class ContentFactory(object): DEFAULT_DATASEC = "plain", {"nb_copy": "1", "distance": "0"} def __init__(self, conf): self.conf = conf self.logger = get_logger(conf) self.container_client = ContainerClient(conf) def get(self, container_id, content_id): try: meta, chunks = self.container_client.content_show( cid=container_id, content=content_id) except NotFound: raise ContentNotFound("Content %s/%s not found" % (container_id, content_id)) chunk_method = meta['chunk-method'] storage_method = STORAGE_METHODS.load(chunk_method) cls = ECContent if storage_method.ec else PlainContent return cls(self.conf, container_id, meta, chunks, storage_method) def new(self, container_id, path, size, policy): meta, chunks = self.container_client.content_prepare(cid=container_id, path=path, size=size, stgpol=policy) chunk_method = meta['chunk-method'] storage_method = STORAGE_METHODS.load(chunk_method) cls = ECContent if storage_method.ec else PlainContent return cls(self.conf, container_id, meta, chunks, storage_method) def change_policy(self, container_id, content_id, new_policy): old_content = self.get(container_id, content_id) if old_content.stgpol == new_policy: return old_content new_content = self.new(container_id, old_content.path, old_content.length, new_policy) stream = old_content.fetch() new_content.create(GeneratorIO(stream)) # the old content is automatically deleted because the new content has # the same name (but not the same id) return new_content
class ContentFactory(object): DEFAULT_DATASEC = "plain", {"nb_copy": "1", "distance": "0"} def __init__(self, conf): self.conf = conf self.logger = get_logger(conf) self.container_client = ContainerClient(conf) def get(self, container_id, content_id): try: meta, chunks = self.container_client.content_show( cid=container_id, content=content_id) except NotFound: raise ContentNotFound("Content %s/%s not found" % (container_id, content_id)) chunk_method = meta['chunk_method'] storage_method = STORAGE_METHODS.load(chunk_method) cls = ECContent if storage_method.ec else PlainContent return cls(self.conf, container_id, meta, chunks, storage_method) def new(self, container_id, path, size, policy): meta, chunks = self.container_client.content_prepare( cid=container_id, path=path, size=size, stgpol=policy) chunk_method = meta['chunk_method'] storage_method = STORAGE_METHODS.load(chunk_method) cls = ECContent if storage_method.ec else PlainContent return cls(self.conf, container_id, meta, chunks, storage_method) def change_policy(self, container_id, content_id, new_policy): old_content = self.get(container_id, content_id) if old_content.stgpol == new_policy: return old_content new_content = self.new(container_id, old_content.path, old_content.length, new_policy) stream = old_content.fetch() new_content.create(GeneratorIO(stream)) # the old content is automatically deleted because the new content has # the same name (but not the same id) return new_content
class TestRebuilderCrawler(BaseTestCase): def setUp(self): super(TestRebuilderCrawler, self).setUp() self.namespace = self.conf['namespace'] self.account = self.conf['account'] self.gridconf = {"namespace": self.namespace} self.container_client = ContainerClient(self.gridconf) self.blob_client = BlobClient() self.container_name = "TestRebuilderCrawler%d" % int(time.time()) self.container_client.container_create(acct=self.account, ref=self.container_name) def _push_content(self, content): for c in content.chunks: self.blob_client.chunk_put(c.url, c.get_create_xattr(), c.data) self.container_client.content_create(acct=content.account, ref=content.container_name, path=content.content_name, size=content.size, checksum=content.hash, content_id=content.content_id, stgpol=content.stgpol, data=content.get_create_meta2()) def tearDown(self): super(TestRebuilderCrawler, self).tearDown() def test_rebuild_chunk(self): # push a new content content = TestContent(self.conf, self.account, self.container_name, "mycontent", "TWOCOPIES") data = "azerty" content.add_chunk(data, pos='0', rawx=0) content.add_chunk(data, pos='0', rawx=1) self._push_content(content) # rebuild the first rawx rebuilder = BlobRebuilderWorker(self.gridconf, None, self.conf['rawx'][0]['addr']) rebuilder.chunk_rebuild(content.container_id, content.content_id, content.chunks[0].id) # check meta2 information _, res = self.container_client.content_show(acct=content.account, ref=content.container_name, content=content.content_id) new_chunk_info = None for c in res: if (c['url'] != content.chunks[0].url and c['url'] != content.chunks[1].url): new_chunk_info = c new_chunk_id = new_chunk_info['url'].split('/')[-1] self.assertEqual(new_chunk_info['hash'], content.chunks[0].hash) self.assertEqual(new_chunk_info['pos'], content.chunks[0].pos) self.assertEqual(new_chunk_info['size'], content.chunks[0].size) # check chunk information meta, stream = self.blob_client.chunk_get(new_chunk_info['url']) self.assertEqual(meta['content_size'], str(content.chunks[0].size)) self.assertEqual(meta['content_path'], content.content_name) self.assertEqual(meta['content_cid'], content.container_id) self.assertEqual(meta['content_id'], content.content_id) self.assertEqual(meta['chunk_id'], new_chunk_id) self.assertEqual(meta['chunk_pos'], content.chunks[0].pos) self.assertEqual(meta['content_version'], content.version) self.assertEqual(meta['chunk_hash'], content.chunks[0].hash) self.assertEqual(stream.next(), content.chunks[0].data) # check rtime flag in rdir rdir_client = RdirClient(self.gridconf) res = rdir_client.chunk_fetch(self.conf['rawx'][0]['addr']) key = (content.container_id, content.content_id, content.chunks[0].id) for i_container, i_content, i_chunk, i_value in res: if (i_container, i_content, i_chunk) == key: check_value = i_value self.assertIsNotNone(check_value.get('rtime')) @unittest.skipIf(len(get_config()['rawx']) != 3, "The number of rawx must be 3") def test_rebuild_no_spare(self): # push a new content content = TestContent(self.conf, self.account, self.container_name, "mycontent", "THREECOPIES") data = "azerty" content.add_chunk(data, pos='0', rawx=0) content.add_chunk(data, pos='0', rawx=1) content.add_chunk(data, pos='0', rawx=2) self._push_content(content) # rebuild the first rawx rebuilder = BlobRebuilderWorker(self.gridconf, None, self.conf['rawx'][0]['addr']) self.assertRaises(SpareChunkException, rebuilder.chunk_rebuild, content.container_id, content.content_id, content.chunks[0].id) def test_rebuild_upload_failed(self): # push a new content content = TestContent(self.conf, self.account, self.container_name, "mycontent", "TWOCOPIES") data = "azerty" content.add_chunk(data, pos='0', rawx=0) content.add_chunk(data, pos='0', rawx=1) self._push_content(content) # rebuild the first rawx rebuilder = BlobRebuilderWorker(self.gridconf, None, self.conf['rawx'][0]['addr']) # Force upload to raise an exception with patch('oio.content.content.BlobClient') as MockClass: instance = MockClass.return_value instance.chunk_copy.side_effect = Exception("xx") self.assertRaises(UnrecoverableContent, rebuilder.chunk_rebuild, content.container_id, content.content_id, content.chunks[0].id) def test_rebuild_nonexistent_chunk(self): rebuilder = BlobRebuilderWorker(self.gridconf, None, self.conf['rawx'][0]['addr']) # try to rebuild an nonexistant chunk self.assertRaises(OrphanChunk, rebuilder.chunk_rebuild, 64 * '0', 32 * '0', 64 * '0') def test_rebuild_orphan_chunk(self): # push a new content content = TestContent(self.conf, self.account, self.container_name, "mycontent", "TWOCOPIES") data = "azerty" content.add_chunk(data, pos='0', rawx=0) content.add_chunk(data, pos='0', rawx=1) self._push_content(content) # rebuild the first rawx rebuilder = BlobRebuilderWorker(self.gridconf, None, self.conf['rawx'][0]['addr']) # try to rebuild an nonexistant chunk self.assertRaises(OrphanChunk, rebuilder.chunk_rebuild, content.container_id, content.content_id, 64 * '0') def test_rebuild_with_no_copy(self): # push a new content content = TestContent(self.conf, self.account, self.container_name, "mycontent", "SINGLE") data = "azerty" content.add_chunk(data, pos='0', rawx=0) self._push_content(content) # rebuild the first rawx rebuilder = BlobRebuilderWorker(self.gridconf, None, self.conf['rawx'][0]['addr']) # try to rebuild chunk without copy self.assertRaises(UnrecoverableContent, rebuilder.chunk_rebuild, content.container_id, content.content_id, content.chunks[0].id)
class BlobAuditorWorker(object): def __init__(self, conf, logger, volume): self.conf = conf self.logger = logger self.volume = volume self.run_time = 0 self.passes = 0 self.errors = 0 self.orphan_chunks = 0 self.faulty_chunks = 0 self.corrupted_chunks = 0 self.last_reported = 0 self.chunks_run_time = 0 self.bytes_running_time = 0 self.bytes_processed = 0 self.total_bytes_processed = 0 self.total_chunks_processed = 0 self.report_interval = int_value(conf.get('report_interval'), 3600) self.max_chunks_per_second = int_value(conf.get('chunks_per_second'), 30) self.max_bytes_per_second = int_value(conf.get('bytes_per_second'), 10000000) self.container_client = ContainerClient(conf) def audit_pass(self): self.namespace, self.address = check_volume(self.volume) start_time = report_time = time.time() total_errors = 0 total_corrupted = 0 total_orphans = 0 total_faulty = 0 audit_time = 0 paths = paths_gen(self.volume) for path in paths: loop_time = time.time() self.safe_chunk_audit(path) self.chunks_run_time = ratelimit(self.chunks_run_time, self.max_chunks_per_second) self.total_chunks_processed += 1 now = time.time() if now - self.last_reported >= self.report_interval: self.logger.info( '%(start_time)s ' '%(passes)d ' '%(corrupted)d ' '%(faulty)d ' '%(orphans)d ' '%(errors)d ' '%(c_rate).2f ' '%(b_rate).2f ' '%(total).2f ' '%(audit_time).2f' '%(audit_rate).2f' % { 'start_time': time.ctime(report_time), 'passes': self.passes, 'corrupted': self.corrupted_chunks, 'faulty': self.faulty_chunks, 'orphans': self.orphan_chunks, 'errors': self.errors, 'c_rate': self.passes / (now - report_time), 'b_rate': self.bytes_processed / (now - report_time), 'total': (now - start_time), 'audit_time': audit_time, 'audit_rate': audit_time / (now - start_time) }) report_time = now total_corrupted += self.corrupted_chunks total_orphans += self.orphan_chunks total_faulty += self.faulty_chunks total_errors += self.errors self.passes = 0 self.corrupted_chunks = 0 self.orphan_chunks = 0 self.faulty_chunks = 0 self.errors = 0 self.bytes_processed = 0 self.last_reported = now audit_time += (now - loop_time) elapsed = (time.time() - start_time) or 0.000001 self.logger.info( '%(elapsed).02f ' '%(corrupted)d ' '%(faulty)d ' '%(orphans)d ' '%(errors)d ' '%(chunk_rate).2f ' '%(bytes_rate).2f ' '%(audit_time).2f ' '%(audit_rate).2f' % { 'elapsed': elapsed, 'corrupted': total_corrupted + self.corrupted_chunks, 'faulty': total_faulty + self.faulty_chunks, 'orphans': total_orphans + self.orphan_chunks, 'errors': total_errors + self.errors, 'chunk_rate': self.total_chunks_processed / elapsed, 'bytes_rate': self.total_bytes_processed / elapsed, 'audit_time': audit_time, 'audit_rate': audit_time / elapsed }) def safe_chunk_audit(self, path): try: self.chunk_audit(path) except exc.FaultyChunk as err: self.faulty_chunks += 1 self.logger.error('ERROR faulty chunk %s: %s', path, err) except exc.CorruptedChunk as err: self.corrupted_chunks += 1 self.logger.error('ERROR corrupted chunk %s: %s', path, err) except exc.OrphanChunk as err: self.orphan_chunks += 1 self.logger.error('ERROR orphan chunk %s: %s', path, err) except Exception: self.errors += 1 self.logger.exception('ERROR while auditing chunk %s', path) self.passes += 1 def chunk_audit(self, path): with open(path) as f: try: meta = read_chunk_metadata(f) except exc.MissingAttribute as e: raise exc.FaultyChunk('Missing extended attribute %s' % e) size = int(meta['chunk_size']) md5_checksum = meta['chunk_hash'].lower() reader = ChunkReader(f, size, md5_checksum) with closing(reader): for buf in reader: buf_len = len(buf) self.bytes_running_time = ratelimit( self.bytes_running_time, self.max_bytes_per_second, increment=buf_len) self.bytes_processed += buf_len self.total_bytes_processed += buf_len try: container_id = meta['container_id'] content_path = meta['content_path'] content_attr, data = self.container_client.content_show( cid=container_id, path=content_path) # Check chunk data chunk_data = None metachunks = set() for c in data: if c['url'].endswith(meta['chunk_id']): metachunks.add(c['pos'].split('.', 2)[0]) chunk_data = c if not chunk_data: raise exc.OrphanChunk('Not found in content') if chunk_data['size'] != int(meta['chunk_size']): raise exc.FaultyChunk('Invalid chunk size found') if chunk_data['hash'] != meta['chunk_hash']: raise exc.FaultyChunk('Invalid chunk hash found') if chunk_data['pos'] != meta['chunk_pos']: raise exc.FaultyChunk('Invalid chunk position found') except exc.NotFound: raise exc.OrphanChunk('Chunk not found in container')
class Checker(object): def __init__(self, namespace, concurrency=50, error_file=None): self.pool = GreenPool(concurrency) self.error_file = error_file if self.error_file: f = open(self.error_file, 'a') self.error_writer = csv.writer(f, delimiter=' ') conf = {'namespace': namespace} self.account_client = AccountClient(conf) self.container_client = ContainerClient(conf) self.blob_client = BlobClient() 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 = {} def write_error(self, target): error = [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 check_chunk(self, target): chunk = target.chunk obj_listing = self.check_obj(target) error = False if chunk not in obj_listing: print(' Chunk %s missing in object listing' % target) error = True # checksum = None else: # TODO check checksum match # checksum = obj_listing[chunk]['hash'] pass try: self.blob_client.chunk_head(chunk) except exc.NotFound as e: self.chunk_not_found += 1 error = True print(' Not found chunk "%s": %s' % (target, str(e))) except Exception as e: self.chunk_exceptions += 1 error = True print(' Exception chunk "%s": %s' % (target, str(e))) if error and self.error_file: self.write_error(target) self.chunks_checked += 1 def check_obj(self, target, recurse=False): account = target.account container = target.container obj = target.obj if (account, container, obj) in self.running: self.running[(account, container, obj)].wait() if (account, container, obj) in self.list_cache: return self.list_cache[(account, container, obj)] self.running[(account, container, obj)] = Event() print('Checking object "%s"' % target) container_listing = self.check_container(target) error = False if obj not in container_listing: print(' Object %s missing in container listing' % target) error = True # checksum = None else: # TODO check checksum match # checksum = container_listing[obj]['hash'] pass results = [] try: _, resp = self.container_client.content_show(acct=account, ref=container, path=obj) except exc.NotFound as e: self.object_not_found += 1 error = True print(' Not found object "%s": %s' % (target, str(e))) except Exception as e: self.object_exceptions += 1 error = True print(' Exception object "%s": %s' % (target, str(e))) else: results = resp chunk_listing = dict() for chunk in results: chunk_listing[chunk['url']] = chunk self.objects_checked += 1 self.list_cache[(account, container, obj)] = chunk_listing self.running[(account, container, obj)].send(True) del self.running[(account, container, obj)] if recurse: for chunk in chunk_listing: t = target.copy() t.chunk = chunk self.pool.spawn_n(self.check_chunk, t) if error and self.error_file: self.write_error(target) return chunk_listing def check_container(self, target, recurse=False): account = target.account container = target.container if (account, container) in self.running: self.running[(account, container)].wait() if (account, container) in self.list_cache: return self.list_cache[(account, container)] self.running[(account, container)] = Event() print('Checking container "%s"' % target) account_listing = self.check_account(target) error = False if container not in account_listing: error = True print(' Container %s missing in account listing' % target) marker = None results = [] while True: try: resp = self.container_client.container_list(acct=account, ref=container, marker=marker) except exc.NotFound as e: self.container_not_found += 1 error = True print(' Not found container "%s": %s' % (target, str(e))) break except Exception as e: self.container_exceptions += 1 error = True print(' Exception container "%s": %s' % (target, str(e))) break if resp['objects']: marker = resp['objects'][-1]['name'] else: break results.extend(resp['objects']) container_listing = dict() for obj in results: container_listing[obj['name']] = obj self.containers_checked += 1 self.list_cache[(account, container)] = container_listing self.running[(account, container)].send(True) del self.running[(account, container)] if recurse: for obj in container_listing: t = target.copy() t.obj = obj self.pool.spawn_n(self.check_obj, t, True) if error and self.error_file: self.write_error(target) return container_listing def check_account(self, target, recurse=False): account = target.account if account in self.running: self.running[account].wait() if account in self.list_cache: return self.list_cache[account] self.running[account] = Event() print('Checking account "%s"' % target) error = False marker = None results = [] while True: try: resp = self.account_client.containers_list(account, marker=marker) except Exception as e: self.account_exceptions += 1 error = True print(' Exception account "%s": %s' % (target, str(e))) break if resp['listing']: marker = resp['listing'][-1][0] else: break results.extend(resp['listing']) containers = dict() for e in results: containers[e[0]] = (e[1], e[2]) self.list_cache[account] = containers self.running[account].send(True) del self.running[account] self.accounts_checked += 1 if recurse: for container in containers: t = target.copy() t.container = container self.pool.spawn_n(self.check_container, t, True) if error and self.error_file: self.write_error(target) return containers def check(self, target): if target.chunk and target.obj and target.container: self.pool.spawn_n(self.check_chunk, target) elif target.obj and target.container: self.pool.spawn_n(self.check_obj, target, True) elif target.container: self.pool.spawn_n(self.check_container, target, True) else: self.pool.spawn_n(self.check_account, target, True) def wait(self): self.pool.waitall() def report(self): 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: _report_stat("Missing accounts", self.account_not_found) if self.account_exceptions: _report_stat("Exceptions", self.account_not_found) print() _report_stat("Containers checked", self.containers_checked) if self.container_not_found: _report_stat("Missing containers", self.container_not_found) if self.container_exceptions: _report_stat("Exceptions", self.container_exceptions) print() _report_stat("Objects checked", self.objects_checked) if self.object_not_found: _report_stat("Missing objects", self.object_not_found) if self.object_exceptions: _report_stat("Exceptions", self.object_exceptions) print() _report_stat("Chunks checked", self.chunks_checked) if self.chunk_not_found: _report_stat("Missing chunks", self.chunk_not_found) if self.chunk_exceptions: _report_stat("Exceptions", self.chunk_exceptions)
