple):
ranges = (range_, )
else:
ranges = range_
stream = self.api.object_fetch(self.account, name, name,
ranges=ranges)[1]
data = ""
for chunk in stream:
data += chunk
return data
def test_object_fetch_range_start(self):
"""From 0 to somewhere"""
name = random_str(16)
_, data = self._upload_data(name)
end = 666
fdata = self._fetch_range(name, (0, end))
self.assertEqual(len(fdata), end + 1)
self.assertEqual(fdata, data[0:end + 1])
def test_object_fetch_range_end(self):
"""From somewhere to end"""
name = random_str(16)
chunks, data = self._upload_data(name)
start = 666
last = max(chunks.keys())
end = chunks[last][0]['offset'] + chunks[last][0]['size']
fdata = self._fetch_range(name, (start, end))
self.assertEqual(len(fdata), len(data) - start)
self.assertEqual(fdata, data[start:])
def test_object_fetch_range_metachunk_start(self):
"""From the start of the second metachunk to somewhere"""
name = random_str(16)
chunks, data = self._upload_data(name)
start = chunks[1][0]['offset']
end = start + 666
fdata = self._fetch_range(name, (start, end))
self.assertEqual(len(fdata), end - start + 1)
self.assertEqual(fdata, data[start:end + 1])
def test_object_fetch_range_metachunk_end(self):
"""From somewhere to end of the first metachunk"""
name = random_str(16)
chunks, data = self._upload_data(name)
start = 666
end = chunks[0][0]['size'] - 1
fdata = self._fetch_range(name, (start, end))
self.assertEqual(len(fdata), end - start + 1)
self.assertEqual(fdata, data[start:end + 1])
def test_object_fetch_range_2_metachunks(self):
"""
From somewhere in the first metachunk
to somewhere in the second metachunk
"""
name = random_str(16)
chunks, data = self._upload_data(name)
start = 666
end = start + chunks[0][0]['size'] - 1
fdata = self._fetch_range(name, (start, end))
self.assertEqual(len(fdata), end - start + 1)
self.assertEqual(fdata, data[start:end + 1])
def test_object_fetch_several_ranges(self):
"""
Download several ranges at once.
"""
name = random_str(16)
chunks, data = self._upload_data(name)
start = 666
end = start + chunks[0][0]['size'] - 1
fdata = self._fetch_range(name, ((start, end), (end + 1, end + 2)))
self.assertEqual(len(fdata), end - start + 3)
self.assertEqual(fdata, data[start:end + 3])
# Notice that we download some bytes from the second metachunk
# before some from the first.
fdata = self._fetch_range(
name, ((chunks[0][0]['size'], chunks[0][0]['size'] + 2), (0, 1),
(1, 2), (4, 6)))
self.assertEqual(len(fdata), 10)
self.assertEqual(
fdata, data[chunks[0][0]['size']:chunks[0][0]['size'] + 3] +
data[0:2] + data[1:3] + data[4:7])
def test_object_create_then_append(self):
"""Create an object then append data"""
name = random_str(16)
self.api.object_create(self.account,
name,
data="1" * 128,
obj_name=name)
_, size, _ = self.api.object_create(self.account,
name,
data="2" * 128,
obj_name=name,
append=True)
self.assertEqual(size, 128)
_, data = self.api.object_fetch(self.account, name, name)
data = "".join(data)
self.assertEqual(len(data), 256)
self.assertEqual(data, "1" * 128 + "2" * 128)
def test_object_create_from_append(self):
"""Create an object with append operation"""
name = random_str(16)
self.api.container_create(self.account, name)
self.api.object_create(self.account,
name,
data="1" * 128,
obj_name=name,
append=True)
_, data = self.api.object_fetch(self.account, name, name)
data = "".join(data)
self.assertEqual(len(data), 128)
self.assertEqual(data, "1" * 128)
def test_container_object_create_from_append(self):
"""Try to create container and object with append operation"""
name = random_str(16)
_chunks, size, checksum = self.api.object_create(self.account,
name,
data="1" * 128,
obj_name=name,
append=True)
self.assertEqual(size, 128)
meta = self.api.object_get_properties(self.account, name, name)
self.assertEqual(meta.get('hash', "").lower(), checksum.lower())
def test_container_refresh(self):
account = random_str(32)
# container_refresh on unknown container
name = random_str(32)
self.assertRaises(exc.NoSuchContainer, self.api.container_refresh,
account, name)
self.api.container_create(account, name)
time.sleep(0.5) # ensure container event have been processed
