Python InternalClient.iter_objects方法代码示例

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]
    def download(self, acc, container, u_agent, delay=0, request_tries=3):
        self.logger.info('Prefetching objects with InternalClient with ' + str(delay) + ' seconds of delay.')
        time.sleep(delay)
        swift = InternalClient(PROXY_PATH, u_agent, request_tries=request_tries)
        headers = {}

        prefetch_list = []
        bytes_count = 0
        for o in swift.iter_objects(acc, container):
            if bytes_count + int(o['bytes']) < self.cache_max_size:
                prefetch_list.append(o['name'])
                bytes_count += int(o['bytes'])
            else:
                break

        for name in prefetch_list:
            object_path = '/v1/' + acc + '/' + container + '/' + name
            oid = hashlib.md5(object_path).hexdigest()

            status, resp_headers, it = swift.get_object(acc, container, name, headers, ACCEPTABLE_STATUS)

            object_size = int(resp_headers.get('Content-Length'))
            object_etag = resp_headers.get('Etag')

            object_storage_policy_id = '0'  # FIXME hardcoded
            to_evict = self.cache.access_cache("PUT", oid, object_size, object_etag, object_storage_policy_id)
            for ev_object_id in to_evict:
                os.remove(os.path.join(self.cache_path, ev_object_id))

            self.logger.info('Prefetch Filter - Object ' + name + ' stored in cache with ID: ' + oid)
            with open(os.path.join(self.cache_path, oid), 'w') as f:
                for el in it:
                    f.write(el)

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]

#.........这里部分代码省略.........
        return success

    def _iter_containers(self):
        """
        Generate a list of containers to process.
        """
        # hit most recent container first instead of waiting on the updaters
        current_container = get_reconciler_container_name(time.time())
        yield current_container
        container_gen = self.swift.iter_containers(MISPLACED_OBJECTS_ACCOUNT)
        self.logger.debug('looking for containers in %s',
                          MISPLACED_OBJECTS_ACCOUNT)
        while True:
            one_page = None
            try:
                one_page = list(itertools.islice(
                    container_gen, constraints.CONTAINER_LISTING_LIMIT))
            except UnexpectedResponse as err:
                self.logger.error('Error listing containers in '
                                  'account %s (%s)',
                                  MISPLACED_OBJECTS_ACCOUNT, err)

            if not one_page:
                # don't generally expect more than one page
                break
            # reversed order since we expect older containers to be empty
            for c in reversed(one_page):
                # encoding here is defensive
                container = c['name'].encode('utf8')
                if container == current_container:
                    continue  # we've already hit this one this pass
                yield container

    def _iter_objects(self, container):
        """
        Generate a list of objects to process.

        :param container: the name of the container to process

        If the given container is empty and older than reclaim_age this
        processor will attempt to reap it.
        """
        self.logger.debug('looking for objects in %s', container)
        found_obj = False
        try:
            for raw_obj in self.swift.iter_objects(
                    MISPLACED_OBJECTS_ACCOUNT, container):
                found_obj = True
                yield raw_obj
        except UnexpectedResponse as err:
            self.logger.error('Error listing objects in container %s (%s)',
                              container, err)
        if float(container) < time.time() - self.reclaim_age and \
                not found_obj:
            # Try to delete old empty containers so the queue doesn't
            # grow without bound. It's ok if there's a conflict.
            self.swift.delete_container(
                MISPLACED_OBJECTS_ACCOUNT, container,
                acceptable_statuses=(2, 404, 409, 412))

    def reconcile(self):
        """
        Main entry point for processing misplaced objects.

        Iterate over all queue entries and delegate to reconcile_object.
        """

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]
class TestObjectExpirer(ReplProbeTest):

    def setUp(self):
        self.expirer = Manager(['object-expirer'])
        self.expirer.start()
        err = self.expirer.stop()
        if err:
            raise unittest.SkipTest('Unable to verify object-expirer service')

        conf_files = []
        for server in self.expirer.servers:
            conf_files.extend(server.conf_files())
        conf_file = conf_files[0]
        self.client = InternalClient(conf_file, 'probe-test', 3)

        super(TestObjectExpirer, self).setUp()
        self.container_name = 'container-%s' % uuid.uuid4()
        self.object_name = 'object-%s' % uuid.uuid4()
        self.brain = BrainSplitter(self.url, self.token, self.container_name,
                                   self.object_name)

    def _check_obj_in_container_listing(self):
        for obj in self.client.iter_objects(self.account,
                                            self.container_name):

            if self.object_name == obj['name']:
                return True

        return False

    @unittest.skipIf(len(ENABLED_POLICIES) < 2, "Need more than one policy")
    def test_expirer_object_split_brain(self):
        old_policy = random.choice(ENABLED_POLICIES)
        wrong_policy = random.choice([p for p in ENABLED_POLICIES
                                      if p != old_policy])
        # create an expiring object and a container with the wrong policy
        self.brain.stop_primary_half()
        self.brain.put_container(int(old_policy))
        self.brain.put_object(headers={'X-Delete-After': 2})
        # get the object timestamp
        metadata = self.client.get_object_metadata(
            self.account, self.container_name, self.object_name,
            headers={'X-Backend-Storage-Policy-Index': int(old_policy)})
        create_timestamp = Timestamp(metadata['x-timestamp'])
        self.brain.start_primary_half()
        # get the expiring object updates in their queue, while we have all
        # the servers up
        Manager(['object-updater']).once()
        self.brain.stop_handoff_half()
        self.brain.put_container(int(wrong_policy))
        # don't start handoff servers, only wrong policy is available