class TestECContent(BaseTestCase): def setUp(self): super(TestECContent, self).setUp() if len(self.conf['services']['rawx']) < 12: self.skipTest("Not enough rawx. " "EC tests needs at least 12 rawx to run") self.namespace = self.conf['namespace'] self.account = self.conf['account'] self.chunk_size = self.conf['chunk_size'] self.gridconf = {"namespace": self.namespace} self.content_factory = ContentFactory(self.gridconf) self.container_client = ContainerClient(self.gridconf) self.blob_client = BlobClient() self.container_name = "TestECContent%f" % time.time() self.container_client.container_create(acct=self.account, ref=self.container_name) self.container_id = cid_from_name(self.account, self.container_name).upper() self.content = random_str(64) self.stgpol = "EC" self.size = 1024*1024 + 320 self.k = 6 self.m = 3 def tearDown(self): super(TestECContent, self).tearDown() def random_chunks(self, nb): l = random.sample(xrange(self.k + self.m), nb) return ["0.%s" % i for i in l] def _test_create(self, data_size): # generate random test data data = random_data(data_size) # using factory create new EC content content = self.content_factory.new( self.container_id, self.content, len(data), self.stgpol) # verify the factory gave us an ECContent self.assertEqual(type(content), ECContent) # perform the content creation content.create(StringIO(data)) meta, chunks = self.container_client.content_show( cid=self.container_id, content=content.content_id) # verify metadata chunks = ChunksHelper(chunks) self.assertEqual(meta['hash'], md5_data(data)) self.assertEqual(meta['length'], str(len(data))) self.assertEqual(meta['policy'], self.stgpol) self.assertEqual(meta['name'], self.content) metachunk_nb = int(math.ceil(float(len(data)) / self.chunk_size)) \ if len(data) != 0 else 1 # verify each metachunk for metapos in range(metachunk_nb): chunks_at_pos = content.chunks.filter(metapos=metapos) for chunk in chunks_at_pos: meta, stream = self.blob_client.chunk_get(chunk.url) self.assertEqual(meta['metachunk_size'], str(chunk.size)) self.assertEqual(meta['metachunk_hash'], chunk.checksum) self.assertEqual(meta['content_path'], self.content) self.assertEqual(meta['container_id'], self.container_id) self.assertEqual(meta['content_id'], meta['content_id']) self.assertEqual(meta['chunk_id'], chunk.id) self.assertEqual(meta['chunk_pos'], chunk.pos) self.assertEqual(meta['chunk_hash'], md5_stream(stream)) def test_create_0_byte(self): self._test_create(0) def test_create_1_byte(self): self._test_create(1) def test_create(self): self._test_create(DAT_LEGIT_SIZE) def _test_rebuild(self, data_size, broken_pos_list): # generate test data data = os.urandom(data_size) # create initial content old_content = self.content_factory.new( self.container_id, self.content, len(data), self.stgpol) # verify factory work as intended self.assertEqual(type(old_content), ECContent) # perform initial content creation old_content.create(StringIO(data)) uploaded_content = self.content_factory.get(self.container_id, old_content.content_id) # break the content old_info = {} for pos in broken_pos_list: old_info[pos] = {} c = uploaded_content.chunks.filter(pos=pos)[0] old_info[pos]["url"] = c.url old_info[pos]["id"] = c.id old_info[pos]["hash"] = c.checksum chunk_id_to_rebuild = c.id meta, stream = self.blob_client.chunk_get(c.url) old_info[pos]["dl_meta"] = meta old_info[pos]["dl_hash"] = md5_stream(stream) # delete the chunk self.blob_client.chunk_delete(c.url) # rebuild the broken chunks uploaded_content.rebuild_chunk(chunk_id_to_rebuild) rebuilt_content = self.content_factory.get(self.container_id, uploaded_content.content_id) # sanity check self.assertEqual(type(rebuilt_content), ECContent) # verify rebuild result for pos in broken_pos_list: c = rebuilt_content.chunks.filter(pos=pos)[0] rebuilt_meta, rebuilt_stream = self.blob_client.chunk_get(c.url) self.assertEqual(rebuilt_meta["chunk_id"], c.id) self.assertEqual(md5_stream(rebuilt_stream), old_info[pos]["dl_hash"]) self.assertEqual(c.checksum, old_info[pos]["hash"]) self.assertNotEqual(c.url, old_info[pos]["url"]) del old_info[pos]["dl_meta"]["chunk_id"] del rebuilt_meta["chunk_id"] self.assertEqual(rebuilt_meta, old_info[pos]["dl_meta"]) def test_content_0_byte_rebuild(self): self._test_rebuild(0, self.random_chunks(1)) def test_content_0_byte_rebuild_advanced(self): self._test_rebuild(0, self.random_chunks(3)) def test_content_1_byte_rebuild(self): self._test_rebuild(1, self.random_chunks(1)) def test_content_1_byte_rebuild_advanced(self): self._test_rebuild(1, self.random_chunks(3)) def test_content_rebuild(self): self._test_rebuild(DAT_LEGIT_SIZE, self.random_chunks(1)) def test_content_rebuild_advanced(self): self._test_rebuild(DAT_LEGIT_SIZE, self.random_chunks(3)) def test_content_rebuild_unrecoverable(self): self.assertRaises( UnrecoverableContent, self._test_rebuild, DAT_LEGIT_SIZE, self.random_chunks(4)) def _new_content(self, data, broken_pos_list=[]): old_content = self.content_factory.new( self.container_id, self.content, len(data), self.stgpol) self.assertEqual(type(old_content), ECContent) old_content.create(StringIO(data)) # break content for pos in broken_pos_list: c = old_content.chunks.filter(pos=pos)[0] self.blob_client.chunk_delete(c.url) # get the new structure of the uploaded content return self.content_factory.get(self.container_id, old_content.content_id) def test_orphan_chunk(self): content = self._new_content(random_data(10)) self.assertRaises(OrphanChunk, content.rebuild_chunk, "invalid") def _test_fetch(self, data_size, broken_pos_list=None): broken_pos_list = broken_pos_list or [] test_data = random_data(data_size) content = self._new_content(test_data, broken_pos_list) data = "".join(content.fetch()) self.assertEqual(len(data), len(test_data)) self.assertEqual(md5_data(data), md5_data(test_data)) # verify that chunks are broken for pos in broken_pos_list: chunk = content.chunks.filter(pos=pos)[0] self.assertRaises( NotFound, self.blob_client.chunk_delete, chunk.url) def test_fetch_content_0_byte(self): self._test_fetch(0) def test_fetch_content_1_byte(self): self._test_fetch(1) def test_fetch_content(self): self._test_fetch(DAT_LEGIT_SIZE) def test_fetch_content_0_byte_broken(self): self._test_fetch(0, self.random_chunks(3)) def test_fetch_content_1_byte_broken(self): self._test_fetch(1, self.random_chunks(3)) def test_fetch_content_broken(self): self._test_fetch(DAT_LEGIT_SIZE, self.random_chunks(3)) def test_fetch_content_unrecoverable(self): broken_chunks = self.random_chunks(4) self.assertRaises( OioException, self._test_fetch, DAT_LEGIT_SIZE, broken_chunks)
class ObjectStorageApi(object): """ The Object Storage API. High level API that wraps `AccountClient`, `ContainerClient` and `DirectoryClient` classes. Every method that takes a `kwargs` argument accepts the at least the following keywords: - `headers`: `dict` of extra headers to pass to the proxy - `connection_timeout`: `float` - `read_timeout`: `float` - `write_timeout`: `float` """ TIMEOUT_KEYS = ('connection_timeout', 'read_timeout', 'write_timeout') def __init__(self, namespace, logger=None, **kwargs): """ Initialize the object storage API. :param namespace: name of the namespace to interract with :type namespace: `str` :keyword connection_timeout: connection timeout towards rawx services :type connection_timeout: `float` seconds :keyword read_timeout: timeout for rawx responses and data reads from the caller (when uploading) :type read_timeout: `float` seconds :keyword write_timeout: timeout for rawx write requests :type write_timeout: `float` seconds :keyword pool_manager: a pooled connection manager that will be used for all HTTP based APIs (except rawx) :type pool_manager: `urllib3.PoolManager` """ self.namespace = namespace conf = {"namespace": self.namespace} self.logger = logger or get_logger(conf) self.timeouts = {tok: float_value(tov, None) for tok, tov in kwargs.items() if tok in self.__class__.TIMEOUT_KEYS} from oio.account.client import AccountClient from oio.container.client import ContainerClient from oio.directory.client import DirectoryClient self.directory = DirectoryClient(conf, logger=self.logger, **kwargs) self.container = ContainerClient(conf, logger=self.logger, **kwargs) # In AccountClient, "endpoint" is the account service, not the proxy acct_kwargs = kwargs.copy() acct_kwargs["proxy_endpoint"] = acct_kwargs.pop("endpoint", None) self.account = AccountClient(conf, logger=self.logger, **acct_kwargs) def _patch_timeouts(self, kwargs): """ Insert timeout settings from this class's constructor into `kwargs`, if they are not already there. """ for tok, tov in self.timeouts.items(): if tok not in kwargs: kwargs[tok] = tov def account_create(self, account, **kwargs): """ Create an account. :param account: name of the account to create :type account: `str` :returns: `True` if the account has been created """ return self.account.account_create(account, **kwargs) @handle_account_not_found def account_delete(self, account, **kwargs): """ Delete an account. :param account: name of the account to delete :type account: `str` """ self.account.account_delete(account, **kwargs) @handle_account_not_found def account_show(self, account, **kwargs): """ Get information about an account. """ return self.account.account_show(account, **kwargs) def account_list(self, **kwargs): """ List known accounts. Notice that account creation is asynchronous, and an autocreated account may appear in the listing only after several seconds. """ return self.account.account_list(**kwargs) @handle_account_not_found def account_update(self, account, metadata, to_delete=None, **kwargs): warnings.warn("You'd better use account_set_properties()", DeprecationWarning, stacklevel=2) self.account.account_update(account, metadata, to_delete, **kwargs) @handle_account_not_found def account_set_properties(self, account, properties, **kwargs): self.account.account_update(account, properties, None, **kwargs) @handle_account_not_found def account_del_properties(self, account, properties, **kwargs): self.account.account_update(account, None, properties, **kwargs) def container_create(self, account, container, properties=None, **kwargs): """ Create a container. :param account: account in which to create the container :type account: `str` :param container: name of the container :type container: `str` :param properties: properties to set on the container :type properties: `dict` :returns: True if the container has been created, False if it already exists """ return self.container.container_create(account, container, properties=properties, **kwargs) @handle_container_not_found @ensure_headers @ensure_request_id def container_touch(self, account, container, **kwargs): """ Trigger a notification about the container state. :param account: account from which to delete the container :type account: `str` :param container: name of the container :type container: `str` """ self.container.container_touch(account, container, **kwargs) def container_create_many(self, account, containers, properties=None, **kwargs): """ Create Many containers :param account: account in which to create the containers :type account: `str` :param containers: names of the containers :type containers: `list` :param properties: properties to set on the containers :type properties: `dict` """ return self.container.container_create_many(account, containers, properties=properties, **kwargs) @handle_container_not_found def container_delete(self, account, container, **kwargs): """ Delete a container. :param account: account from which to delete the container :type account: `str` :param container: name of the container :type container: `str` """ self.container.container_delete(account, container, **kwargs) @handle_account_not_found def container_list(self, account, limit=None, marker=None, end_marker=None, prefix=None, delimiter=None, **kwargs): """ Get the list of containers of an account. :param account: account from which to get the container list :type account: `str` :keyword limit: maximum number of results to return :type limit: `int` :keyword marker: name of the container from where to start the listing :type marker: `str` :keyword end_marker: :keyword prefix: :keyword delimiter: :return: the list of containers of an account :rtype: `list` of items (`list`) with 4 fields: name, number of objects, number of bytes, and 1 if the item is a prefix or 0 if the item is actually a container """ resp = self.account.container_list(account, limit=limit, marker=marker, end_marker=end_marker, prefix=prefix, delimiter=delimiter, **kwargs) return resp["listing"] @handle_container_not_found def container_show(self, account, container, **kwargs): """ Get information about a container (user properties). :param account: account in which the container is :type account: `str` :param container: name of the container :type container: `str` :returns: a `dict` with "properties" containing a `dict` of user properties. """ return self.container.container_show(account, container, **kwargs) @handle_container_not_found def container_snapshot(self, account, container, dst_account, dst_container, batch=100, **kwargs): """ Create a copy of the container (only the content of the database) :param account: account in which the target is :type account: `str` :param container: name of the target :type container: `str` :param dst_account: account in which the snapshot will be. :type dst_account: `str` :param dst_container: name of the snapshot :type dst_container: `str` """ try: self.container.container_freeze(account, container) self.container.container_snapshot( account, container, dst_account, dst_container) resp = self.object_list(dst_account, dst_container) obj_gen = resp['objects'] target_beans = [] copy_beans = [] for obj in obj_gen: data = self.object_locate( account, container, obj["name"]) chunks = [chunk['url'] for chunk in data[1]] copies = self._generate_copy(chunks) fullpath = self._generate_fullpath( dst_account, dst_container, obj['name'], obj['version']) self._send_copy(chunks, copies, fullpath[0]) t_beans, c_beans = self._prepare_update_meta2( data[1], copies, dst_account, dst_container, obj['content']) target_beans.extend(t_beans) copy_beans.extend(c_beans) if len(target_beans) > batch: self.container.container_raw_update( target_beans, copy_beans, dst_account, dst_container, frozen=True) target_beans = [] copy_beans = [] if target_beans: self.container.container_raw_update( target_beans, copy_beans, dst_account, dst_container, frozen=True) finally: self.container.container_enable(account, container) @handle_container_not_found def container_get_properties(self, account, container, properties=None, **kwargs): """ Get information about a container (user and system properties). :param account: account in which the container is :type account: `str` :param container: name of the container :type container: `str` :param properties: *ignored* :returns: a `dict` with "properties" and "system" entries, containing respectively a `dict` of user properties and a `dict` of system properties. """ return self.container.container_get_properties(account, container, properties=properties, **kwargs) @handle_container_not_found def container_set_properties(self, account, container, properties=None, clear=False, **kwargs): """ Set properties on a container. :param account: name of the account :type account: `str` :param container: name of the container where to set properties :type container: `str` :param properties: a dictionary of properties :type properties: `dict` :param clear: :type clear: `bool` :keyword system: dictionary of system properties to set """ return self.container.container_set_properties( account, container, properties, clear=clear, **kwargs) @handle_container_not_found def container_del_properties(self, account, container, properties, **kwargs): """ Delete properties of a container. :param account: name of the account :type account: `str` :param container: name of the container to deal with :type container: `str` :param properties: a list of property keys :type properties: `list` """ return self.container.container_del_properties( account, container, properties, **kwargs) def container_update(self, account, container, metadata, clear=False, **kwargs): warnings.warn("You'd better use container_set_properties()", DeprecationWarning) if not metadata: self.container_del_properties( account, container, [], **kwargs) else: self.container_set_properties( account, container, metadata, clear, **kwargs) @handle_container_not_found @ensure_headers @ensure_request_id def object_create(self, account, container, file_or_path=None, data=None, etag=None, obj_name=None, mime_type=None, metadata=None, policy=None, key_file=None, append=False, properties=None, **kwargs): """ Create an object or append data to object in *container* of *account* with data taken from either *data* (`str` or `generator`) or *file_or_path* (path to a file or file-like object). The object will be named after *obj_name* if specified, or after the base name of *file_or_path*. :param account: name of the account where to create the object :type account: `str` :param container: name of the container where to create the object :type container: `str` :param file_or_path: file-like object or path to a file from which to read object data :type file_or_path: `str` or file-like object :param data: object data (if `file_or_path` is not set) :type data: `str` or `generator` :keyword etag: entity tag of the object :type etag: `str` :keyword obj_name: name of the object to create. If not set, will use the base name of `file_or_path`. :keyword mime_type: MIME type of the object :type mime_type: `str` :keyword properties: a dictionary of properties :type properties: `dict` :keyword policy: name of the storage policy :type policy: `str` :keyword key_file: :param append: if set, data will be append to existing object (or object will be created if unset) :type append: `bool` :returns: `list` of chunks, size and hash of the what has been uploaded """ if (data, file_or_path) == (None, None): raise exc.MissingData() src = data if data is not None else file_or_path if src is file_or_path: if isinstance(file_or_path, basestring): if not os.path.exists(file_or_path): raise exc.FileNotFound("File '%s' not found." % file_or_path) file_name = os.path.basename(file_or_path) else: try: file_name = os.path.basename(file_or_path.name) except AttributeError: file_name = None obj_name = obj_name or file_name elif isgenerator(src): file_or_path = GeneratorIO(src) src = file_or_path if not obj_name: raise exc.MissingName( "No name for the object has been specified" ) sysmeta = {'mime_type': mime_type, 'etag': etag} if metadata: warnings.warn( "You'd better use 'properties' instead of 'metadata'", DeprecationWarning, stacklevel=4) if not properties: properties = metadata else: properties.update(metadata) if src is data: return self._object_create( account, container, obj_name, BytesIO(data), sysmeta, properties=properties, policy=policy, key_file=key_file, append=append, **kwargs) elif hasattr(file_or_path, "read"): return self._object_create( account, container, obj_name, src, sysmeta, properties=properties, policy=policy, key_file=key_file, append=append, **kwargs) else: with open(file_or_path, "rb") as f: return self._object_create( account, container, obj_name, f, sysmeta, properties=properties, policy=policy, key_file=key_file, append=append, **kwargs) @ensure_headers @ensure_request_id def object_touch(self, account, container, obj, version=None, **kwargs): """ Trigger a notification about an object (as if it just had been created). :param account: name of the account where to create the object :type account: `str` :param container: name of the container where to create the object :type container: `str` :param obj: name of the object to touch """ self.container.content_touch(account, container, obj, version=version, **kwargs) def object_drain(self, account, container, obj, version=None, **kwargs): """ Remove all the chunks of a content, but keep all the metadata. :param account: name of the account where the object is present :type account: `str` :param container: name of the container where the object is present :type container: `str` :param obj: name of the object to drain """ self.container.content_drain(account, container, obj, version=version, **kwargs) @handle_object_not_found @ensure_headers @ensure_request_id def object_delete(self, account, container, obj, version=None, **kwargs): """ Delete an object from a container. If versioning is enabled and no version is specified, the object will be marked as deleted but not actually deleted. :param account: name of the account the object belongs to :type account: `str` :param container: name of the container the object belongs to :type container: `str` :param obj: name of the object to delete :param version: version of the object to delete :returns: True on success """ return self.container.content_delete(account, container, obj, version=version, **kwargs) @ensure_headers @ensure_request_id def object_delete_many(self, account, container, objs, **kwargs): return self.container.content_delete_many( account, container, objs, **kwargs) @handle_object_not_found @ensure_headers @ensure_request_id def object_truncate(self, account, container, obj, version=None, size=None, **kwargs): """ Truncate object at specified size. Only shrink is supported. A download may occur if size is not on chunk boundaries. :param account: name of the account in which the object is stored :param container: name of the container in which the object is stored :param obj: name of the object to query :param version: version of the object to query :param size: new size of object """ # code copied from object_fetch (should be factorized !) meta, raw_chunks = self.object_locate( account, container, obj, version=version, **kwargs) chunk_method = meta['chunk_method'] storage_method = STORAGE_METHODS.load(chunk_method) chunks = _sort_chunks(raw_chunks, storage_method.ec) for pos in sorted(chunks.keys()): chunk = chunks[pos][0] if (size >= chunk['offset'] and size <= chunk['offset'] + chunk['size']): break else: raise exc.OioException("No chunk found at position %d" % size) if chunk['offset'] != size: # retrieve partial chunk ret = self.object_fetch(account, container, obj, version=version, ranges=[(chunk['offset'], size-1)]) # TODO implement a proper object_update pos = int(chunk['pos'].split('.')