# container_refresh on existing container
self.api.container_refresh(account, name)
time.sleep(0.5) # ensure container event have been processed
res = self.api.container_list(account, prefix=name)
name_container, nb_objects, nb_bytes, _ = res[0]
self.assertEqual(name_container, name)
self.assertEqual(nb_objects, 0)
self.assertEqual(nb_bytes, 0)
self.api.object_create(account, name, data="data", obj_name=name)
time.sleep(0.5) # ensure container event have been processed
# container_refresh on existing container with data
self.api.container_refresh(account, name)
time.sleep(0.5) # ensure container event have been processed
res = self.api.container_list(account, prefix=name)
name_container, nb_objects, nb_bytes, _ = res[0]
self.assertEqual(name_container, name)
self.assertEqual(nb_objects, 1)
self.assertEqual(nb_bytes, 4)
self.api.object_delete(account, name, name)
time.sleep(0.5) # ensure container event have been processed
self.api.container_delete(account, name)
time.sleep(0.5) # ensure container event have been processed
# container_refresh on deleted container
self.assertRaises(exc.NoSuchContainer, self.api.container_refresh,
account, name)
self.api.account_delete(account)
def test_container_refresh_user_not_found(self):
name = random_str(32)
self.api.account.container_update(name, name, {"mtime": time.time()})
self.api.container_refresh(name, name)
containers = self.api.container_list(name)
self.assertEqual(len(containers), 0)
self.api.account_delete(name)
def test_account_refresh(self):
# account_refresh on unknown account
account = random_str(32)
self.assertRaises(exc.NoSuchAccount, self.api.account_refresh, account)
# account_refresh on existing account
self.api.account_create(account)
self.api.account_refresh(account)
time.sleep(0.5) # ensure container event have been processed
res = self.api.account_show(account)
self.assertEqual(res["bytes"], 0)
self.assertEqual(res["objects"], 0)
self.assertEqual(res["containers"], 0)
name = random_str(32)
self.api.object_create(account, name, data="data", obj_name=name)
time.sleep(0.5) # ensure container event have been processed
self.api.account_refresh(account)
time.sleep(0.5) # ensure container event have been processed
res = self.api.account_show(account)
self.assertEqual(res["bytes"], 4)
self.assertEqual(res["objects"], 1)
self.assertEqual(res["containers"], 1)
self.api.object_delete(account, name, name)
time.sleep(0.5) # ensure container event have been processed
self.api.container_delete(account, name)
time.sleep(0.5) # ensure container event have been processed
self.api.account_delete(account)
# account_refresh on deleted account
self.assertRaises(exc.NoSuchAccount, self.api.account_refresh, account)
def test_all_accounts_refresh(self):
# clear accounts
accounts = self.api.account_list()
for account in accounts:
try:
self.api.account_flush(account)
self.api.account_delete(account)
except exc.NoSuchAccount: # account remove in the meantime
pass
# all_accounts_refresh with 0 account
self.api.all_accounts_refresh()
# all_accounts_refresh with 2 account
account1 = random_str(32)
self.api.account_create(account1)
account2 = random_str(32)
self.api.account_create(account2)
self.api.all_accounts_refresh()
res = self.api.account_show(account1)
self.assertEqual(res["bytes"], 0)
self.assertEqual(res["objects"], 0)
self.assertEqual(res["containers"], 0)
res = self.api.account_show(account2)
self.assertEqual(res["bytes"], 0)
self.assertEqual(res["objects"], 0)
self.assertEqual(res["containers"], 0)
self.api.account_delete(account1)
self.api.account_delete(account2)
def test_account_flush(self):
# account_flush on unknown account
account = random_str(32)
self.assertRaises(exc.NoSuchAccount, self.api.account_flush, account)
# account_flush on existing account
name1 = random_str(32)
self.api.container_create(account, name1)
name2 = random_str(32)
self.api.container_create(account, name2)
time.sleep(0.5) # ensure container event have been processed
self.api.account_flush(account)
containers = self.api.container_list(account)
self.assertEqual(len(containers), 0)
res = self.api.account_show(account)
self.assertEqual(res["bytes"], 0)