[0]) self.object_create(account, container, obj_name=obj, data=ret[1], meta_pos=pos, content_id=meta['id']) return self.container.content_truncate(account, container, obj, version=version, size=size, **kwargs) @handle_container_not_found def object_list(self, account, container, limit=None, marker=None, delimiter=None, prefix=None, end_marker=None, properties=False, versions=False, deleted=False, **kwargs): """ Lists objects inside a container. :param properties: if True, list object properties along with objects :param versions: if True, list all versions of objects :param deleted: if True, list also the deleted objects :returns: a dict which contains * 'objects': the list of objects * 'prefixes': common prefixes (only if delimiter and prefix are set) * 'properties': a dict of container properties * 'system': a dict of system metadata """ _, resp_body = self.container.content_list( account, container, limit=limit, marker=marker, end_marker=end_marker, prefix=prefix, delimiter=delimiter, properties=properties, versions=versions, deleted=deleted, **kwargs) for obj in resp_body['objects']: mtype = obj.get('mime-type') if mtype is not None: obj['mime_type'] = mtype del obj['mime-type'] version = obj.get('ver') if version is not None: obj['version'] = version del obj['ver'] return resp_body @handle_object_not_found def object_locate(self, account, container, obj, version=None, **kwargs): """ Get a description of the object along with the list of its chunks. :param account: name of the account in which the object is stored :param container: name of the container in which the object is stored :param obj: name of the object to query :param version: version of the object to query :returns: a tuple with object metadata `dict` as first element and chunk `list` as second element """ obj_meta, chunks = self.container.content_locate( account, container, obj, version=version, **kwargs) return obj_meta, chunks def object_analyze(self, *args, **kwargs): """ :deprecated: use `object_locate` """ warnings.warn("You'd better use object_locate()", DeprecationWarning) return self.object_locate(*args, **kwargs) @ensure_headers @ensure_request_id def object_fetch(self, account, container, obj, version=None, ranges=None, key_file=None, **kwargs): meta, raw_chunks = self.object_locate( account, container, obj, version=version, **kwargs) chunk_method = meta['chunk_method'] storage_method = STORAGE_METHODS.load(chunk_method) chunks = _sort_chunks(raw_chunks, storage_method.ec) meta['container_id'] = cid_from_name(account, container).upper() meta['ns'] = self.namespace self._patch_timeouts(kwargs) if storage_method.ec: stream = fetch_stream_ec(chunks, ranges, storage_method, **kwargs) elif storage_method.backblaze: stream = self._fetch_stream_backblaze(meta, chunks, ranges, storage_method, key_file, **kwargs) else: stream = fetch_stream(chunks, ranges, storage_method, **kwargs) return meta, stream @handle_object_not_found def object_get_properties(self, account, container, obj, **kwargs): return self.container.content_get_properties(account, container, obj, **kwargs) @handle_object_not_found def object_show(self, account, container, obj, version=None, **kwargs): """ Get a description of the content along with its user properties. :param account: name of the account in which the object is stored :param container: name of the container in which the object is stored :param obj: name of the object to query :returns: a `dict` describing the object .. python:: {'hash': '6BF60C17CC15EEA108024903B481738F', 'ctime': '1481031763', 'deleted': 'False', 'properties': { u'projet': u'OpenIO-SDS'}, 'length': '43518', 'hash_method': 'md5', 'chunk_method': 'ec/algo=liberasurecode_rs_vand,k=6,m=3', 'version': '1481031762951972', 'policy': 'EC', 'id': '20BF2194FD420500CD4729AE0B5CBC07', 'mime_type': 'application/octet-stream', 'name': 'Makefile'} """ return self.container.content_show(account, container, obj, version=version, **kwargs) def object_update(self, account, container, obj, metadata, version=None, clear=False, **kwargs): warnings.warn("You'd better use object_set_properties()", DeprecationWarning, stacklevel=2) if clear: self.object_del_properties( account, container, obj, [], version=version, **kwargs) if metadata: self.object_set_properties( account, container, obj, metadata, version=version, **kwargs) @handle_object_not_found def object_set_properties(self, account, container, obj, properties, version=None, **kwargs): return self.container.content_set_properties( account, container, obj, properties={'properties': properties}, version=version, **kwargs) @handle_object_not_found def object_del_properties(self, account, container, obj, properties, version=None, **kwargs): return self.container.content_del_properties( account, container, obj, properties=properties, version=version, **kwargs) def _content_preparer(self, account, container, obj_name, policy=None, **kwargs): # TODO: optimize by asking more than one metachunk at a time obj_meta, first_body = self.container.content_prepare( account, container, obj_name, size=1, stgpol=policy, autocreate=True, **kwargs) storage_method = STORAGE_METHODS.load(obj_meta['chunk_method']) def _fix_mc_pos(chunks, mc_pos): for chunk in chunks: raw_pos = chunk["pos"].split(".") if storage_method.ec: chunk['num'] = int(raw_pos[1]) chunk["pos"] = "%d.%d" % (mc_pos, chunk['num']) else: chunk["pos"] = str(mc_pos) def _metachunk_preparer(): mc_pos = kwargs.get('meta_pos', 0) _fix_mc_pos(first_body, mc_pos) yield first_body while True: mc_pos += 1 _, next_body = self.container.content_prepare( account, container, obj_name, 1, stgpol=policy, autocreate=True, **kwargs) _fix_mc_pos(next_body, mc_pos) yield next_body return obj_meta, _metachunk_preparer def _generate_fullpath(self, account, container_name, path, version): return ['{0}/{1}/{2}/{3}'.format(quote_plus(account), quote_plus(container_name), quote_plus(path), version)] def _object_create(self, account, container, obj_name, source, sysmeta, properties=None, policy=None, key_file=None, **kwargs): self._patch_timeouts(kwargs) obj_meta, chunk_prep = self._content_preparer( account, container, obj_name, policy=policy, **kwargs) obj_meta.update(sysmeta) obj_meta['content_path'] = obj_name obj_meta['container_id'] = cid_from_name(account, container).upper() obj_meta['ns'] = self.namespace obj_meta['full_path'] = self._generate_fullpath(account, container, obj_name, obj_meta['version']) obj_meta['oio_version'] = (obj_meta.get('oio_version') or OIO_VERSION) # XXX content_id is necessary to update an existing object kwargs['content_id'] = kwargs.get('content_id', obj_meta['id']) storage_method = STORAGE_METHODS.load(obj_meta['chunk_method']) if storage_method.ec: handler = ECWriteHandler( source, obj_meta, chunk_prep, storage_method, **kwargs) elif storage_method.backblaze: backblaze_info = self._b2_credentials(storage_method, key_file) handler = BackblazeWriteHandler( source, obj_meta, chunk_prep, storage_method, backblaze_info, **kwargs) else: handler = ReplicatedWriteHandler( source, obj_meta, chunk_prep, storage_method, **kwargs) final_chunks, bytes_transferred, content_checksum = handler.stream() etag = obj_meta.get('etag') if etag and etag.lower() != content_checksum.lower(): raise exc.EtagMismatch( "given etag %s != computed %s" % (etag, content_checksum)) obj_meta['etag'] = content_checksum data = {'chunks': final_chunks, 'properties': properties or {}} # FIXME: we may just pass **obj_meta self.container.content_create( account, container, obj_name, size=bytes_transferred, checksum=content_checksum, data=data, stgpol=obj_meta['policy'], version=obj_meta['version'], mime_type=obj_meta['mime_type'], chunk_method=obj_meta['chunk_method'], **kwargs) return final_chunks, bytes_transferred, content_checksum def _b2_credentials(self, storage_method, key_file): key_file = key_file or '/etc/oio/sds/b2-appkey.conf' try: return BackblazeUtils.get_credentials(storage_method, key_file) except BackblazeUtilsException as err: raise exc.ConfigurationException(str(err)) def _fetch_stream_backblaze(self, meta, chunks, ranges, storage_method, key_file, **kwargs): backblaze_info = self._b2_credentials(storage_method, key_file) total_bytes = 0 current_offset = 0 size = None offset = 0 for pos in range(len(chunks)): if ranges: offset = ranges[pos][0] size = ranges[pos][1] if size is None: size = int(meta["length"]) chunk_size = int(chunks[pos][0]["size"]) if total_bytes >= size: break if current_offset + chunk_size > offset: if current_offset < offset: _offset = offset - current_offset else: _offset = 0 if chunk_size + total_bytes > size: _size = size - total_bytes else: _size = chunk_size handler = BackblazeChunkDownloadHandler( meta, chunks[pos], _offset, _size, backblaze_info=backblaze_info) stream = handler.get_stream() if not stream: raise exc.OioException("Error while downloading") total_bytes += len(stream) yield stream current_offset += chunk_size @handle_container_not_found def container_refresh(self, account, container, attempts=3, **kwargs): for i in range(attempts): try: self.account.container_reset(account, container, time.time()) except exc.Conflict: if i >= attempts - 1: raise try: self.container.container_touch(account, container) except exc.ClientException as e: if e.status != 406 and e.status != 431: raise # CODE_USER_NOTFOUND or CODE_CONTAINER_NOTFOUND metadata = dict() metadata["dtime"] = time.time() self.account.container_update(account, container, metadata) @handle_account_not_found def account_refresh(self, account, **kwargs): self.account.account_refresh(account) containers = self.container_list(account) for container in containers: try: self.container_refresh(account, container[0]) except exc.NoSuchContainer: # container remove in the meantime pass while containers: marker = containers[-1][0] containers = self.container_list(account, marker=marker) if containers: for container in containers: try: self.container_refresh(account, container[0]) except exc.NoSuchContainer: # container remove in the meantime pass def all_accounts_refresh(self, **kwargs): accounts = self.account_list() for account in accounts: try: self.account_refresh(account) except exc.NoSuchAccount: # account remove in the meantime pass @handle_account_not_found def account_flush(self, account): self.account.account_flush(account) def _random_buffer(self, dictionary, n): return ''.join(random.choice(dictionary) for _ in range(n)) def _generate_copy(self, chunks, random_hex=60): # random_hex is the number of hexadecimals characters to generate for # the copy path copies = [] for c in chunks: tmp = ''.join([c[:-random_hex], self._random_buffer('0123456789ABCDEF', random_hex)]) copies.append(tmp) return copies def _send_copy(self, targets, copies, fullpath): headers = {"x-oio-chunk-meta-full-path": fullpath} if not hasattr(self, "blob_client"): from oio.blob.client import BlobClient self.blob_client = BlobClient() for t, c in zip(targets, copies): self.blob_client.chunk_link(t, c, headers=headers).status def _prepare_update_meta2(self, targets, copies, account, container, content): targets_beans = [] copies_beans = [] for t, c in zip(targets, copies): targets_beans.append(self._meta2bean(t['url'], t, content)) copies_beans.append(self._meta2bean(c, t, content)) return targets_beans, copies_beans def _meta2bean(self, url, meta, content): return {"type": "chunk", "id": url, "hash": meta['hash'], "size": int(meta["size"]), "pos": meta["pos"], "content": content}
class BlobAuditorWorker(object): def __init__(self, conf, logger, volume): self.conf = conf self.logger = logger self.volume = volume self.run_time = 0 self.passes = 0 self.errors = 0 self.orphan_chunks = 0 self.faulty_chunks = 0 self.corrupted_chunks = 0 self.last_reported = 0 self.chunks_run_time = 0 self.bytes_running_time = 0 self.bytes_processed = 0 self.total_bytes_processed = 0 self.total_chunks_processed = 0 self.report_interval = int_value(conf.get("report_interval"), 3600) self.max_chunks_per_second = int_value(conf.get("chunks_per_second"), 30) self.max_bytes_per_second = int_value(conf.get("bytes_per_second"), 10000000) self.container_client = ContainerClient(conf) def audit_pass(self): self.namespace, self.address = check_volume(self.volume) start_time = report_time = time.time() total_errors = 0 total_corrupted = 0 total_orphans = 0 total_faulty = 0 audit_time = 0 paths = paths_gen(self.volume) for path in paths: loop_time = time.time() self.safe_chunk_audit(path) self.chunks_run_time = ratelimit(self.chunks_run_time, self.max_chunks_per_second) self.total_chunks_processed += 1 now = time.time() if now - self.last_reported >= self.report_interval: self.logger.info( "%(start_time)s " "%(passes)d " "%(corrupted)d " "%(faulty)d " "%(orphans)d " "%(errors)d " "%(c_rate).2f " "%(b_rate).2f " "%(total).2f " "%(audit_time).2f" "%(audit_rate).2f" % { "start_time": time.ctime(report_time), "passes": self.passes, "corrupted": self.corrupted_chunks, "faulty": self.faulty_chunks, "orphans": self.orphan_chunks, "errors": self.errors, "c_rate": self.passes / (now - report_time), "b_rate": self.bytes_processed / (now - report_time), "total": (now - start_time), "audit_time": audit_time, "audit_rate": audit_time / (now - start_time), } ) report_time = now total_corrupted += self.corrupted_chunks total_orphans += self.orphan_chunks total_faulty += self.faulty_chunks total_errors += self.errors self.passes = 0 self.corrupted_chunks = 0 self.orphan_chunks = 0 self.faulty_chunks = 0 self.errors = 0 self.bytes_processed = 0 self.last_reported = now audit_time += now - loop_time elapsed = (time.time() - start_time) or 0.000001 self.logger.info( "%(elapsed).02f " "%(corrupted)d " "%(faulty)d " "%(orphans)d " "%(errors)d " "%(chunk_rate).2f " "%(bytes_rate).2f " "%(audit_time).2f " "%(audit_rate).2f" % { "elapsed": elapsed, "corrupted": total_corrupted + self.corrupted_chunks, "faulty": total_faulty + self.faulty_chunks, "orphans": total_orphans + self.orphan_chunks, "errors": total_errors + self.errors, "chunk_rate": self.total_chunks_processed / elapsed, "bytes_rate": self.total_bytes_processed / elapsed, "audit_time": audit_time, "audit_rate": audit_time / elapsed, } ) def safe_chunk_audit(self, path): try: self.chunk_audit(path) except exc.FaultyChunk as err: self.faulty_chunks += 1 self.logger.error("ERROR faulty chunk %s: %s", path, err) except exc.CorruptedChunk as err: self.corrupted_chunks += 1 self.logger.error("ERROR corrupted chunk %s: %s", path, err) except exc.OrphanChunk as err: self.orphan_chunks += 1 self.logger.error("ERROR orphan chunk %s: %s", path, err) except Exception: self.errors += 1 self.logger.exception("ERROR while auditing chunk %s", path) self.passes += 1 def chunk_audit(self, path): with open(path) as f: try: meta = read_chunk_metadata(f) except exc.MissingAttribute as e: raise exc.FaultyChunk("Missing extended attribute %s" % e) size = int(meta["chunk_size"]) md5_checksum = meta["chunk_hash"].lower() reader = ChunkReader(f, size, md5_checksum) with closing(reader): for buf in reader: buf_len = len(buf) self.bytes_running_time = ratelimit( self.bytes_running_time, self.max_bytes_per_second, increment=buf_len ) self.bytes_processed += buf_len self.total_bytes_processed += buf_len try: container_id = meta["container_id"] content_path = meta["content_path"] content_attr, data = self.container_client.content_show(cid=container_id, path=content_path) # Check chunk data chunk_data = None metachunks = set() for c in data: if c["url"].endswith(meta["chunk_id"]): metachunks.add(c["pos"].split(".", 2)[0]) chunk_data = c if not chunk_data: raise exc.OrphanChunk("Not found in content") if chunk_data["size"] != int(meta["chunk_size"]): raise exc.FaultyChunk("Invalid chunk size found") if chunk_data["hash"] != meta["chunk_hash"]: raise exc.FaultyChunk("Invalid chunk hash found") if chunk_data["pos"] != meta["chunk_pos"]: raise exc.FaultyChunk("Invalid chunk position found") except exc.NotFound: raise exc.OrphanChunk("Chunk not found in container")
class TestDupContent(BaseTestCase): def setUp(self): super(TestDupContent, self).setUp() if len(self.conf['rawx']) < 3: self.skipTest("Not enough rawx. " "Dup tests needs more than 2 rawx to run") self.namespace = self.conf['namespace'] self.account = self.conf['account'] self.chunk_size = self.conf['chunk_size'] self.gridconf = {"namespace": self.namespace} self.content_factory = ContentFactory(self.gridconf) self.container_client = ContainerClient(self.gridconf) self.blob_client = BlobClient() self.container_name = "TestDupContent%f" % time.time() self.container_client.container_create(acct=self.account, ref=self.container_name) self.container_id = cid_from_name(self.account, self.container_name).upper() def tearDown(self): super(TestDupContent, self).tearDown() def _test_upload(self, stgpol, data_size): data = random_data(data_size) content = self.content_factory.new(self.container_id, "titi", len(data), stgpol) self.assertEqual(type(content), DupContent) content.upload(StringIO.StringIO(data)) meta, chunks = self.container_client.content_show( cid=self.container_id, content=content.content_id) chunks = ChunksHelper(chunks) self.assertEqual(meta['hash'], md5_data(data)) self.assertEqual(meta['length'], str(len(data))) self.assertEqual(meta['policy'], stgpol) self.assertEqual(meta['name'], "titi") metachunk_nb = int(math.ceil(float(len(data)) / self.chunk_size)) if metachunk_nb == 0: metachunk_nb = 1 # special case for empty content if stgpol == "THREECOPIES": nb_copy = 3 elif stgpol == "TWOCOPIES": nb_copy = 2 elif stgpol == "SINGLE": nb_copy = 1 self.assertEqual(len(chunks), metachunk_nb * nb_copy) for pos in range(metachunk_nb): chunks_at_pos = chunks.filter(pos=pos) self.assertEqual(len(chunks_at_pos), nb_copy) data_begin = pos * self.chunk_size data_end = pos * self.chunk_size + self.chunk_size chunk_hash = md5_data(data[data_begin:data_end]) for chunk in chunks_at_pos: meta, stream = self.blob_client.chunk_get(chunk.url) self.assertEqual(md5_stream(stream), chunk_hash) self.assertEqual(meta['content_size'], str(len(data))) self.assertEqual(meta['content_path'], "titi") self.assertEqual(meta['content_cid'], self.container_id) self.assertEqual(meta['content_id'], meta['content_id']) self.assertEqual(meta['chunk_id'], chunk.id) self.assertEqual(meta['chunk_pos'], str(pos)) self.assertEqual(meta['chunk_hash'], chunk_hash) def test_twocopies_upload_0_byte(self): self._test_upload("TWOCOPIES", 0) def test_twocopies_upload_1_byte(self): self._test_upload("TWOCOPIES", 1) def test_twocopies_upload_chunksize_bytes(self): self._test_upload("TWOCOPIES", self.chunk_size) def test_twocopies_upload_chunksize_plus_1_bytes(self): self._test_upload("TWOCOPIES", self.chunk_size + 1) def test_single_upload_0_byte(self): self._test_upload("SINGLE", 0) def test_single_upload_chunksize_plus_1_bytes(self): self._test_upload("SINGLE", self.chunk_size + 1) def test_chunks_cleanup_when_upload_failed(self): data = random_data(2 * self.chunk_size) content = self.content_factory.new(self.container_id, "titi", len(data), "TWOCOPIES") self.assertEqual(type(content), DupContent) # set bad url for position 1 for chunk in content.chunks.filter(pos=1): chunk.url = "http://127.0.0.1:9/DEADBEEF" self.assertRaises(Exception, content.upload, StringIO.StringIO(data)) for chunk in content.chunks.exclude(pos=1): self.assertRaises(NotFound, self.blob_client.chunk_head, chunk.url) def _new_content(self, stgpol, data, broken_pos_list=[]): old_content = self.content_factory.new(self.container_id, "titi", len(data), stgpol) self.assertEqual(type(old_content), DupContent) old_content.upload(StringIO.StringIO(data)) broken_chunks_info = {} for pos, idx in broken_pos_list: c = old_content.chunks.filter(pos=pos)[idx] meta, stream = self.blob_client.chunk_get(c.url) if pos not in broken_chunks_info: broken_chunks_info[pos] = {} broken_chunks_info[pos][idx] = { "url": c.url, "id": c.id, "hash": c.hash, "dl_meta": meta, "dl_hash": md5_stream(stream) } self.blob_client.chunk_delete(c.url) # get the new structure of the uploaded content return (self.content_factory.get( self.container_id, old_content.content_id), broken_chunks_info) def _test_rebuild(self, stgpol, data_size, broken_pos_list, full_rebuild_pos): data = random_data(data_size) content, broken_chunks_info = self._new_content(stgpol, data, broken_pos_list) rebuild_pos, rebuild_idx = full_rebuild_pos rebuild_chunk_info = broken_chunks_info[rebuild_pos][rebuild_idx] content.rebuild_chunk(rebuild_chunk_info["id"]) # get the new structure of the content rebuilt_content = self.content_factory.get(self.container_id, content.content_id) self.assertEqual(type(rebuilt_content), DupContent) # find the rebuilt chunk for c in rebuilt_content.chunks.filter(pos=rebuild_pos): if len(content.chunks.filter(id=c.id)) > 0: # not the rebuilt chunk # if this chunk is broken, it must not have been rebuilt for b_c_i in broken_chunks_info[rebuild_pos].values(): if c.id == b_c_i["id"]: with ExpectedException(NotFound): _, _ = self.blob_client.chunk_get(c.url) continue meta, stream = self.blob_client.chunk_get(c.url) self.assertEqual(meta["chunk_id"], c.id) self.assertEqual(md5_stream(stream), rebuild_chunk_info["dl_hash"]) self.assertEqual(c.hash, rebuild_chunk_info["hash"]) self.assertThat(c.url, NotEquals(rebuild_chunk_info["url"])) del meta["chunk_id"] del rebuild_chunk_info["dl_meta"]["chunk_id"] self.assertEqual(meta, rebuild_chunk_info["dl_meta"]) def test_2copies_content_0_byte_1broken_rebuild_pos_0_idx_0(self): self._test_rebuild("TWOCOPIES", 0, [(0, 0)], (0, 0)) def test_2copies_content_1_byte_1broken_rebuild_pos_0_idx_1(self): self._test_rebuild("TWOCOPIES", 1, [(0, 1)], (0, 1)) def test_3copies_content_chunksize_bytes_2broken_rebuild_pos_0_idx_1(self): if len(self.conf['rawx']) <= 3: self.skipTest("Need more than 3 rawx") self._test_rebuild("THREECOPIES", self.chunk_size, [(0, 0), (0, 1)], (0, 1)) def test_3copies_content_2xchksize_bytes_2broken_rebuild_pos_1_idx_2(self): if len(self.conf['rawx']) <= 3: self.skipTest("Need more than 3 rawx") self._test_rebuild("THREECOPIES", 2 * self.chunk_size, [(1, 0), (1, 2)], (1, 2)) def test_2copies_content_0_byte_2broken_rebuild_pos_0_idx_0(self): with ExpectedException(UnrecoverableContent): self._test_rebuild("TWOCOPIES", 0, [(0, 0), (0, 1)], (0, 0)) def _test_download(self, stgpol, data_size, broken_pos_list): data = random_data(data_size) content, _ = self._new_content(stgpol, data, broken_pos_list) downloaded_data = "".join(content.download()) self.assertEqual(downloaded_data, data) for pos, idx in broken_pos_list: # check nothing has been rebuilt c = content.chunks.filter(pos=pos)[0] self.assertRaises(NotFound, self.blob_client.chunk_delete, c.url) def test_twocopies_download_content_0_byte_without_broken_chunks(self): self._test_download("TWOCOPIES", 0, []) def test_twocopies_download_content_0_byte_with_broken_0_0(self): self._test_download("TWOCOPIES", 0, [(0, 0)]) def test_twocopies_download_content_1_byte_without_broken_chunks(self): self._test_download("TWOCOPIES", 1, []) def test_twocopies_download_content_1_byte_with_broken_0_0(self): self._test_download("TWOCOPIES", 1, [(0, 0)]) def test_twocopies_download_chunksize_bytes_without_broken_chunks(self): self._test_download("TWOCOPIES", self.chunk_size, []) def test_twocopies_download_2xchuksize_bytes_with_broken_0_0_and_1_0(self): self._test_download("TWOCOPIES", self.chunk_size * 2, [(0, 0), (1, 0)]) def test_twocopies_download_content_chunksize_bytes_2_broken_chunks(self): data = random_data(self.chunk_size) content, _ = self._new_content("TWOCOPIES", data, [(0, 0), (0, 1)]) gen = content.download() self.assertRaises(UnrecoverableContent, gen.next) def test_single_download_content_1_byte_without_broken_chunks(self): self._test_download("SINGLE", 1, []) def test_single_download_chunksize_bytes_plus_1_without_broken_chunk(self): self._test_download("SINGLE", self.chunk_size * 2, [])
class TestRainContent(BaseTestCase): def setUp(self): super(TestRainContent, self).setUp() if len(self.conf['rawx']) < 12: self.skipTest("Not enough rawx. " "Rain tests needs more than 12 rawx to run") self.namespace = self.conf['namespace'] self.account = self.conf['account'] self.chunk_size = self.conf['chunk_size'] self.gridconf = {"namespace": self.namespace} self.content_factory = ContentFactory(self.gridconf) self.container_client = ContainerClient(self.gridconf) self.blob_client = BlobClient() self.container_name = "TestRainContent%f" % time.time() self.container_client.container_create(acct=self.account, ref=self.container_name) self.container_id = cid_from_name(self.account, self.container_name).upper() def tearDown(self): super(TestRainContent, self).tearDown() def _test_upload(self, data_size): data = random_data(data_size) content = self.content_factory.new(self.container_id, "titi", len(data), "RAIN") k = 6 m = 2 self.assertEqual(type(content), RainContent) content.upload(StringIO.StringIO(data)) meta, chunks = self.container_client.content_show( cid=self.container_id, content=content.content_id) chunks = ChunksHelper(chunks) self.assertEqual(meta['hash'], md5_data(data)) self.assertEqual(meta['length'], str(len(data))) self.assertEqual(meta['policy'], "RAIN") self.assertEqual(meta['name'], "titi") metachunk_nb = int(math.ceil(float(len(data)) / self.chunk_size)) if metachunk_nb == 0: metachunk_nb = 1 # special case for empty content nb_chunks_min = metachunk_nb * (k + m) - (k - 1) nb_chunks_max = metachunk_nb * (k + m) self.assertEquals(len(chunks) >= nb_chunks_min, True) self.assertEquals(len(chunks) <= nb_chunks_max, True) for metapos in range(metachunk_nb): chunks_at_pos = content.chunks.filter(metapos=metapos) data_chunks_at_pos = chunks_at_pos.filter(is_parity=False) parity_chunks_at_pos = chunks_at_pos.filter(is_parity=True) if metapos < metachunk_nb - 1: self.assertEqual(len(data_chunks_at_pos), k) else: self.assertEquals(len(data_chunks_at_pos) >= 1, True) self.assertEquals(len(data_chunks_at_pos) <= k, True) self.assertEqual(len(parity_chunks_at_pos), m) for chunk in chunks_at_pos: meta, stream = self.blob_client.chunk_get(chunk.url) self.assertEqual(md5_stream(stream), chunk.hash) self.assertEqual(meta['content_size'], str(len(data))) self.assertEqual(meta['content_path'], "titi") self.assertEqual(meta['content_cid'], self.container_id) self.assertEqual(meta['content_id'], meta['content_id']) self.assertEqual(meta['chunk_id'], chunk.id) self.assertEqual(meta['chunk_pos'], chunk.pos) self.assertEqual(meta['chunk_hash'], chunk.hash) data_begin = metapos * self.chunk_size data_end = metapos * self.chunk_size + self.chunk_size target_metachunk_hash = md5_data(data[data_begin:data_end]) metachunk_hash = hashlib.md5() for chunk in data_chunks_at_pos: meta, stream = self.blob_client.chunk_get(chunk.url) for d in stream: metachunk_hash.update(d) self.assertEqual(metachunk_hash.hexdigest().upper(), target_metachunk_hash) def test_upload_0_byte(self): self._test_upload(0) def test_upload_1_byte(self): self._test_upload(1) def test_upload_chunksize_bytes(self): self._test_upload(self.chunk_size) def test_upload_chunksize_plus_1_bytes(self): self._test_upload(self.chunk_size + 1) def test_chunks_cleanup_when_upload_failed(self): data = random_data(2 * self.chunk_size) content = self.content_factory.new(self.container_id, "titi", len(data), "RAIN") self.assertEqual(type(content), RainContent) # set bad url for position 1 for chunk in content.chunks.filter(pos="1.p0"): chunk.url = "http://127.0.0.1:9/DEADBEEF" self.assertRaises(Exception, content.upload, StringIO.StringIO(data)) for chunk in content.chunks.exclude(pos="1.p0"): self.assertRaises(NotFound, self.blob_client.chunk_head, chunk.url) def _test_rebuild(self, data_size, broken_pos_list): data = os.urandom(data_size) old_content = self.content_factory.new(self.container_id, "titi", len(data), "RAIN") self.assertEqual(type(old_content), RainContent) old_content.upload(StringIO.StringIO(data)) # get the new structure of the uploaded content uploaded_content = self.content_factory.get(self.container_id, old_content.content_id) old_info = {} for pos in broken_pos_list: old_info[pos] = {} c = uploaded_content.chunks.filter(pos=pos)[0] old_info[pos]["url"] = c.url old_info[pos]["id"] = c.id old_info[pos]["hash"] = c.hash chunk_id_to_rebuild = c.id meta, stream = self.blob_client.chunk_get(c.url) old_info[pos]["dl_meta"] = meta old_info[pos]["dl_hash"] = md5_stream(stream) # delete the chunk self.blob_client.chunk_delete(c.url) # rebuild the broken chunks uploaded_content.rebuild_chunk(chunk_id_to_rebuild) # get the new structure of the content rebuilt_content = self.content_factory.get(self.container_id, uploaded_content.content_id) self.assertEqual(type(rebuilt_content), RainContent) for pos in broken_pos_list: c = rebuilt_content.chunks.filter(pos=pos)[0] rebuilt_meta, rebuilt_stream = self.blob_client.chunk_get(c.url) self.assertEqual(rebuilt_meta["chunk_id"], c.id) self.assertEqual(md5_stream(rebuilt_stream), old_info[pos]["dl_hash"]) self.assertEqual(c.hash, old_info[pos]["hash"]) self.assertThat(c.url, NotEquals(old_info[pos]["url"])) del old_info[pos]["dl_meta"]["chunk_id"] del rebuilt_meta["chunk_id"] self.assertEqual(rebuilt_meta, old_info[pos]["dl_meta"]) def test_content_0_byte_rebuild_pos_0_0(self): self._test_rebuild(0, ["0.0"]) def test_content_0_byte_rebuild_pos_0_0_and_0_p0(self): self._test_rebuild(0, ["0.0", "0.p0"]) def test_content_1_byte_rebuild_pos_0_0(self): self._test_rebuild(1, ["0.0"]) def test_content_1_byte_rebuild_pos_0_p0(self): self._test_rebuild(1, ["0.p0"]) def test_content_1_byte_rebuild_pos_0_0_and_0_p0(self): self._test_rebuild(1, ["0.0", "0.p0"]) def test_content_chunksize_bytes_rebuild_pos_0_0(self): self._test_rebuild(self.conf["chunk_size"], ["0.0"]) def test_content_chunksize_bytes_rebuild_pos_0_0_and_0_1(self): self._test_rebuild(self.conf["chunk_size"], ["0.0", "0.1"]) def test_content_chunksize_bytes_rebuild_pos_0_0_and_0_p0(self): self._test_rebuild(self.conf["chunk_size"], ["0.0", "0.p0"]) def test_content_chunksize_bytes_rebuild_pos_0_p0_and_0_p1(self): self._test_rebuild(self.conf["chunk_size"], ["0.p0", "0.p1"]) def test_content_chunksize_bytes_rebuild_more_than_k_chunk(self): self.assertRaises(UnrecoverableContent, self._test_rebuild, self.conf["chunk_size"], ["0.0", "0.1", "0.2"]) def _new_content(self, data, broken_pos_list=[]): old_content = self.content_factory.new(self.container_id, "titi", len(data), "RAIN") self.assertEqual(type(old_content), RainContent) old_content.upload(StringIO.StringIO(data)) for pos in broken_pos_list: c = old_content.chunks.filter(pos=pos)[0] self.blob_client.chunk_delete(c.url) # get the new structure of the uploaded content return self.content_factory.get(self.container_id, old_content.content_id) def test_orphan_chunk(self): content = self._new_content(random_data(10)) self.assertRaises(OrphanChunk, content.rebuild_chunk, "uNkNoWnId") def test_rebuild_on_the_fly(self): data = random_data(self.conf["chunk_size"]) content = self._new_content(data, ["0.0", "0.p0"]) stream = content.rebuild_metachunk("0", on_the_fly=True) dl_data = "".join(stream) self.assertEqual(dl_data, data) del_chunk_0_0 = content.chunks.filter(pos="0.0")[0] del_chunk_0_p0 = content.chunks.filter(pos="0.p0")[0] self.assertRaises(NotFound, self.blob_client.chunk_get, del_chunk_0_0.url) self.assertRaises(NotFound, self.blob_client.chunk_get, del_chunk_0_p0.url) def _test_download(self, data_size, broken_pos_list): data = random_data(data_size) content = self._new_content(data, broken_pos_list) downloaded_data = "".join(content.download()) self.assertEqual(downloaded_data, data) for pos in broken_pos_list: c = content.chunks.filter(pos=pos)[0] self.assertRaises(NotFound, self.blob_client.chunk_delete, c.url) def test_download_content_0_byte_without_broken_chunks(self): self._test_download(0, []) def test_download_content_1_byte_without_broken_chunks(self): self._test_download(1, []) def test_download_content_chunksize_bytes_without_broken_chunks(self): self._test_download(self.conf["chunk_size"], []) def test_download_content_chunksize_plus_1_without_broken_chunks(self): self._test_download(self.conf["chunk_size"] + 1, []) def test_download_content_0_byte_with_broken_0_0_and_0_p0(self): self._test_download(0, ["0.0", "0.p0"]) def test_download_content_1_byte_with_broken_0_0_and_0_p0(self): self._test_download(1, ["0.0", "0.p0"]) def test_download_content_2xchunksize_with_broken_0_2_and_1_0(self): self._test_download(2 * self.conf["chunk_size"], ["0.2", "1.0"]) def test_download_content_chunksize_bytes_with_3_broken_chunks(self): data = random_data(self.conf["chunk_size"]) content = self._new_content(data, ["0.0", "0.1", "0.2"]) gen = content.download() self.assertRaises(UnrecoverableContent, gen.next) def test_download_interrupt_close(self): data = random_data(self.conf["chunk_size"]) content = self._new_content(data, ["0.p0"]) download_iter = content.download() self.assertEqual(download_iter.next(), data[0:READ_CHUNK_SIZE - 1]) download_iter.close()
class ContentFactory(object): def __init__(self, conf): self.conf = conf self.logger = get_logger(conf) self.cs_client = ConscienceClient(conf) self.container_client = ContainerClient(conf) self.ns_info = self.cs_client.info() def _extract_datasec(self, stgpol_name): try: stgpol = self.ns_info["storage_policy"][stgpol_name] except KeyError: self.logger.error("Storage policy '%s' not found" % stgpol_name) raise InconsistentContent("Storage policy not found") stgclass_name, datasec_name, datatreat_name = stgpol.split(':') if datasec_name == 'NONE': return "DUP", {"nb_copy": "1", "distance": "0"} try: datasec = self.ns_info["data_security"][datasec_name] except KeyError: self.logger.error("Data security '%s' not found" % datasec_name) raise InconsistentContent("Data security not found") ds_type, ds_args = datasec.split(':') args = {} for arg in ds_args.split('|'): key, value = arg.split('=') args[key] = value return ds_type, args def get(self, container_id, content_id): try: meta, chunks = self.container_client.content_show( cid=container_id, content=content_id) except NotFound: raise ContentNotFound("Content %s/%s not found" % (container_id, content_id)) pol_type, pol_args = self._extract_datasec(meta['policy']) if pol_type == "DUP": return DupContent(self.conf, container_id, meta, chunks, pol_args) elif pol_type == "RAIN": return RainContent(self.conf, container_id, meta, chunks, pol_args) raise InconsistentContent("Unknown storage policy") def new(self, container_id, path, size, policy): meta, chunks = self.container_client.content_prepare(cid=container_id, path=path, size=size, stgpol=policy) pol_type, pol_args = self._extract_datasec(meta['policy']) if pol_type == "DUP": return DupContent(self.conf, container_id, meta, chunks, pol_args) elif pol_type == "RAIN": return RainContent(self.conf, container_id, meta, chunks, pol_args) raise InconsistentContent("Unknown storage policy") def change_policy(self, container_id, content_id, new_policy): old_content = self.get(container_id, content_id) if old_content.stgpol_name == new_policy: return old_content new_content = self.new(container_id, old_content.path, old_content.length, new_policy) stream = old_content.download() new_content.upload(GeneratorIO(stream)) # the old content is automatically deleted because the new content has # the same name (but not the same id) return new_content
class TestRainContent(BaseTestCase): def setUp(self): super(TestRainContent, self).setUp() if len(self.conf['rawx']) < 12: self.skipTest("Not enough rawx. " "Rain tests needs more than 12 rawx to run") self.namespace = self.conf['namespace'] self.account = self.conf['account'] self.chunk_size = self.conf['chunk_size'] self.gridconf = {"namespace": self.namespace} self.content_factory = ContentFactory(self.gridconf) self.container_client = ContainerClient(self.gridconf) self.blob_client = BlobClient() self.container_name = "TestRainContent%f" % time.time() self.container_client.container_create(acct=self.account, ref=self.container_name) self.container_id = cid_from_name(self.account, self.container_name).upper() def tearDown(self): super(TestRainContent, self).tearDown() def _test_upload(self, data_size): data = random_data(data_size) content = self.content_factory.new(self.container_id, "titi", len(data), "RAIN") k = 6 m = 2 self.assertEqual(type(content), RainContent) content.upload(StringIO.StringIO(data)) meta, chunks = self.container_client.content_show( cid=self.container_id, content=content.content_id) chunks = ChunksHelper(chunks) self.assertEqual(meta['hash'], md5_data(data)) self.assertEqual(meta['length'], str(len(data))) self.assertEqual(meta['policy'], "RAIN") self.assertEqual(meta['name'], "titi") metachunk_nb = int(math.ceil(float(len(data)) / self.chunk_size)) if metachunk_nb == 0: metachunk_nb = 1 # special case for empty content nb_chunks_min = metachunk_nb * (1 + m) nb_chunks_max = metachunk_nb * (k + m) self.assertGreaterEqual(len(chunks), nb_chunks_min) self.assertLessEqual(len(chunks), nb_chunks_max) for metapos in range(metachunk_nb): chunks_at_pos = content.chunks.filter(metapos=metapos) data_chunks_at_pos = chunks_at_pos.filter(is_parity=False) parity_chunks_at_pos = chunks_at_pos.filter(is_parity=True) self.assertEquals(len(data_chunks_at_pos) >= 1, True) self.assertEquals(len(data_chunks_at_pos) <= k, True) self.assertEqual(len(parity_chunks_at_pos), m) for chunk in chunks_at_pos: meta, stream = self.blob_client.chunk_get(chunk.url) self.assertEqual(md5_stream(stream), chunk.hash) self.assertEqual(meta['content_size'], str(len(data))) self.assertEqual(meta['content_path'], "titi") self.assertEqual(meta['content_cid'], self.container_id) self.assertEqual(meta['content_id'], meta['content_id']) self.assertEqual(meta['chunk_id'], chunk.id) self.assertEqual(meta['chunk_pos'], chunk.pos) self.assertEqual(meta['chunk_hash'], chunk.hash) data_begin = metapos * self.chunk_size data_end = metapos * self.chunk_size + self.chunk_size target_metachunk_hash = md5_data(data[data_begin:data_end]) metachunk_hash = hashlib.md5() for chunk in data_chunks_at_pos: meta, stream = self.blob_client.chunk_get(chunk.url) for d in stream: metachunk_hash.update(d) self.assertEqual(metachunk_hash.hexdigest().upper(), target_metachunk_hash) def test_upload_0_byte(self): self._test_upload(0) def test_upload_1_byte(self): self._test_upload(1) def test_upload_chunksize_bytes(self): self._test_upload(self.chunk_size) def test_upload_chunksize_plus_1_bytes(self): self._test_upload(self.chunk_size + 1) def test_chunks_cleanup_when_upload_failed(self): data = random_data(2 * self.chunk_size) content = self.content_factory.new(self.container_id, "titi", len(data), "RAIN") self.assertEqual(type(content), RainContent) # set bad url for position 1 for chunk in content.chunks.filter(pos="1.p0"): chunk.url = "http://127.0.0.1:9/DEADBEEF" self.assertRaises(Exception, content.upload, StringIO.StringIO(data)) for chunk in content.chunks.exclude(pos="1.p0"): self.assertRaises(NotFound, self.blob_client.chunk_head, chunk.url) def _test_rebuild(self, data_size, broken_pos_list): data = os.urandom(data_size) old_content = self.content_factory.new(self.container_id, "titi", len(data), "RAIN") self.assertEqual(type(old_content), RainContent) old_content.upload(StringIO.StringIO(data)) # get the new structure of the uploaded content uploaded_content = self.content_factory.get(self.container_id, old_content.content_id) old_info = {} for pos in broken_pos_list: old_info[pos] = {} c = uploaded_content.chunks.filter(pos=pos)[0] old_info[pos]["url"] = c.url old_info[pos]["id"] = c.id old_info[pos]["hash"] = c.hash chunk_id_to_rebuild = c.id meta, stream = self.blob_client.chunk_get(c.url) old_info[pos]["dl_meta"] = meta old_info[pos]["dl_hash"] = md5_stream(stream) # delete the chunk self.blob_client.chunk_delete(c.url) # rebuild the broken chunks uploaded_content.rebuild_chunk(chunk_id_to_rebuild) # get the new structure of the content rebuilt_content = self.content_factory.get(self.container_id, uploaded_content.content_id) self.assertEqual(type(rebuilt_content), RainContent) for pos in broken_pos_list: c = rebuilt_content.chunks.filter(pos=pos)[0] rebuilt_meta, rebuilt_stream = self.blob_client.chunk_get(c.url) self.assertEqual(rebuilt_meta["chunk_id"], c.id) self.assertEqual(md5_stream(rebuilt_stream), old_info[pos]["dl_hash"]) self.assertEqual(c.hash, old_info[pos]["hash"]) self.assertThat(c.url, NotEquals(old_info[pos]["url"])) del old_info[pos]["dl_meta"]["chunk_id"] del rebuilt_meta["chunk_id"] self.assertEqual(rebuilt_meta, old_info[pos]["dl_meta"]) def test_content_0_byte_rebuild_pos_0_0(self): self._test_rebuild(0, ["0.0"]) def test_content_0_byte_rebuild_pos_0_0_and_0_p0(self): self._test_rebuild(0, ["0.0", "0.p0"]) def test_content_1_byte_rebuild_pos_0_0(self): self._test_rebuild(1, ["0.0"]) def test_content_1_byte_rebuild_pos_0_p0(self): self._test_rebuild(1, ["0.p0"]) def test_content_1_byte_rebuild_pos_0_0_and_0_p0(self): self._test_rebuild(1, ["0.0", "0.p0"]) def test_content_chunksize_bytes_rebuild_pos_0_0(self): self._test_rebuild(self.conf["chunk_size"], ["0.0"]) def test_content_chunksize_bytes_rebuild_pos_0_0_and_0_1(self): self._test_rebuild(self.conf["chunk_size"], ["0.0", "0.1"]) def test_content_chunksize_bytes_rebuild_pos_0_0_and_0_p0(self): self._test_rebuild(self.conf["chunk_size"], ["0.0", "0.p0"]) def test_content_chunksize_bytes_rebuild_pos_0_p0_and_0_p1(self): self._test_rebuild(self.conf["chunk_size"], ["0.p0", "0.p1"]) def test_content_chunksize_bytes_rebuild_more_than_k_chunk(self): self.assertRaises(UnrecoverableContent, self._test_rebuild, self.conf["chunk_size"], ["0.0", "0.1", "0.2"]) def _new_content(self, data, broken_pos_list=[]): old_content = self.content_factory.new(self.container_id, "titi", len(data), "RAIN") self.assertEqual(type(old_content), RainContent) old_content.upload(StringIO.StringIO(data)) for pos in broken_pos_list: c = old_content.chunks.filter(pos=pos)[0] self.blob_client.chunk_delete(c.url) # get the new structure of the uploaded content return self.content_factory.get(self.container_id, old_content.content_id) def test_orphan_chunk(self): content = self._new_content(random_data(10)) self.assertRaises(OrphanChunk, content.rebuild_chunk, "uNkNoWnId") def test_rebuild_on_the_fly(self): data = random_data(self.conf["chunk_size"]) content = self._new_content(data, ["0.0", "0.p0"]) stream = content.rebuild_metachunk("0", on_the_fly=True) dl_data = "".join(stream) self.assertEqual(dl_data, data) del_chunk_0_0 = content.chunks.filter(pos="0.0")[0] del_chunk_0_p0 = content.chunks.filter(pos="0.p0")[0] self.assertRaises(NotFound, self.blob_client.chunk_get, del_chunk_0_0.url) self.assertRaises(NotFound, self.blob_client.chunk_get, del_chunk_0_p0.url) def _test_download(self, data_size, broken_pos_list): data = random_data(data_size) content = self._new_content(data, broken_pos_list) downloaded_data = "".join(content.download()) self.assertEqual(downloaded_data, data) for pos in broken_pos_list: c = content.chunks.filter(pos=pos)[0] self.assertRaises(NotFound, self.blob_client.chunk_delete, c.url) def test_download_content_0_byte_without_broken_chunks(self): self._test_download(0, []) def test_download_content_1_byte_without_broken_chunks(self): self._test_download(1, []) def test_download_content_chunksize_bytes_without_broken_chunks(self): self._test_download(self.conf["chunk_size"], []) def test_download_content_chunksize_plus_1_without_broken_chunks(self): self._test_download(self.conf["chunk_size"] + 1, []) def test_download_content_0_byte_with_broken_0_0_and_0_p0(self): self._test_download(0, ["0.0", "0.p0"]) def test_download_content_1_byte_with_broken_0_0_and_0_p0(self): self._test_download(1, ["0.0", "0.p0"]) def test_download_content_2xchunksize_with_broken_0_2_and_1_0(self): self._test_download(2 * self.conf["chunk_size"], ["0.2", "1.0"]) def test_download_content_chunksize_bytes_with_3_broken_chunks(self): data = random_data(self.conf["chunk_size"]) content = self._new_content(data, ["0.0", "0.1", "0.2"]) gen = content.download() self.assertRaises(UnrecoverableContent, gen.next) def test_download_interrupt_close(self): data = random_data(self.conf["chunk_size"]) content = self._new_content(data, ["0.p0"]) download_iter = content.download() dl_data = "" for buf in download_iter: dl_data += buf self.assertEqual(len(dl_data), len(data)) self.assertEqual(dl_data, data) download_iter.close()
class BlobRebuilderWorker(object): def __init__(self, conf, logger, volume): self.conf = conf self.logger = logger or get_logger(conf) self.volume = volume self.run_time = 0 self.passes = 0 self.errors = 0 self.last_reported = 0 self.chunks_run_time = 0 self.bytes_running_time = 0 self.bytes_processed = 0 self.total_bytes_processed = 0 self.total_chunks_processed = 0 self.dry_run = true_value( conf.get('dry_run', False)) self.report_interval = int_value( conf.get('report_interval'), 3600) self.max_chunks_per_second = int_value( conf.get('chunks_per_second'), 30) self.max_bytes_per_second = int_value( conf.get('bytes_per_second'), 10000000) self.rdir_fetch_limit = int_value( conf.get('rdir_fetch_limit'), 100) self.blob_client = BlobClient() self.container_client = ContainerClient(conf) self.rdir_client = RdirClient(conf) def rebuilder_pass_with_lock(self): self.rdir_client.admin_lock(self.volume, "rebuilder on %s" % gethostname()) try: self.rebuilder_pass() finally: self.rdir_client.admin_unlock(self.volume) def rebuilder_pass(self): start_time = report_time = time.time() total_errors = 0 rebuilder_time = 0 chunks = self.rdir_client.chunk_fetch(self.volume, limit=self.rdir_fetch_limit, rebuild=True) for container_id, content_id, chunk_id, data in chunks: loop_time = time.time() if self.dry_run: self.dryrun_chunk_rebuild(container_id, content_id, chunk_id) else: self.safe_chunk_rebuild(container_id, content_id, chunk_id) self.chunks_run_time = ratelimit( self.chunks_run_time, self.max_chunks_per_second ) self.total_chunks_processed += 1 now = time.time() if now - self.last_reported >= self.report_interval: self.logger.info( '%(start_time)s ' '%(passes)d ' '%(errors)d ' '%(c_rate).2f ' '%(b_rate).2f ' '%(total).2f ' '%(rebuilder_time).2f' '%(rebuilder_rate).2f' % { 'start_time': time.ctime(report_time), 'passes': self.passes, 'errors': self.errors, 'c_rate': self.passes / (now - report_time), 'b_rate': self.bytes_processed / (now - report_time), 'total': (now - start_time), 'rebuilder_time': rebuilder_time, 'rebuilder_rate': rebuilder_time / (now - start_time) } ) report_time = now total_errors += self.errors self.passes = 0 self.bytes_processed = 0 self.last_reported = now rebuilder_time += (now - loop_time) elapsed = (time.time() - start_time) or 0.000001 self.logger.info( '%(elapsed).02f ' '%(errors)d ' '%(chunk_rate).2f ' '%(bytes_rate).2f ' '%(rebuilder_time).2f ' '%(rebuilder_rate).2f' % { 'elapsed': elapsed, 'errors': total_errors + self.errors, 'chunk_rate': self.total_chunks_processed / elapsed, 'bytes_rate': self.total_bytes_processed / elapsed, 'rebuilder_time': rebuilder_time, 'rebuilder_rate': rebuilder_time / elapsed } ) def dryrun_chunk_rebuild(self, container_id, content_id, chunk_id): self.logger.info("[dryrun] Rebuilding " "container %s, content %s, chunk %s" % (container_id, content_id, chunk_id)) self.passes += 1 def safe_chunk_rebuild(self, container_id, content_id, chunk_id): self.logger.info('Rebuilding (container %s, content %s, chunk %s)' % (container_id, content_id, chunk_id)) try: self.chunk_rebuild(container_id, content_id, chunk_id) except Exception as e: self.errors += 1 self.logger.error('ERROR while rebuilding chunk %s|%s|%s) : %s', container_id, content_id, chunk_id, e) self.passes += 1 def _meta2_get_chunks_at_pos(self, container_id, content_id, chunk_id): current_chunk_url = 'http://%s/%s' % (self.volume, chunk_id) try: data = self.container_client.content_show( cid=container_id, content=content_id) except exc.NotFound: raise exc.OrphanChunk('Content not found') current_chunk = None for c in data: if c['url'] == current_chunk_url: current_chunk = c break if not current_chunk: raise exc.OrphanChunk('Chunk not found in content') duplicate_chunks = [] for c in data: if c['pos'] == current_chunk['pos'] \ and c['url'] != current_chunk['url']: duplicate_chunks.append(c) if len(duplicate_chunks) == 0: raise exc.UnrecoverableContent('No copy of missing chunk') return current_chunk, duplicate_chunks def _meta2_get_spare_chunk(self, container_id, content_id, notin, broken): spare_data = {'notin': notin, 'broken': [broken], 'size': 0} try: spare_resp = self.container_client.content_spare( cid=container_id, content=content_id, data=spare_data) except ClientException as e: raise exc.SpareChunkException('No spare chunk (%s)' % e.message) return spare_resp['chunks'][0] def _meta2_replace_chunk(self, container_id, content_id, current_chunk, new_chunk): old = [{'type': 'chunk', 'id': current_chunk['url'], 'hash': current_chunk['hash'], 'size': current_chunk['size'], 'pos': current_chunk['pos'], 'content': content_id}] new = [{'type': 'chunk', 'id': new_chunk['id'], 'hash': current_chunk['hash'], 'size': current_chunk['size'], 'pos': current_chunk['pos'], 'content': content_id}] update_data = {'old': old, 'new': new} self.container_client.container_raw_update( cid=container_id, data=update_data) # TODO rain support def chunk_rebuild(self, container_id, content_id, chunk_id): current_chunk, duplicate_chunks = self._meta2_get_chunks_at_pos( container_id, content_id, chunk_id) spare_chunk = self._meta2_get_spare_chunk( container_id, content_id, duplicate_chunks, current_chunk) uploaded = False for src in duplicate_chunks: try: self.blob_client.chunk_copy(src['url'], spare_chunk['id']) self.logger.debug('copy chunk from %s to %s', src['url'], spare_chunk['id']) uploaded = True break except Exception as e: self.logger.debug('Failed to copy chunk from %s to %s: %s', src['url'], spare_chunk['id'], type(e)) if not uploaded: raise exc.UnrecoverableContent('No copy available ' 'of missing chunk') self._meta2_replace_chunk(container_id, content_id, current_chunk, spare_chunk) self.rdir_client.chunk_push(self.volume, container_id, content_id, chunk_id, rtime=int(time.time())) self.bytes_processed += current_chunk['size'] self.total_bytes_processed += current_chunk['size']
class TestDupContent(BaseTestCase): def setUp(self): super(TestDupContent, self).setUp() if len(self.conf['rawx']) < 3: self.skipTest("Not enough rawx. " "Dup tests needs more than 2 rawx to run") self.namespace = self.conf['namespace'] self.account = self.conf['account'] self.chunk_size = self.conf['chunk_size'] self.gridconf = {"namespace": self.namespace} self.content_factory = ContentFactory(self.gridconf) self.container_client = ContainerClient(self.gridconf) self.blob_client = BlobClient() self.container_name = "TestDupContent%f" % time.time() self.container_client.container_create(acct=self.account, ref=self.container_name) self.container_id = cid_from_name(self.account, self.container_name).upper() def tearDown(self): super(TestDupContent, self).tearDown() def _test_upload(self, stgpol, data_size): data = random_data(data_size) content = self.content_factory.new(self.container_id, "titi", len(data), stgpol) self.assertEqual(type(content), DupContent) content.upload(StringIO.StringIO(data)) meta, chunks = self.container_client.content_show( cid=self.container_id, content=content.content_id) chunks = ChunksHelper(chunks) self.assertEqual(meta['hash'], md5_data(data)) self.assertEqual(meta['length'], str(len(data))) self.assertEqual(meta['policy'], stgpol) self.assertEqual(meta['name'], "titi") metachunk_nb = int(math.ceil(float(len(data)) / self.chunk_size)) if metachunk_nb == 0: metachunk_nb = 1 # special case for empty content if stgpol == "THREECOPIES": nb_copy = 3 elif stgpol == "TWOCOPIES": nb_copy = 2 elif stgpol == "SINGLE": nb_copy = 1 self.assertEqual(len(chunks), metachunk_nb * nb_copy) for pos in range(metachunk_nb): chunks_at_pos = chunks.filter(pos=pos) self.assertEqual(len(chunks_at_pos), nb_copy) data_begin = pos * self.chunk_size data_end = pos * self.chunk_size + self.chunk_size chunk_hash = md5_data(data[data_begin:data_end]) for chunk in chunks_at_pos: meta, stream = self.blob_client.chunk_get(chunk.url) self.assertEqual(md5_stream(stream), chunk_hash) self.assertEqual(meta['content_size'], str(len(data))) self.assertEqual(meta['content_path'], "titi") self.assertEqual(meta['content_cid'], self.container_id) self.assertEqual(meta['content_id'], meta['content_id']) self.assertEqual(meta['chunk_id'], chunk.id) self.assertEqual(meta['chunk_pos'], str(pos)) self.assertEqual(meta['chunk_hash'], chunk_hash) def test_twocopies_upload_0_byte(self): self._test_upload("TWOCOPIES", 0) def test_twocopies_upload_1_byte(self): self._test_upload("TWOCOPIES", 1) def test_twocopies_upload_chunksize_bytes(self): self._test_upload("TWOCOPIES", self.chunk_size) def test_twocopies_upload_chunksize_plus_1_bytes(self): self._test_upload("TWOCOPIES", self.chunk_size + 1) def test_single_upload_0_byte(self): self._test_upload("SINGLE", 0) def test_single_upload_chunksize_plus_1_bytes(self): self._test_upload("SINGLE", self.chunk_size + 1) def test_chunks_cleanup_when_upload_failed(self): data = random_data(2 * self.chunk_size) content = self.content_factory.new(self.container_id, "titi", len(data), "TWOCOPIES") self.assertEqual(type(content), DupContent) # set bad url for position 1 for chunk in content.chunks.filter(pos=1): chunk.url = "http://127.0.0.1:9/DEADBEEF" self.assertRaises(Exception, content.upload, StringIO.StringIO(data)) for chunk in content.chunks.exclude(pos=1): self.assertRaises(NotFound, self.blob_client.chunk_head, chunk.url) def _new_content(self, stgpol, data, broken_pos_list): old_content = self.content_factory.new(self.container_id, "titi", len(data), stgpol) self.assertEqual(type(old_content), DupContent) old_content.upload(StringIO.StringIO(data)) broken_chunks_info = {} for pos, idx in broken_pos_list: c = old_content.chunks.filter(pos=pos)[idx] meta, stream = self.blob_client.chunk_get(c.url) if pos not in broken_chunks_info: broken_chunks_info[pos] = {} broken_chunks_info[pos][idx] = { "url": c.url, "id": c.id, "hash": c.hash, "dl_meta": meta, "dl_hash": md5_stream(stream) } self.blob_client.chunk_delete(c.url) # get the new structure of the uploaded content return (self.content_factory.get(self.container_id, old_content.content_id), broken_chunks_info) def _test_rebuild(self, stgpol, data_size, broken_pos_list, full_rebuild_pos): data = random_data(data_size) content, broken_chunks_info = self._new_content( stgpol, data, broken_pos_list) rebuild_pos, rebuild_idx = full_rebuild_pos rebuild_chunk_info = broken_chunks_info[rebuild_pos][rebuild_idx] content.rebuild_chunk(rebuild_chunk_info["id"]) # get the new structure of the content rebuilt_content = self.content_factory.get(self.container_id, content.content_id) self.assertEqual(type(rebuilt_content), DupContent) # find the rebuilt chunk for c in rebuilt_content.chunks.filter(pos=rebuild_pos): if len(content.chunks.filter(id=c.id)) > 0: # not the rebuilt chunk # if this chunk is broken, it must not have been rebuilt for b_c_i in broken_chunks_info[rebuild_pos].values(): if c.id == b_c_i["id"]: with ExpectedException(NotFound): _, _ = self.blob_client.chunk_get(c.url) continue meta, stream = self.blob_client.chunk_get(c.url) self.assertEqual(meta["chunk_id"], c.id) self.assertEqual(md5_stream(stream), rebuild_chunk_info["dl_hash"]) self.assertEqual(c.hash, rebuild_chunk_info["hash"]) self.assertThat(c.url, NotEquals(rebuild_chunk_info["url"])) del meta["chunk_id"] del rebuild_chunk_info["dl_meta"]["chunk_id"] self.assertEqual(meta, rebuild_chunk_info["dl_meta"]) def test_2copies_content_0_byte_1broken_rebuild_pos_0_idx_0(self): self._test_rebuild("TWOCOPIES", 0, [(0, 0)], (0, 0)) def test_2copies_content_1_byte_1broken_rebuild_pos_0_idx_1(self): self._test_rebuild("TWOCOPIES", 1, [(0, 1)], (0, 1)) def test_3copies_content_chunksize_bytes_2broken_rebuild_pos_0_idx_1(self): self._test_rebuild("THREECOPIES", self.chunk_size, [(0, 0), (0, 1)], (0, 1)) def test_3copies_content_2xchksize_bytes_2broken_rebuild_pos_1_idx_2(self): self._test_rebuild("THREECOPIES", 2 * self.chunk_size, [(1, 0), (1, 2)], (1, 2)) def test_2copies_content_0_byte_2broken_rebuild_pos_0_idx_0(self): with ExpectedException(UnrecoverableContent): self._test_rebuild("TWOCOPIES", 0, [(0, 0), (0, 1)], (0, 0)) def _test_download(self, stgpol, data_size, broken_pos_list): data = random_data(data_size) content, _ = self._new_content(stgpol, data, broken_pos_list) downloaded_data = "".join(content.download()) self.assertEqual(downloaded_data, data) for pos, idx in broken_pos_list: # check nothing has been rebuilt c = content.chunks.filter(pos=pos)[0] self.assertRaises(NotFound, self.blob_client.chunk_delete, c.url) def test_twocopies_download_content_0_byte_without_broken_chunks(self): self._test_download("TWOCOPIES", 0, []) def test_twocopies_download_content_0_byte_with_broken_0_0(self): self._test_download("TWOCOPIES", 0, [(0, 0)]) def test_twocopies_download_content_1_byte_without_broken_chunks(self): self._test_download("TWOCOPIES", 1, []) def test_twocopies_download_content_1_byte_with_broken_0_0(self): self._test_download("TWOCOPIES", 1, [(0, 0)]) def test_twocopies_download_chunksize_bytes_without_broken_chunks(self): self._test_download("TWOCOPIES", self.chunk_size, []) def test_twocopies_download_2xchuksize_bytes_with_broken_0_0_and_1_0(self): self._test_download("TWOCOPIES", self.chunk_size * 2, [(0, 0), (1, 0)]) def test_twocopies_download_content_chunksize_bytes_2_broken_chunks(self): data = random_data(self.chunk_size) content, _ = self._new_content("TWOCOPIES", data, [(0, 0), (0, 1)]) gen = content.download() self.assertRaises(UnrecoverableContent, gen.next) def test_single_download_content_1_byte_without_broken_chunks(self): self._test_download("SINGLE", 1, []) def test_single_download_chunksize_bytes_plus_1_without_broken_chunk(self): self._test_download("SINGLE", self.chunk_size * 2, [])