self.assertEqual(res["objects"], 0)
self.assertEqual(res["containers"], 0)
self.api.container_delete(account, name1)
self.api.container_delete(account, name2)
time.sleep(0.5) # ensure container event have been processed
self.api.account_delete(account)
# account_flush on deleted account
self.assertRaises(exc.NoSuchAccount, self.api.account_flush, account)
def test_object_create_then_truncate(self):
"""Create an object then truncate data"""
name = random_str(16)
self.api.object_create(self.account,
name,
data="1" * 128,
obj_name=name)
self.api.object_truncate(self.account, name, name, size=64)
_, data = self.api.object_fetch(self.account, name, name)
data = "".join(data)
self.assertEqual(len(data), 64)
self.assertEqual(data, "1" * 64)
def test_object_create_append_then_truncate(self):
"""Create an object, append data then truncate on chunk boundary"""
name = random_str(16)
self.api.object_create(self.account,
name,
data="1" * 128,
obj_name=name)
_, size, _ = self.api.object_create(self.account,
name,
data="2" * 128,
obj_name=name,
append=True)
self.assertEqual(size, 128)
self.api.object_truncate(self.account, name, name, size=128)
_, data = self.api.object_fetch(self.account, name, name)
data = "".join(data)
self.assertEqual(len(data), 128)
self.assertEqual(data, "1" * 128)
self.api.object_truncate(self.account, name, name, size=128)
def test_object_create_then_invalid_truncate(self):
"""Create an object, append data then try to truncate outside object
range"""
name = random_str(16)
self.api.object_create(self.account,
name,
data="1" * 128,
obj_name=name)
self.assertRaises(exc.OioException,
self.api.object_truncate,
self.account,
name,
name,
size=-1)
self.assertRaises(exc.OioException,
self.api.object_truncate,
self.account,
name,
name,
size=129)
def test_container_snapshot(self):
name = random_str(16)
self.api.container_create(self.account, name)
test_object = "test_object"
self.api.object_create(self.account,
name,
data="0" * 128,
obj_name=test_object)
# Snapshot cannot have same name and same account
self.assertRaises(exc.ClientException, self.api.container_snapshot,
self.account, name, self.account, name)
snapshot_name = random_str(16)
self.assertNotEqual(snapshot_name, name)
# Non existing snapshot should work
self.api.container_snapshot(self.account, name, self.account,
snapshot_name)
# Already taken snapshot name should failed
self.assertRaises(exc.ClientException, self.api.container_snapshot,
self.account, name, self.account, snapshot_name)
# Check Container Frozen so create should failed
self.assertRaises(exc.ServiceBusy,
self.api.object_create,
self.account,
snapshot_name,
data="1" * 128,
obj_name="should_not_be_created")
# fullpath is set on every chunk
chunk_list = self.api.object_locate(self.account, name, test_object)[1]
# check that every chunk is different from the target
snapshot_list = self.api.object_locate(self.account, snapshot_name,
test_object)[1]
for c, t in zip(chunk_list, snapshot_list):
self.assertNotEqual(c['url'], t['url'])
# check target can be used
self.api.object_create(self.account,
name,
data="0" * 128,
obj_name="should_be_created")
# Create and send copy of a object
url_list = [c['url'] for c in chunk_list]
copy_list = self.api._generate_copy(url_list)
# every chunks should have the fullpath
fullpath = self.api._generate_fullpath(self.account, snapshot_name,
'copy', 12456)
self.api._send_copy(url_list, copy_list, fullpath[0])
# check that every copy exists
pool_manager = get_pool_manager()
for c in copy_list:
r = pool_manager.request('HEAD', c)
self.assertEqual(r.status, 200)
self.assertIn(fullpath[0],
r.headers["X-oio-chunk-meta-full-path"].split(','))
# Snapshot on non existing container should failed
self.assertRaises(exc.NoSuchContainer, self.api.container_snapshot,
random_str(16), random_str(16), random_str(16),
random_str(16))
# Snapshot need to have a account
self.assertRaises(exc.ClientException, self.api.container_snapshot,
self.account, name, None, random_str(16))
# Snapshot need to have a name
self.assertRaises(exc.ClientException, self.api.container_snapshot,
self.account, name, random_str(16), None)