class BlobMoverWorker(object): def __init__(self, conf, logger, volume): self.conf = conf self.logger = logger or get_logger(conf) self.volume = volume self.run_time = 0 self.passes = 0 self.errors = 0 self.last_reported = 0 self.last_usage_check = 0 self.chunks_run_time = 0 self.bytes_running_time = 0 self.bytes_processed = 0 self.total_bytes_processed = 0 self.total_chunks_processed = 0 self.usage_target = int_value( conf.get('usage_target'), 0) self.usage_check_interval = int_value( conf.get('usage_check_interval'), 3600) self.report_interval = int_value( conf.get('report_interval'), 3600) self.max_chunks_per_second = int_value( conf.get('chunks_per_second'), 30) self.max_bytes_per_second = int_value( conf.get('bytes_per_second'), 10000000) self.blob_client = BlobClient() self.container_client = ContainerClient(conf) def mover_pass(self): self.namespace, self.address = check_volume(self.volume) start_time = report_time = time.time() total_errors = 0 mover_time = 0 paths = paths_gen(self.volume) for path in paths: loop_time = time.time() now = time.time() if now - self.last_usage_check >= self.usage_check_interval: used, total = statfs(self.volume) usage = (float(used) / total) * 100 if usage <= self.usage_target: self.logger.info( 'current usage %.2f%%: target reached (%.2f%%)', usage, self.usage_target) self.last_usage_check = now break self.safe_chunk_move(path) self.chunks_run_time = ratelimit( self.chunks_run_time, self.max_chunks_per_second ) self.total_chunks_processed += 1 now = time.time() if now - self.last_reported >= self.report_interval: self.logger.info( '%(start_time)s ' '%(passes)d ' '%(errors)d ' '%(c_rate).2f ' '%(b_rate).2f ' '%(total).2f ' '%(mover_time).2f' '%(mover_rate).2f' % { 'start_time': time.ctime(report_time), 'passes': self.passes, 'errors': self.errors, 'c_rate': self.passes / (now - report_time), 'b_rate': self.bytes_processed / (now - report_time), 'total': (now - start_time), 'mover_time': mover_time, 'mover_rate': mover_time / (now - start_time) } ) report_time = now total_errors += self.errors self.passes = 0 self.bytes_processed = 0 self.last_reported = now mover_time += (now - loop_time) elapsed = (time.time() - start_time) or 0.000001 self.logger.info( '%(elapsed).02f ' '%(errors)d ' '%(chunk_rate).2f ' '%(bytes_rate).2f ' '%(mover_time).2f ' '%(mover_rate).2f' % { 'elapsed': elapsed, 'errors': total_errors + self.errors, 'chunk_rate': self.total_chunks_processed / elapsed, 'bytes_rate': self.total_bytes_processed / elapsed, 'mover_time': mover_time, 'mover_rate': mover_time / elapsed } ) def safe_chunk_move(self, path): try: self.chunk_move(path) except Exception as e: self.errors += 1 self.logger.error('ERROR while moving chunk %s: %s', path, e) self.passes += 1 def load_chunk_metadata(self, path): with open(path) as f: return read_chunk_metadata(f) def chunk_move(self, path): meta = self.load_chunk_metadata(path) content_cid = meta['content_cid'] content_path = meta['content_path'] chunk_url = 'http://%s/%s' % \ (self.address, meta['chunk_id']) try: data = self.container_client.content_show( cid=content_cid, path=content_path) except exc.NotFound: raise exc.OrphanChunk('Content not found') current_chunk = None notin = [] for c in data: if c['pos'] == meta['chunk_pos']: notin.append(c) for c in notin: if c['url'] == chunk_url: current_chunk = c notin.remove(c) if not current_chunk: raise exc.OrphanChunk('Chunk not found in content') spare_data = {'notin': notin, 'broken': [current_chunk], 'size': 0} spare_resp = self.container_client.content_spare( cid=content_cid, path=content_path, data=spare_data) new_chunk = spare_resp['chunks'][0] self.blob_client.chunk_copy( current_chunk['url'], new_chunk['id']) old = [{'type': 'chunk', 'id': current_chunk['url'], 'hash': meta['chunk_hash'], 'size': int(meta['chunk_size'])}] new = [{'type': 'chunk', 'id': new_chunk['id'], 'hash': meta['chunk_hash'], 'size': int(meta['chunk_size'])}] update_data = {'old': old, 'new': new} self.container_client.container_raw_update( cid=content_cid, data=update_data) self.blob_client.chunk_delete(current_chunk['url']) self.logger.info( 'moved chunk %s to %s', current_chunk['url'], new_chunk['id'])
class BlobMoverWorker(object): def __init__(self, conf, logger, volume): self.conf = conf self.logger = logger or get_logger(conf) self.volume = volume self.run_time = 0 self.passes = 0 self.errors = 0 self.last_reported = 0 self.last_usage_check = 0 self.chunks_run_time = 0 self.bytes_running_time = 0 self.bytes_processed = 0 self.total_bytes_processed = 0 self.total_chunks_processed = 0 self.usage_target = int_value(conf.get('usage_target'), 0) self.usage_check_interval = int_value(conf.get('usage_check_interval'), 3600) self.report_interval = int_value(conf.get('report_interval'), 3600) self.max_chunks_per_second = int_value(conf.get('chunks_per_second'), 30) self.max_bytes_per_second = int_value(conf.get('bytes_per_second'), 10000000) self.blob_client = BlobClient() self.container_client = ContainerClient(conf) def mover_pass(self): self.namespace, self.address = check_volume(self.volume) start_time = report_time = time.time() total_errors = 0 mover_time = 0 paths = paths_gen(self.volume) for path in paths: loop_time = time.time() now = time.time() if now - self.last_usage_check >= self.usage_check_interval: used, total = statfs(self.volume) usage = (float(used) / total) * 100 if usage <= self.usage_target: self.logger.info( 'current usage %.2f%%: target reached (%.2f%%)', usage, self.usage_target) self.last_usage_check = now break self.safe_chunk_move(path) self.chunks_run_time = ratelimit(self.chunks_run_time, self.max_chunks_per_second) self.total_chunks_processed += 1 now = time.time() if now - self.last_reported >= self.report_interval: self.logger.info( '%(start_time)s ' '%(passes)d ' '%(errors)d ' '%(c_rate).2f ' '%(b_rate).2f ' '%(total).2f ' '%(mover_time).2f' '%(mover_rate).2f' % { 'start_time': time.ctime(report_time), 'passes': self.passes, 'errors': self.errors, 'c_rate': self.passes / (now - report_time), 'b_rate': self.bytes_processed / (now - report_time), 'total': (now - start_time), 'mover_time': mover_time, 'mover_rate': mover_time / (now - start_time) }) report_time = now total_errors += self.errors self.passes = 0 self.bytes_processed = 0 self.last_reported = now mover_time += (now - loop_time) elapsed = (time.time() - start_time) or 0.000001 self.logger.info( '%(elapsed).02f ' '%(errors)d ' '%(chunk_rate).2f ' '%(bytes_rate).2f ' '%(mover_time).2f ' '%(mover_rate).2f' % { 'elapsed': elapsed, 'errors': total_errors + self.errors, 'chunk_rate': self.total_chunks_processed / elapsed, 'bytes_rate': self.total_bytes_processed / elapsed, 'mover_time': mover_time, 'mover_rate': mover_time / elapsed }) def safe_chunk_move(self, path): try: self.chunk_move(path) except Exception as e: self.errors += 1 self.logger.error('ERROR while moving chunk %s: %s', path, e) self.passes += 1 def load_chunk_metadata(self, path): with open(path) as f: return read_chunk_metadata(f) def chunk_move(self, path): meta = self.load_chunk_metadata(path) content_cid = meta['content_cid'] content_path = meta['content_path'] chunk_url = 'http://%s/%s' % \ (self.address, meta['chunk_id']) try: _, data = self.container_client.content_show(cid=content_cid, path=content_path) except exc.NotFound: raise exc.OrphanChunk('Content not found') current_chunk = None notin = [] for c in data: if c['pos'] == meta['chunk_pos']: notin.append(c) for c in notin: if c['url'] == chunk_url: current_chunk = c notin.remove(c) if not current_chunk: raise exc.OrphanChunk('Chunk not found in content') spare_data = {'notin': notin, 'broken': [current_chunk], 'size': 0} spare_resp = self.container_client.content_spare(cid=content_cid, path=content_path, data=spare_data) new_chunk = spare_resp['chunks'][0] self.blob_client.chunk_copy(current_chunk['url'], new_chunk['id']) old = [{ 'type': 'chunk', 'id': current_chunk['url'], 'hash': meta['chunk_hash'], 'size': int(meta['chunk_size']) }] new = [{ 'type': 'chunk', 'id': new_chunk['id'], 'hash': meta['chunk_hash'], 'size': int(meta['chunk_size']) }] update_data = {'old': old, 'new': new} self.container_client.container_raw_update(cid=content_cid, data=update_data) self.blob_client.chunk_delete(current_chunk['url']) self.logger.info('moved chunk %s to %s', current_chunk['url'], new_chunk['id'])
class TestPlainContent(BaseTestCase): def setUp(self): super(TestPlainContent, self).setUp() if len(self.conf['services']['rawx']) < 4: self.skipTest("Plain tests needs more than 3 rawx to run") self.namespace = self.conf['namespace'] self.account = self.conf['account'] self.chunk_size = self.conf['chunk_size'] self.gridconf = {"namespace": self.namespace} self.content_factory = ContentFactory(self.gridconf) self.container_client = ContainerClient(self.gridconf) self.blob_client = BlobClient() self.container_name = "TestPlainContent-%f" % time.time() self.container_client.container_create(acct=self.account, ref=self.container_name) self.container_id = cid_from_name(self.account, self.container_name).upper() self.content = random_str(64) self.stgpol = "SINGLE" self.stgpol_twocopies = "TWOCOPIES" self.stgpol_threecopies = "THREECOPIES" def _test_create(self, stgpol, data_size): data = random_data(data_size) content = self.content_factory.new(self.container_id, self.content, len(data), stgpol) content.create(StringIO(data)) meta, chunks = self.container_client.content_show( cid=self.container_id, content=content.content_id) self.assertEqual(meta['hash'], md5_data(data)) self.assertEqual(meta['length'], str(len(data))) self.assertEqual(meta['policy'], stgpol) self.assertEqual(meta['name'], self.content) metachunk_nb = int(math.ceil(float(len(data)) / self.chunk_size)) if metachunk_nb == 0: metachunk_nb = 1 # special case for empty content chunks = ChunksHelper(chunks) # TODO NO NO NO if stgpol == self.stgpol_threecopies: nb_copy = 3 elif stgpol == self.stgpol_twocopies: nb_copy = 2 elif stgpol == self.stgpol: nb_copy = 1 self.assertEqual(len(chunks), metachunk_nb * nb_copy) for pos in range(metachunk_nb): chunks_at_pos = chunks.filter(pos=pos) self.assertEqual(len(chunks_at_pos), nb_copy) data_begin = pos * self.chunk_size data_end = pos * self.chunk_size + self.chunk_size chunk_hash = md5_data(data[data_begin:data_end]) for chunk in chunks_at_pos: meta, stream = self.blob_client.chunk_get(chunk.url) self.assertEqual(md5_stream(stream), chunk_hash) self.assertEqual(meta['content_path'], self.content) self.assertEqual(meta['container_id'], self.container_id) self.assertEqual(meta['content_id'], meta['content_id']) self.assertEqual(meta['chunk_id'], chunk.id) self.assertEqual(meta['chunk_pos'], str(pos)) self.assertEqual(meta['chunk_hash'], chunk_hash) def test_twocopies_create_0_byte(self): self._test_create(self.stgpol_twocopies, 0) def test_twocopies_create_1_byte(self): self._test_create(self.stgpol_twocopies, 1) def test_twocopies_create_chunksize_bytes(self): self._test_create(self.stgpol_twocopies, self.chunk_size) def test_twocopies_create_chunksize_plus_1_bytes(self): self._test_create(self.stgpol_twocopies, self.chunk_size + 1) def test_single_create_0_byte(self): self._test_create(self.stgpol, 0) def test_single_create_chunksize_plus_1_bytes(self): self._test_create(self.stgpol, self.chunk_size + 1) def _new_content(self, stgpol, data, broken_pos_list=[]): old_content = self.content_factory.new(self.container_id, self.content, len(data), stgpol) old_content.create(StringIO(data)) broken_chunks_info = {} for pos, idx in broken_pos_list: c = old_content.chunks.filter(pos=pos)[idx] meta, stream = self.blob_client.chunk_get(c.url) if pos not in broken_chunks_info: broken_chunks_info[pos] = {} broken_chunks_info[pos][idx] = { "url": c.url, "id": c.id, "hash": c.checksum, "dl_meta": meta, "dl_hash": md5_stream(stream) } self.blob_client.chunk_delete(c.url) # get the new structure of the uploaded content return (self.content_factory.get(self.container_id, old_content.content_id), broken_chunks_info) def _test_rebuild(self, stgpol, data_size, broken_pos_list, full_rebuild_pos): data = random_data(data_size) content, broken_chunks_info = self._new_content( stgpol, data, broken_pos_list) rebuild_pos, rebuild_idx = full_rebuild_pos rebuild_chunk_info = broken_chunks_info[rebuild_pos][rebuild_idx] content.rebuild_chunk(rebuild_chunk_info["id"]) # get the new structure of the content rebuilt_content = self.content_factory.get(self.container_id, content.content_id) # find the rebuilt chunk for c in rebuilt_content.chunks.filter(pos=rebuild_pos): if len(content.chunks.filter(id=c.id)) > 0: # not the rebuilt chunk # if this chunk is broken, it must not have been rebuilt for b_c_i in broken_chunks_info[rebuild_pos].values(): if c.id == b_c_i["id"]: with ExpectedException(NotFound): _, _ = self.blob_client.chunk_get(c.url) continue meta, stream = self.blob_client.chunk_get(c.url) self.assertEqual(meta["chunk_id"], c.id) self.assertEqual(md5_stream(stream), rebuild_chunk_info["dl_hash"]) self.assertEqual(c.checksum, rebuild_chunk_info["hash"]) self.assertThat(c.url, NotEquals(rebuild_chunk_info["url"])) del meta["chunk_id"] del rebuild_chunk_info["dl_meta"]["chunk_id"] self.assertEqual(meta, rebuild_chunk_info["dl_meta"]) def test_2copies_content_0_byte_1broken_rebuild_pos_0_idx_0(self): self._test_rebuild(self.stgpol_twocopies, 0, [(0, 0)], (0, 0)) def test_2copies_content_1_byte_1broken_rebuild_pos_0_idx_1(self): self._test_rebuild(self.stgpol_twocopies, 1, [(0, 1)], (0, 1)) def test_3copies_content_chunksize_bytes_2broken_rebuild_pos_0_idx_1(self): if len(self.conf['services']['rawx']) <= 3: self.skipTest("Need more than 3 rawx") self._test_rebuild(self.stgpol_threecopies, self.chunk_size, [(0, 0), (0, 1)], (0, 1)) def test_3copies_content_2xchksize_bytes_2broken_rebuild_pos_1_idx_2(self): self._test_rebuild(self.stgpol_threecopies, 2 * self.chunk_size, [(1, 0), (1, 2)], (1, 2)) def test_2copies_content_0_byte_2broken_rebuild_pos_0_idx_0(self): with ExpectedException(UnrecoverableContent): self._test_rebuild(self.stgpol_twocopies, 0, [(0, 0), (0, 1)], (0, 0)) def _test_fetch(self, stgpol, data_size, broken_pos_list): data = random_data(data_size) content, _ = self._new_content(stgpol, data, broken_pos_list) fetched_data = "".join(content.fetch()) self.assertEqual(fetched_data, data) for pos, idx in broken_pos_list: # check nothing has been rebuilt c = content.chunks.filter(pos=pos)[0] self.assertRaises(NotFound, self.blob_client.chunk_delete, c.url) def test_twocopies_fetch_content_0_byte_without_broken_chunks(self): self._test_fetch(self.stgpol_twocopies, 0, []) def test_twocopies_fetch_content_0_byte_with_broken_0_0(self): self._test_fetch(self.stgpol_twocopies, 0, [(0, 0)]) def test_twocopies_fetch_content_1_byte_without_broken_chunks(self): self._test_fetch(self.stgpol_twocopies, 1, []) def test_twocopies_fetch_content_1_byte_with_broken_0_0(self): self._test_fetch(self.stgpol_twocopies, 1, [(0, 0)]) def test_twocopies_fetch_chunksize_bytes_without_broken_chunks(self): self._test_fetch(self.stgpol_twocopies, self.chunk_size, []) def test_twocopies_fetch_2xchuksize_bytes_with_broken_0_0_and_1_0(self): self._test_fetch(self.stgpol_twocopies, self.chunk_size * 2, [(0, 0), (1, 0)]) def test_twocopies_fetch_content_chunksize_bytes_2_broken_chunks(self): data = random_data(self.chunk_size) content, _ = self._new_content(self.stgpol_twocopies, data, [(0, 0), (0, 1)]) gen = content.fetch() self.assertRaises(UnrecoverableContent, gen.next) def test_single_fetch_content_1_byte_without_broken_chunks(self): self._test_fetch(self.stgpol, 1, []) def test_single_fetch_chunksize_bytes_plus_1_without_broken_chunk(self): self._test_fetch(self.stgpol, self.chunk_size * 2, [])
class Checker(object): def __init__(self, namespace, concurrency=50, error_file=None, rebuild_file=None): self.pool = GreenPool(concurrency) self.error_file = error_file if self.error_file: f = open(self.error_file, 'a') self.error_writer = csv.writer(f, delimiter=' ') self.rebuild_file = rebuild_file if self.rebuild_file: fd = open(self.rebuild_file, 'a') self.rebuild_writer = csv.writer(fd, delimiter='|') conf = {'namespace': namespace} self.account_client = AccountClient(conf) self.container_client = ContainerClient(conf) self.blob_client = BlobClient() 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 = {} def write_error(self, target): error = [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, obj_meta, ct_meta): cid = ct_meta['properties']['sys.name'].split('.', 1)[0] self.rebuild_writer.writerow((cid, obj_meta['id'], target.chunk)) def _check_chunk_xattr(self, target, obj_meta, xattr_meta): error = False # 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): print(" Chunk %s '%s' xattr (%s) " "differs from size in meta2 (%s)" % (target, attr_key, xattr_meta.get(attr_key), obj_meta['size'])) error = True attr_key = attr_prefix + 'chunk_hash' if obj_meta['hash'] != xattr_meta.get(attr_key): print(" Chunk %s '%s' xattr (%s) " "differs from hash in meta2 (%s)" % (target, attr_key, xattr_meta.get(attr_key), obj_meta['hash'])) error = True return error def check_chunk(self, target): chunk = target.chunk obj_listing, obj_meta = self.check_obj(target) error = False if chunk not in obj_listing: print(' Chunk %s missing from object listing' % target) error = True db_meta = dict() else: db_meta = obj_listing[chunk] try: xattr_meta = self.blob_client.chunk_head(chunk) except exc.NotFound as e: self.chunk_not_found += 1 error = True print(' Not found chunk "%s": %s' % (target, str(e))) except Exception as e: self.chunk_exceptions += 1 error = True print(' Exception chunk "%s": %s' % (target, str(e))) else: if db_meta: error = self._check_chunk_xattr(target, db_meta, xattr_meta) if error: if self.error_file: self.write_error(target) if self.rebuild_file: self.write_rebuilder_input( target, obj_meta, self.list_cache[(target.account, target.container)][1]) self.chunks_checked += 1 def check_obj(self, target, recurse=False): account = target.account container = target.container obj = target.obj if (account, container, obj) in self.running: self.running[(account, container, obj)].wait() if (account, container, obj) in self.list_cache: return self.list_cache[(account, container, obj)] self.running[(account, container, obj)] = Event() print('Checking object "%s"' % target) container_listing, ct_meta = self.check_container(target) error = False if obj not in container_listing: print(' Object %s missing from container listing' % target) error = True # checksum = None else: # TODO check checksum match # checksum = container_listing[obj]['hash'] pass results = [] meta = dict() try: meta, results = self.container_client.content_show( acct=account, ref=container, path=obj) except exc.NotFound as e: self.object_not_found += 1 error = True print(' Not found object "%s": %s' % (target, str(e))) except Exception as e: self.object_exceptions += 1 error = True print(' Exception object "%s": %s' % (target, str(e))) chunk_listing = dict() for chunk in results: chunk_listing[chunk['url']] = chunk self.objects_checked += 1 self.list_cache[(account, container, obj)] = (chunk_listing, meta) self.running[(account, container, obj)].send(True) del self.running[(account, container, obj)] if recurse: for chunk in chunk_listing: t = target.copy() t.chunk = chunk self.pool.spawn_n(self.check_chunk, t) if error and self.error_file: self.write_error(target) return chunk_listing, meta def check_container(self, target, recurse=False): account = target.account container = target.container if (account, container) in self.running: self.running[(account, container)].wait() if (account, container) in self.list_cache: return self.list_cache[(account, container)] self.running[(account, container)] = Event() print('Checking container "%s"' % target) account_listing = self.check_account(target) error = False if container not in account_listing: error = True print(' Container %s missing from account listing' % target) marker = None results = [] ct_meta = dict() while True: try: resp = self.container_client.container_list( acct=account, ref=container, marker=marker) except exc.NotFound as e: self.container_not_found += 1 error = True print(' Not found container "%s": %s' % (target, str(e))) break except Exception as e: self.container_exceptions += 1 error = True print(' Exception container "%s": %s' % (target, str(e))) break if resp['objects']: marker = resp['objects'][-1]['name'] results.extend(resp['objects']) else: ct_meta = resp ct_meta.pop('objects') break container_listing = dict() for obj in results: container_listing[obj['name']] = obj self.containers_checked += 1 self.list_cache[(account, container)] = container_listing, ct_meta self.running[(account, container)].send(True) del self.running[(account, container)] if recurse: for obj in container_listing: t = target.copy() t.obj = obj self.pool.spawn_n(self.check_obj, t, True) if error and self.error_file: self.write_error(target) return container_listing, ct_meta def check_account(self, target, recurse=False): account = target.account if account in self.running: self.running[account].wait() if account in self.list_cache: return self.list_cache[account] self.running[account] = Event() print('Checking account "%s"' % target) error = False marker = None results = [] while True: try: resp = self.account_client.containers_list( account, marker=marker) except Exception as e: self.account_exceptions += 1 error = True print(' Exception account "%s": %s' % (target, str(e))) break if resp['listing']: marker = resp['listing'][-1][0] else: break results.extend(resp['listing']) containers = dict() for e in results: containers[e[0]] = (e[1], e[2]) self.list_cache[account] = containers self.running[account].send(True) del self.running[account] self.accounts_checked += 1 if recurse: for container in containers: t = target.copy() t.container = container self.pool.spawn_n(self.check_container, t, True) if error and self.error_file: self.write_error(target) return containers def check(self, target): if target.chunk and target.obj and target.container: self.pool.spawn_n(self.check_chunk, target) elif target.obj and target.container: self.pool.spawn_n(self.check_obj, target, True) elif target.container: self.pool.spawn_n(self.check_container, target, True) else: self.pool.spawn_n(self.check_account, target, True) def wait(self): self.pool.waitall() def report(self): 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: _report_stat("Missing accounts", self.account_not_found) if self.account_exceptions: _report_stat("Exceptions", self.account_not_found) print() _report_stat("Containers checked", self.containers_checked) if self.container_not_found: _report_stat("Missing containers", self.container_not_found) if self.container_exceptions: _report_stat("Exceptions", self.container_exceptions) print() _report_stat("Objects checked", self.objects_checked) if self.object_not_found: _report_stat("Missing objects", self.object_not_found) if self.object_exceptions: _report_stat("Exceptions", self.object_exceptions) print() _report_stat("Chunks checked", self.chunks_checked) if self.chunk_not_found: _report_stat("Missing chunks", self.chunk_not_found) if self.chunk_exceptions: _report_stat("Exceptions", self.chunk_exceptions)
class TestRebuilderCrawler(BaseTestCase): def setUp(self): super(TestRebuilderCrawler, self).setUp() self.namespace = self.conf['namespace'] self.account = self.conf['account'] self.gridconf = {"namespace": self.namespace} self.container_client = ContainerClient(self.gridconf) self.blob_client = BlobClient() self.container_name = "TestRebuilderCrawler%d" % int(time.time()) self.container_client.container_create(acct=self.account, ref=self.container_name) def _push_content(self, content): for c in content.chunks: self.blob_client.chunk_put(c.url, c.get_create_xattr(), c.data) self.container_client.content_create(acct=content.account, ref=content.container_name, path=content.content_name, size=content.size, checksum=content.hash, content_id=content.content_id, stgpol=content.stgpol, data=content.get_create_meta2()) def tearDown(self): super(TestRebuilderCrawler, self).tearDown() def test_rebuild_chunk(self): # push a new content content = TestContent(self.conf, self.account, self.container_name, "mycontent", "TWOCOPIES") data = "azerty" content.add_chunk(data, pos='0', rawx=0) content.add_chunk(data, pos='0', rawx=1) self._push_content(content) # rebuild the first rawx rebuilder = BlobRebuilderWorker(self.gridconf, None, self.conf['rawx'][0]['addr']) rebuilder.chunk_rebuild(content.container_id, content.content_id, content.chunks[0].id) # check meta2 information _, res = self.container_client.content_show(acct=content.account, ref=content.container_name, content=content.content_id) new_chunk_info = None for c in res: if (c['url'] != content.chunks[0].url and c['url'] != content.chunks[1].url): new_chunk_info = c new_chunk_id = new_chunk_info['url'].split('/')[-1] self.assertEqual(new_chunk_info['hash'], content.chunks[0].hash) self.assertEqual(new_chunk_info['pos'], content.chunks[0].pos) self.assertEqual(new_chunk_info['size'], content.chunks[0].size) # check chunk information meta, stream = self.blob_client.chunk_get(new_chunk_info['url']) self.assertEqual(meta['content_size'], str(content.chunks[0].size)) self.assertEqual(meta['content_path'], content.content_name) self.assertEqual(meta['content_cid'], content.container_id) self.assertEqual(meta['content_id'], content.content_id) self.assertEqual(meta['chunk_id'], new_chunk_id) self.assertEqual(meta['chunk_pos'], content.chunks[0].pos) self.assertEqual(meta['content_version'], content.version) self.assertEqual(meta['chunk_hash'], content.chunks[0].hash) self.assertEqual(stream.next(), content.chunks[0].data) # check rtime flag in rdir rdir_client = RdirClient(self.gridconf) res = rdir_client.chunk_fetch(self.conf['rawx'][0]['addr']) key = (content.container_id, content.content_id, content.chunks[0].id) for i_container, i_content, i_chunk, i_value in res: if (i_container, i_content, i_chunk) == key: check_value = i_value self.assertIsNotNone(check_value.get('rtime')) @unittest.skipIf( len(get_config()['rawx']) != 3, "The number of rawx must be 3") def test_rebuild_no_spare(self): # push a new content content = TestContent(self.conf, self.account, self.container_name, "mycontent", "THREECOPIES") data = "azerty" content.add_chunk(data, pos='0', rawx=0) content.add_chunk(data, pos='0', rawx=1) content.add_chunk(data, pos='0', rawx=2) self._push_content(content) # rebuild the first rawx rebuilder = BlobRebuilderWorker(self.gridconf, None, self.conf['rawx'][0]['addr']) self.assertRaises(SpareChunkException, rebuilder.chunk_rebuild, content.container_id, content.content_id, content.chunks[0].id) def test_rebuild_upload_failed(self): # push a new content content = TestContent(self.conf, self.account, self.container_name, "mycontent", "TWOCOPIES") data = "azerty" content.add_chunk(data, pos='0', rawx=0) content.add_chunk(data, pos='0', rawx=1) self._push_content(content) # rebuild the first rawx rebuilder = BlobRebuilderWorker(self.gridconf, None, self.conf['rawx'][0]['addr']) # Force upload to raise an exception with patch('oio.content.content.BlobClient') as MockClass: instance = MockClass.return_value instance.chunk_copy.side_effect = Exception("xx") self.assertRaises(UnrecoverableContent, rebuilder.chunk_rebuild, content.container_id, content.content_id, content.chunks[0].id) def test_rebuild_nonexistent_chunk(self): rebuilder = BlobRebuilderWorker(self.gridconf, None, self.conf['rawx'][0]['addr']) # try to rebuild an nonexistant chunk self.assertRaises(OrphanChunk, rebuilder.chunk_rebuild, 64 * '0', 32 * '0', 64 * '0') def test_rebuild_orphan_chunk(self): # push a new content content = TestContent(self.conf, self.account, self.container_name, "mycontent", "TWOCOPIES") data = "azerty" content.add_chunk(data, pos='0', rawx=0) content.add_chunk(data, pos='0', rawx=1) self._push_content(content) # rebuild the first rawx rebuilder = BlobRebuilderWorker(self.gridconf, None, self.conf['rawx'][0]['addr']) # try to rebuild an nonexistant chunk self.assertRaises(OrphanChunk, rebuilder.chunk_rebuild, content.container_id, content.content_id, 64 * '0') def test_rebuild_with_no_copy(self): # push a new content content = TestContent(self.conf, self.account, self.container_name, "mycontent", "SINGLE") data = "azerty" content.add_chunk(data, pos='0', rawx=0) self._push_content(content) # rebuild the first rawx rebuilder = BlobRebuilderWorker(self.gridconf, None, self.conf['rawx'][0]['addr']) # try to rebuild chunk without copy self.assertRaises(UnrecoverableContent, rebuilder.chunk_rebuild, content.container_id, content.content_id, content.chunks[0].id)
class BlobAuditorWorker(object): def __init__(self, conf, logger, volume): self.conf = conf self.logger = logger self.volume = volume self.run_time = 0 self.passes = 0 self.errors = 0 self.orphan_chunks = 0 self.faulty_chunks = 0 self.corrupted_chunks = 0 self.last_reported = 0 self.chunks_run_time = 0 self.bytes_running_time = 0 self.bytes_processed = 0 self.total_bytes_processed = 0 self.total_chunks_processed = 0 self.report_interval = int_value( conf.get('report_interval'), 3600) self.max_chunks_per_second = int_value( conf.get('chunks_per_second'), 30) self.max_bytes_per_second = int_value( conf.get('bytes_per_second'), 10000000) self.container_client = ContainerClient(conf) def audit_pass(self): self.namespace, self.address = check_volume(self.volume) start_time = report_time = time.time() total_errors = 0 total_corrupted = 0 total_orphans = 0 total_faulty = 0 audit_time = 0 paths = paths_gen(self.volume) for path in paths: loop_time = time.time() self.safe_chunk_audit(path) self.chunks_run_time = ratelimit( self.chunks_run_time, self.max_chunks_per_second ) self.total_chunks_processed += 1 now = time.time() if now - self.last_reported >= self.report_interval: self.logger.info( '%(start_time)s ' '%(passes)d ' '%(corrupted)d ' '%(faulty)d ' '%(orphans)d ' '%(errors)d ' '%(c_rate).2f ' '%(b_rate).2f ' '%(total).2f ' '%(audit_time).2f' '%(audit_rate).2f' % { 'start_time': time.ctime(report_time), 'passes': self.passes, 'corrupted': self.corrupted_chunks, 'faulty': self.faulty_chunks, 'orphans': self.orphan_chunks, 'errors': self.errors, 'c_rate': self.passes / (now - report_time), 'b_rate': self.bytes_processed / (now - report_time), 'total': (now - start_time), 'audit_time': audit_time, 'audit_rate': audit_time / (now - start_time) } ) report_time = now total_corrupted += self.corrupted_chunks total_orphans += self.orphan_chunks total_faulty += self.faulty_chunks total_errors += self.errors self.passes = 0 self.corrupted_chunks = 0 self.orphan_chunks = 0 self.faulty_chunks = 0 self.errors = 0 self.bytes_processed = 0 self.last_reported = now audit_time += (now - loop_time) elapsed = (time.time() - start_time) or 0.000001 self.logger.info( '%(elapsed).02f ' '%(corrupted)d ' '%(faulty)d ' '%(orphans)d ' '%(errors)d ' '%(chunk_rate).2f ' '%(bytes_rate).2f ' '%(audit_time).2f ' '%(audit_rate).2f' % { 'elapsed': elapsed, 'corrupted': total_corrupted + self.corrupted_chunks, 'faulty': total_faulty + self.faulty_chunks, 'orphans': total_orphans + self.orphan_chunks, 'errors': total_errors + self.errors, 'chunk_rate': self.total_chunks_processed / elapsed, 'bytes_rate': self.total_bytes_processed / elapsed, 'audit_time': audit_time, 'audit_rate': audit_time / elapsed } ) def safe_chunk_audit(self, path): try: self.chunk_audit(path) except exc.FaultyChunk as err: self.faulty_chunks += 1 self.logger.error('ERROR faulty chunk %s: %s', path, err) except exc.CorruptedChunk as err: self.corrupted_chunks += 1 self.logger.error('ERROR corrupted chunk %s: %s', path, err) except exc.OrphanChunk as err: self.orphan_chunks += 1 self.logger.error('ERROR orphan chunk %s: %s', path, err) except Exception: self.errors += 1 self.logger.exception('ERROR while auditing chunk %s', path) self.passes += 1 def chunk_audit(self, path): with open(path) as f: try: meta = read_chunk_metadata(f) except exc.MissingAttribute as e: raise exc.FaultyChunk( 'Missing extended attribute %s' % e) size = int(meta['chunk_size']) md5_checksum = meta['chunk_hash'].lower() reader = ChunkReader(f, size, md5_checksum) with closing(reader): for buf in reader: buf_len = len(buf) self.bytes_running_time = ratelimit( self.bytes_running_time, self.max_bytes_per_second, increment=buf_len) self.bytes_processed += buf_len self.total_bytes_processed += buf_len try: content_cid = meta['content_cid'] content_path = meta['content_path'] content_attr, data = self.container_client.content_show( cid=content_cid, path=content_path) # Check chunk data chunk_data = None metachunks = set() for c in data: if c['url'].endswith(meta['chunk_id']): metachunks.add(c['pos'].split('.', 2)[0]) chunk_data = c if not chunk_data: raise exc.OrphanChunk('Not found in content') if chunk_data['size'] != int(meta['chunk_size']): raise exc.FaultyChunk('Invalid chunk size found') if chunk_data['hash'] != meta['chunk_hash']: raise exc.FaultyChunk('Invalid chunk hash found') if chunk_data['pos'] != meta['chunk_pos']: raise exc.FaultyChunk('Invalid chunk position found') # Check content data if content_attr['length'] != meta['content_size']: raise exc.FaultyChunk('Invalid content size found') if len(metachunks) != int(meta['content_chunksnb']): self.logger.warn('Invalid number of chunks found') raise exc.FaultyChunk('Invalid number of chunks found') except exc.NotFound: raise exc.OrphanChunk('Chunk not found in container')
class ObjectStorageApi(object): """ The Object Storage API. High level API that wraps `AccountClient`, `ContainerClient` and `DirectoryClient` classes. """ def __init__(self, namespace, **kwargs): """ Initialize the object storage API. :param namespace: name of the namespace to interract with :type namespace: `str` :keyword connection_timeout: connection timeout towards rawx services :type connection_timeout: `float` seconds :keyword read_timeout: timeout for rawx responses and data reads from the caller (when uploading) :type read_timeout: `float` seconds :keyword write_timeout: timeout for rawx write requests :type write_timeout: `float` seconds """ self.namespace = namespace self.connection_timeout = utils.float_value( kwargs.get("connection_timeout"), None) self.read_timeout = utils.float_value(kwargs.get("read_timeout"), None) self.write_timeout = utils.float_value(kwargs.get("write_timeout"), None) # FIXME: share session between all the clients self.directory = DirectoryClient({"namespace": self.namespace}, **kwargs) self.account = AccountClient({"namespace": self.namespace}, **kwargs) self.container = ContainerClient({"namespace": self.namespace}, **kwargs) def account_create(self, account, headers=None): """ Create an account. :param account: name of the account to create :type account: `str` :returns: `True` if the account has been created """ return self.account.account_create(account, headers=headers) @handle_account_not_found def account_delete(self, account, headers=None): """ Delete an account. :param account: name of the account to delete :type account: `str` """ self.account.account_delete(account, headers=headers) @handle_account_not_found def account_show(self, account, headers=None): """ Get information about an account. """ return self.account.account_show(account, headers=headers) def account_list(self, headers=None): """ List accounts """ return self.account.account_list(headers=headers) # FIXME: @handle_account_not_found def account_update(self, account, metadata, to_delete=None, headers=None): self.account.account_update(account, metadata, to_delete, headers=headers) @handle_account_not_found def account_set_properties(self, account, properties, headers=None): self.account_update(account, properties, headers=headers) @handle_account_not_found def account_del_properties(self, account, properties, headers=None): self.account_update(account, None, properties, headers=headers) def container_create(self, account, container, properties=None, headers=None, **kwargs): """ Create a container. :param account: account in which to create the container :type account: `str` :param container: name of the container :type container: `str` :param properties: properties to set on the container :type properties: `dict` :keyword headers: extra headers to send to the proxy :type headers: `dict` :returns: True if the container has been created, False if it already exists """ return self.container.container_create(account, container, properties=properties, headers=headers, autocreate=True, **kwargs) @handle_container_not_found def container_touch(self, account, container, headers=None, **kwargs): """ Trigger a notification about the container state. :param account: account from which to delete the container :type account: `str` :param container: name of the container :type container: `str` :keyword headers: extra headers to send to the proxy :type headers: `dict` """ if not headers: headers = dict() if 'X-oio-req-id' not in headers: headers['X-oio-req-id'] = utils.request_id() self.container.container_touch(account, container, headers=headers, **kwargs) def container_create_many(self, account, containers, properties=None, headers=None, **kwargs): """ Create Many containers :param account: account in which to create the containers :type account: `str` :param containers: names of the containers :type containers: `list` :param properties: properties to set on the containers :type properties: `dict` :keyword headers: extra headers to send to the proxy :type headers: `dict` """ return self.container.container_create_many(account, containers, properties=properties, headers=headers, autocreate=True, **kwargs) @handle_container_not_found def container_delete(self, account, container, headers=None, **kwargs): """ Delete a container. :param account: account from which to delete the container :type account: `str` :param container: name of the container :type container: `str` :keyword headers: extra headers to send to the proxy :type headers: `dict` """ self.container.container_delete(account, container, headers=headers, **kwargs) @handle_account_not_found def container_list(self, account, limit=None, marker=None, end_marker=None, prefix=None, delimiter=None, headers=None): """ Get the list of containers of an account. :param account: account from which to get the container list :type account: `str` :keyword limit: maximum number of results to return :type limit: `int` :keyword marker: name of the container from where to start the listing :type marker: `str` :keyword end_marker: :keyword prefix: :keyword delimiter: :keyword headers: extra headers to send to the proxy :type headers: `dict` """ resp = self.account.container_list(account, limit=limit, marker=marker, end_marker=end_marker, prefix=prefix, delimiter=delimiter, headers=headers) return resp["listing"] @handle_container_not_found def container_show(self, account, container, headers=None): """ Get information about a container (user properties). :param account: account in which the container is :type account: `str` :param container: name of the container :type container: `str` :keyword headers: extra headers to send to the proxy :type headers: `dict` :returns: a `dict` with "properties" containing a `dict` of user properties. """ return self.container.container_show(account, container, headers=headers) @handle_container_not_found def container_get_properties(self, account, container, properties=None, headers=None): """ Get information about a container (user and system properties). :param account: account in which the container is :type account: `str` :param container: name of the container :type container: `str` :param properties: *ignored* :keyword headers: extra headers to send to the proxy :type headers: `dict` :returns: a `dict` with "properties" and "system" entries, containing respectively a `dict` of user properties and a `dict` of system properties. """ return self.container.container_get_properties(account, container, properties=properties, headers=headers) @handle_container_not_found def container_set_properties(self, account, container, properties=None, clear=False, headers=None, **kwargs): """ Set properties on a container. :param account: name of the account :type account: `str` :param container: name of the container where to set properties :type container: `str` :param properties: a dictionary of properties :type properties: `dict` :param clear: :type clear: `bool` :param headers: extra headers to pass to the proxy :type headers: `dict` :keyword system: dictionary of system properties to set """ return self.container.container_set_properties(account, container, properties, clear=clear, headers=headers, **kwargs) @handle_container_not_found def container_del_properties(self, account, container, properties, headers=None, **kwargs): return self.container.container_del_properties(account, container, properties, headers=headers, **kwargs) def container_update(self, account, container, metadata, clear=False, headers=None): if not metadata: self.container_del_properties(account, container, [], headers=headers) else: self.container_set_properties(account, container, metadata, clear, headers=headers) @handle_container_not_found def object_create(self, account, container, file_or_path=None, data=None, etag=None, obj_name=None, mime_type=None, metadata=None, policy=None, headers=None, key_file=None, **_kwargs): """ Create an object in *container* of *account* with data taken from either *data* (`str` or `generator`) or *file_or_path* (path to a file or file-like object). The object will be named after *obj_name* if specified, or after the base name of *file_or_path*. :param account: name of the account where to create the object :type account: `str` :param container: name of the container where to create the object :type container: `str` :param file_or_path: file-like object or path to a file from which to read object data :type file_or_path: `str` or file-like object :param data: object data (if `file_or_path` is not set) :type data: `str` or `generator` :keyword etag: entity tag of the object :type etag: `str` :keyword obj_name: name of the object to create. If not set, will use the base name of `file_or_path`. :keyword mime_type: MIME type of the object :type mime_type: `str` :keyword properties: a dictionary of properties :type properties: `dict` :keyword policy: name of the storage policy :type policy: `str` :param headers: extra headers to pass to the proxy :type headers: `dict` :keyword key_file: """ if (data, file_or_path) == (None, None): raise exc.MissingData() src = data if data is not None else file_or_path if src is file_or_path: if isinstance(file_or_path, basestring): if not os.path.exists(file_or_path): raise exc.FileNotFound("File '%s' not found." % file_or_path) file_name = os.path.basename(file_or_path) else: try: file_name = os.path.basename(file_or_path.name) except AttributeError: file_name = None obj_name = obj_name or file_name elif isgenerator(src): file_or_path = utils.GeneratorIO(src) src = file_or_path if not obj_name: raise exc.MissingName("No name for the object has been specified") sysmeta = {'mime_type': mime_type, 'etag': etag} if not headers: headers = dict() if 'X-oio-req-id' not in headers: headers['X-oio-req-id'] = utils.request_id() if src is data: return self._object_create(account, container, obj_name, BytesIO(data), sysmeta, properties=metadata, policy=policy, headers=headers, key_file=key_file) elif hasattr(file_or_path, "read"): return self._object_create(account, container, obj_name, src, sysmeta, properties=metadata, policy=policy, headers=headers, key_file=key_file) else: with open(file_or_path, "rb") as f: return self._object_create(account, container, obj_name, f, sysmeta, properties=metadata, policy=policy, headers=headers, key_file=key_file) def object_touch(self, account, container, obj, headers=None, **kwargs): """ Trigger a notification about an object (as if it just had been created). :param account: name of the account where to create the object :type account: `str` :param container: name of the container where to create the object :type container: `str` :param obj: name of the object to touch :param headers: extra headers to pass to the proxy """ if not headers: headers = dict() if 'X-oio-req-id' not in headers: headers['X-oio-req-id'] = utils.request_id() self.container.content_touch(account, container, obj, headers=headers, **kwargs) @handle_object_not_found def object_delete(self, account, container, obj, headers=None, **kwargs): if not headers: headers = dict() if 'X-oio-req-id' not in headers: headers['X-oio-req-id'] = utils.request_id() return self.container.content_delete(account, container, obj, headers=headers, **kwargs) def object_delete_many(self, account, container, objs, headers=None, **kwargs): if not headers: headers = dict() if 'X-oio-req-id' not in headers: headers['X-oio-req-id'] = utils.request_id() return self.container.content_delete_many(account, container, objs, headers=headers, **kwargs) @handle_container_not_found def object_list(self, account, container, limit=None, marker=None, delimiter=None, prefix=None, end_marker=None, headers=None, properties=False, **kwargs): """ Lists objects inside a container. :returns: a dict which contains * 'objects': the list of objects * 'prefixes': common prefixes (only if delimiter and prefix are set) * 'properties': a dict of container properties * 'system': system metadata """ _, resp_body = self.container.content_list(account, container, limit=limit, marker=marker, end_marker=end_marker, prefix=prefix, delimiter=delimiter, properties=properties, headers=headers, **kwargs) for obj in resp_body['objects']: mtype = obj.get('mime-type') if mtype: obj['mime_type'] = mtype del obj['mime-type'] return resp_body # FIXME: @handle_object_not_found def object_locate(self, account, container, obj, headers=None): obj_meta, body = self.container.content_locate(account, container, obj) return obj_meta, body def object_analyze(self, *args, **kwargs): """ :deprecated: use `object_locate` """ return self.object_locate(*args, **kwargs) def object_fetch(self, account, container, obj, ranges=None, headers=None, key_file=None): if not headers: headers = dict() if 'X-oio-req-id' not in headers: headers['X-oio-req-id'] = utils.request_id() meta, raw_chunks = self.object_locate(account, container, obj, headers=headers) chunk_method = meta['chunk_method'] storage_method = STORAGE_METHODS.load(chunk_method) chunks = _sort_chunks(raw_chunks, storage_method.ec) meta['container_id'] = utils.name2cid(account, container).upper() meta['ns'] = self.namespace if storage_method.ec: stream = self._fetch_stream_ec(meta, chunks, ranges, storage_method, headers) elif storage_method.backblaze: stream = self._fetch_stream_backblaze(meta, chunks, ranges, storage_method, key_file) else: stream = self._fetch_stream(meta, chunks, ranges, storage_method, headers) return meta, stream @handle_object_not_found def object_get_properties(self, account, container, obj, headers=None): return self.container.content_get_properties(account, container, obj) def object_show(self, account, container, obj, headers=None): """ Get a description of the content along with its user properties. :param account: name of the account in which the object is stored :param container: name of the container in which the object is stored :param obj: name of the object to query :returns: a `dict` describing the object .. python:: {'hash': '6BF60C17CC15EEA108024903B481738F', 'ctime': '1481031763', 'deleted': 'False', 'properties': { u'projet': u'OpenIO-SDS'}, 'length': '43518', 'hash_method': 'md5', 'chunk_method': 'ec/algo=liberasurecode_rs_vand,k=6,m=3', 'version': '1481031762951972', 'policy': 'EC', 'id': '20BF2194FD420500CD4729AE0B5CBC07', 'mime_type': 'application/octet-stream', 'name': 'Makefile'} """ return self.container.content_show(account, container, obj, headers=headers) def object_update(self, account, container, obj, metadata, clear=False, headers=None): if clear: self.object_del_properties(account, container, obj, [], headers=headers) if metadata: self.object_set_properties(account, container, obj, metadata, headers=headers) @handle_object_not_found def object_set_properties(self, account, container, obj, properties, clear=False, headers=None, **kwargs): return self.container.content_set_properties( account, container, obj, properties={'properties': properties}, headers=headers, **kwargs) @handle_object_not_found def object_del_properties(self, account, container, obj, properties, headers=None, **kwargs): return self.container.content_del_properties(account, container, obj, properties=properties, headers=headers, **kwargs) # FIXME: remove and call self.container.content_prepare() directly def _content_prepare(self, account, container, obj_name, size, policy=None, headers=None): return self.container.content_prepare(account, container, obj_name, size, stgpol=policy, autocreate=True, headers=headers) def _content_preparer(self, account, container, obj_name, policy=None, headers=None): # TODO: optimize by asking more than one metachunk at a time obj_meta, first_body = self.container.content_prepare(account, container, obj_name, size=1, stgpol=policy, autocreate=True, headers=headers) storage_method = STORAGE_METHODS.load(obj_meta['chunk_method']) def _fix_mc_pos(chunks, mc_pos): for chunk in chunks: raw_pos = chunk["pos"].split(".") if storage_method.ec: chunk['num'] = int(raw_pos[1]) chunk["pos"] = "%d.%d" % (mc_pos, chunk['num']) else: chunk["pos"] = str(mc_pos) def _metachunk_preparer(): mc_pos = 0 _fix_mc_pos(first_body, mc_pos) yield first_body while True: mc_pos += 1 _, next_body = self._content_prepare(account, container, obj_name, 1, policy, headers) _fix_mc_pos(next_body, mc_pos) yield next_body return obj_meta, _metachunk_preparer def _object_create(self, account, container, obj_name, source, sysmeta, properties=None, policy=None, headers=None, key_file=None): obj_meta, chunk_prep = self._content_preparer(account, container, obj_name, policy=policy, headers=headers) obj_meta.update(sysmeta) obj_meta['content_path'] = obj_name obj_meta['container_id'] = utils.name2cid(account, container).upper() obj_meta['ns'] = self.namespace storage_method = STORAGE_METHODS.load(obj_meta['chunk_method']) if storage_method.ec: handler = ECWriteHandler( source, obj_meta, chunk_prep, storage_method, headers=headers, write_timeout=self.write_timeout, read_timeout=self.read_timeout, connection_timeout=self.connection_timeout) elif storage_method.backblaze: backblaze_info = self._b2_credentials(storage_method, key_file) handler = BackblazeWriteHandler(source, obj_meta, chunk_prep, storage_method, headers, backblaze_info) else: handler = ReplicatedWriteHandler( source, obj_meta, chunk_prep, storage_method, headers=headers, write_timeout=self.write_timeout, read_timeout=self.read_timeout, connection_timeout=self.connection_timeout) final_chunks, bytes_transferred, content_checksum = handler.stream() etag = obj_meta.get('etag') if etag and etag.lower() != content_checksum.lower(): raise exc.EtagMismatch("given etag %s != computed %s" % (etag, content_checksum)) obj_meta['etag'] = content_checksum data = {'chunks': final_chunks, 'properties': properties or {}} # FIXME: we may just pass **obj_meta self.container.content_create(account, container, obj_name, size=bytes_transferred, checksum=content_checksum, data=data, content_id=obj_meta['id'], stgpol=obj_meta['policy'], version=obj_meta['version'], mime_type=obj_meta['mime_type'], chunk_method=obj_meta['chunk_method'], headers=headers) return final_chunks, bytes_transferred, content_checksum def _fetch_stream(self, meta, chunks, ranges, storage_method, headers): total_bytes = 0 headers = headers or {} ranges = ranges or [(None, None)] meta_range_list = get_meta_ranges(ranges, chunks) for meta_range_dict in meta_range_list: for pos, meta_range in meta_range_dict.iteritems(): meta_start, meta_end = meta_range if meta_start is not None and meta_end is not None: headers['Range'] = http_header_from_ranges([meta_range]) reader = io.ChunkReader( iter(chunks[pos]), io.READ_CHUNK_SIZE, headers, connection_timeout=self.connection_timeout, response_timeout=self.read_timeout, read_timeout=self.read_timeout) try: it = reader.get_iter() except Exception as err: raise exc.OioException( "Error while downloading position %d: %s" % (pos, err)) for part in it: for d in part['iter']: total_bytes += len(d) yield d def _fetch_stream_ec(self, meta, chunks, ranges, storage_method, headers): ranges = ranges or [(None, None)] meta_range_list = get_meta_ranges(ranges, chunks) for meta_range_dict in meta_range_list: for pos, meta_range in meta_range_dict.iteritems(): meta_start, meta_end = meta_range handler = ECChunkDownloadHandler( storage_method, chunks[pos], meta_start, meta_end, headers, connection_timeout=self.connection_timeout, response_timeout=self.read_timeout, read_timeout=self.read_timeout) stream = handler.get_stream() for part_info in stream: for d in part_info['iter']: yield d stream.close() def _b2_credentials(self, storage_method, key_file): key_file = key_file or '/etc/oio/sds/b2-appkey.conf' try: return BackblazeUtils.get_credentials(storage_method, key_file) except BackblazeUtilsException as err: raise exc.ConfigurationException(str(err)) def _fetch_stream_backblaze(self, meta, chunks, ranges, storage_method, key_file): backblaze_info = self._b2_credentials(storage_method, key_file) total_bytes = 0 current_offset = 0 size = None offset = 0 for pos in range(len(chunks)): if ranges: offset = ranges[pos][0] size = ranges[pos][1] if size is None: size = int(meta["length"]) chunk_size = int(chunks[pos][0]["size"]) if total_bytes >= size: break if current_offset + chunk_size > offset: if current_offset < offset: _offset = offset - current_offset else: _offset = 0 if chunk_size + total_bytes > size: _size = size - total_bytes else: _size = chunk_size handler = BackblazeChunkDownloadHandler( meta, chunks[pos], _offset, _size, backblaze_info=backblaze_info) stream = handler.get_stream() if not stream: raise exc.OioException("Error while downloading") total_bytes += len(stream) yield stream current_offset += chunk_size
class TestDupContent(BaseTestCase): def setUp(self): super(TestDupContent, self).setUp() if len(self.conf['rawx']) < 3: self.skipTest("Not enough rawx. " "Dup tests needs more than 2 rawx to run") self.namespace = self.conf['namespace'] self.account = self.conf['account'] self.chunk_size = self.conf['chunk_size'] self.gridconf = {"namespace": self.namespace} self.content_factory = ContentFactory(self.gridconf) self.container_client = ContainerClient(self.gridconf) self.blob_client = BlobClient() self.container_name = "TestDupContent%f" % time.time() self.container_client.container_create(acct=self.account, ref=self.container_name) self.container_id = cid_from_name(self.account, self.container_name).upper() def tearDown(self): super(TestDupContent, self).tearDown() def _test_upload(self, stgpol, data_size): data = random_data(data_size) content = self.content_factory.new(self.container_id, "titi", len(data), stgpol) self.assertEqual(type(content), DupContent) content.upload(StringIO.StringIO(data)) meta, chunks = self.container_client.content_show( cid=self.container_id, content=content.content_id) chunks = ChunksHelper(chunks) self.assertEqual(meta['hash'], md5_data(data)) self.assertEqual(meta['length'], str(len(data))) self.assertEqual(meta['policy'], stgpol) self.assertEqual(meta['name'], "titi") metachunk_nb = int(math.ceil(float(len(data)) / self.chunk_size)) if metachunk_nb == 0: metachunk_nb = 1 # special case for empty content if stgpol == "THREECOPIES": nb_copy = 3 elif stgpol == "TWOCOPIES": nb_copy = 2 elif stgpol == "SINGLE": nb_copy = 1 self.assertEqual(len(chunks), metachunk_nb * nb_copy) for pos in range(metachunk_nb): chunks_at_pos = chunks.filter(pos=pos) self.assertEqual(len(chunks_at_pos), nb_copy) data_begin = pos * self.chunk_size data_end = pos * self.chunk_size + self.chunk_size chunk_hash = md5_data(data[data_begin:data_end]) for chunk in chunks_at_pos: meta, stream = self.blob_client.chunk_get(chunk.url) self.assertEqual(md5_stream(stream), chunk_hash) self.assertEqual(meta['content_size'], str(len(data))) self.assertEqual(meta['content_path'], "titi") self.assertEqual(meta['content_cid'], self.container_id) self.assertEqual(meta['content_id'], meta['content_id']) self.assertEqual(meta['chunk_id'], chunk.id) self.assertEqual(meta['chunk_pos'], str(pos)) self.assertEqual(meta['chunk_hash'], chunk_hash) def test_twocopies_upload_0_byte(self): self._test_upload("TWOCOPIES", 0) def test_twocopies_upload_1_byte(self): self._test_upload("TWOCOPIES", 1) def test_twocopies_upload_chunksize_bytes(self): self._test_upload("TWOCOPIES", self.chunk_size) def test_twocopies_upload_chunksize_plus_1_bytes(self): self._test_upload("TWOCOPIES", self.chunk_size + 1) def test_single_upload_0_byte(self): self._test_upload("SINGLE", 0) def test_single_upload_chunksize_plus_1_bytes(self): self._test_upload("SINGLE", self.chunk_size + 1) def test_chunks_cleanup_when_upload_failed(self): data = random_data(2 * self.chunk_size) content = self.content_factory.new(self.container_id, "titi", len(data), "TWOCOPIES") self.assertEqual(type(content), DupContent) # set bad url for position 1 for chunk in content.chunks.filter(pos=1): chunk.url = "http://127.0.0.1:9/DEADBEEF" self.assertRaises(Exception, content.upload, StringIO.StringIO(data)) for chunk in content.chunks.exclude(pos=1): self.assertRaises(NotFound, self.blob_client.chunk_head, chunk.url) def _new_content(self, stgpol, data, broken_pos_list): old_content = self.content_factory.new(self.container_id, "titi", len(data), stgpol) self.assertEqual(type(old_content), DupContent) old_content.upload(StringIO.StringIO(data)) for pos, idx in broken_pos_list: c = old_content.chunks.filter(pos=pos)[idx] self.blob_client.chunk_delete(c.url) # get the new structure of the uploaded content return self.content_factory.get(self.container_id, old_content.content_id) def _test_download(self, stgpol, data_size, broken_pos_list): data = random_data(data_size) content = self._new_content(stgpol, data, broken_pos_list) downloaded_data = "".join(content.download()) self.assertEqual(downloaded_data, data) for pos, idx in broken_pos_list: # check nothing has been rebuilt c = content.chunks.filter(pos=pos)[0] self.assertRaises(NotFound, self.blob_client.chunk_delete, c.url) def test_twocopies_download_content_0_byte_without_broken_chunks(self): self._test_download("TWOCOPIES", 0, []) def test_twocopies_download_content_0_byte_with_broken_0_0(self): self._test_download("TWOCOPIES", 0, [(0, 0)]) def test_twocopies_download_content_1_byte_without_broken_chunks(self): self._test_download("TWOCOPIES", 1, []) def test_twocopies_download_content_1_byte_with_broken_0_0(self): self._test_download("TWOCOPIES", 1, [(0, 0)]) def test_twocopies_download_chunksize_bytes_without_broken_chunks(self): self._test_download("TWOCOPIES", self.chunk_size, []) def test_twocopies_download_2xchuksize_bytes_with_broken_0_0_and_1_0(self): self._test_download("TWOCOPIES", self.chunk_size * 2, [(0, 0), (1, 0)]) def test_twocopies_download_content_chunksize_bytes_2_broken_chunks(self): data = random_data(self.chunk_size) content = self._new_content("TWOCOPIES", data, [(0, 0), (0, 1)]) gen = content.download() self.assertRaises(UnrecoverableContent, gen.next) def test_single_download_content_1_byte_without_broken_chunks(self): self._test_download("SINGLE", 1, []) def test_single_download_chunksize_bytes_plus_1_without_broken_chunk(self): self._test_download("SINGLE", self.chunk_size * 2, [])
class ContentFactory(object): def __init__(self, conf): self.conf = conf self.logger = get_logger(conf) self.cs_client = ConscienceClient(conf) self.container_client = ContainerClient(conf) self.ns_info = self.cs_client.info() def _extract_datasec(self, stgpol_name): try: stgpol = self.ns_info["storage_policy"][stgpol_name] except KeyError: self.logger.error("Storage policy '%s' not found" % stgpol_name) raise InconsistentContent("Storage policy not found") stgclass_name, datasec_name, datatreat_name = stgpol.split(':') if datasec_name == 'NONE': return "DUP", {"nb_copy": "1", "distance": "0"} try: datasec = self.ns_info["data_security"][datasec_name] except KeyError: self.logger.error("Data security '%s' not found" % datasec_name) raise InconsistentContent("Data security not found") ds_type, ds_args = datasec.split(':') args = {} for arg in ds_args.split('|'): key, value = arg.split('=') args[key] = value return ds_type, args def get(self, container_id, content_id): try: meta, chunks = self.container_client.content_show( cid=container_id, content=content_id) except NotFound: raise ContentNotFound("Content %s/%s not found" % (container_id, content_id)) pol_type, pol_args = self._extract_datasec(meta['policy']) if pol_type == "DUP": return DupContent(self.conf, container_id, meta, chunks, pol_args) elif pol_type == "RAIN": return RainContent(self.conf, container_id, meta, chunks, pol_args) raise InconsistentContent("Unknown storage policy") def new(self, container_id, path, size, policy): meta, chunks = self.container_client.content_prepare( cid=container_id, path=path, size=size, stgpol=policy) pol_type, pol_args = self._extract_datasec(meta['policy']) if pol_type == "DUP": return DupContent(self.conf, container_id, meta, chunks, pol_args) elif pol_type == "RAIN": return RainContent(self.conf, container_id, meta, chunks, pol_args) raise InconsistentContent("Unknown storage policy") def change_policy(self, container_id, content_id, new_policy): old_content = self.get(container_id, content_id) if old_content.stgpol_name == new_policy: return old_content new_content = self.new(container_id, old_content.path, old_content.length, new_policy) stream = old_content.download() new_content.upload(GeneratorIO(stream)) # the old content is automatically deleted because the new content has # the same name (but not the same id) return new_content