        # make sure auto-created containers get in the account listing
        Manager(['container-updater']).once()
        # this guy should no-op since it's unable to expire the object
        self.expirer.once()

        self.brain.start_handoff_half()
        self.get_to_final_state()

        # validate object is expired
        found_in_policy = None
        metadata = self.client.get_object_metadata(
            self.account, self.container_name, self.object_name,
            acceptable_statuses=(4,),
            headers={'X-Backend-Storage-Policy-Index': int(old_policy)})
        self.assertIn('x-backend-timestamp', metadata)
        self.assertEqual(Timestamp(metadata['x-backend-timestamp']),
                         create_timestamp)

        # but it is still in the listing
        self.assertTrue(self._check_obj_in_container_listing(),
                        msg='Did not find listing for %s' % self.object_name)

        # clear proxy cache
        client.post_container(self.url, self.token, self.container_name, {})
        # run the expirer again after replication
        self.expirer.once()

        # object is not in the listing
        self.assertFalse(self._check_obj_in_container_listing(),
                         msg='Found listing for %s' % self.object_name)

        # and validate object is tombstoned
        found_in_policy = None
        for policy in ENABLED_POLICIES:
            metadata = self.client.get_object_metadata(
                self.account, self.container_name, self.object_name,
                acceptable_statuses=(4,),
                headers={'X-Backend-Storage-Policy-Index': int(policy)})
            if 'x-backend-timestamp' in metadata:
                if found_in_policy:
                    self.fail('found object in %s and also %s' %
                              (found_in_policy, policy))
                found_in_policy = policy
                self.assertIn('x-backend-timestamp', metadata)
                self.assertGreater(Timestamp(metadata['x-backend-timestamp']),
                                   create_timestamp)

    def test_expirer_doesnt_make_async_pendings(self):
#.........这里部分代码省略.........

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]
class ObjectExpirer(Daemon):
    """
    Daemon that queries the internal hidden expiring_objects_account to
    discover objects that need to be deleted.

    :param conf: The daemon configuration.
    """

    def __init__(self, conf):
        self.conf = conf
        self.logger = get_logger(conf, log_route='object-expirer')
        self.interval = int(conf.get('interval') or 300)
        self.expiring_objects_account = \
            (conf.get('auto_create_account_prefix') or '.') + \
            'expiring_objects'
        conf_path = conf.get('__file__') or '/etc/swift/object-expirer.conf'
        request_tries = int(conf.get('request_tries') or 3)
        self.swift = InternalClient(conf_path,
                                    'Swift Object Expirer',
                                    request_tries)
        self.report_interval = int(conf.get('report_interval') or 300)
        self.report_first_time = self.report_last_time = time()
        self.report_objects = 0
        self.recon_cache_path = conf.get('recon_cache_path',
                                         '/var/cache/swift')
        self.rcache = join(self.recon_cache_path, 'object.recon')
        self.concurrency = int(conf.get('concurrency', 1))
        if self.concurrency < 1:
            raise ValueError("concurrency must be set to at least 1")
        self.processes = int(self.conf.get('processes', 0))
        self.process = int(self.conf.get('process', 0))

    def report(self, final=False):
        """
        Emits a log line report of the progress so far, or the final progress
        is final=True.

        :param final: Set to True for the last report once the expiration pass
                      has completed.
        """
        if final:
            elapsed = time() - self.report_first_time
            self.logger.info(_('Pass completed in %ds; %d objects expired') %
                             (elapsed, self.report_objects))
            dump_recon_cache({'object_expiration_pass': elapsed,
                              'expired_last_pass': self.report_objects},
                             self.rcache, self.logger)
        elif time() - self.report_last_time >= self.report_interval:
            elapsed = time() - self.report_first_time
            self.logger.info(_('Pass so far %ds; %d objects expired') %
                             (elapsed, self.report_objects))
            self.report_last_time = time()

    def run_once(self, *args, **kwargs):
        """
        Executes a single pass, looking for objects to expire.

        :param args: Extra args to fulfill the Daemon interface; this daemon
                     has no additional args.
        :param kwargs: Extra keyword args to fulfill the Daemon interface; this
                       daemon accepts processes and process keyword args.
                       These will override the values from the config file if
                       provided.
        """
        processes, process = self.get_process_values(kwargs)
        pool = GreenPool(self.concurrency)
        containers_to_delete = []
        self.report_first_time = self.report_last_time = time()
        self.report_objects = 0
        try:
            self.logger.debug(_('Run begin'))
            containers, objects = \
                self.swift.get_account_info(self.expiring_objects_account)
            self.logger.info(_('Pass beginning; %s possible containers; %s '
                               'possible objects') % (containers, objects))
            for c in self.swift.iter_containers(self.expiring_objects_account):
                container = c['name']
                timestamp = int(container)
                if timestamp > int(time()):
                    break
                containers_to_delete.append(container)
                for o in self.swift.iter_objects(self.expiring_objects_account,
                                                 container):
                    obj = o['name'].encode('utf8')
                    if processes > 0:
                        obj_process = int(
                            hashlib.md5('%s/%s' % (container, obj)).
                            hexdigest(), 16)
                        if obj_process % processes != process:
                            continue
                    timestamp, actual_obj = obj.split('-', 1)
                    timestamp = int(timestamp)
                    if timestamp > int(time()):
                        break
                    pool.spawn_n(
                        self.delete_object, actual_obj, timestamp,
                        container, obj)
            pool.waitall()
            for container in containers_to_delete:
                try:
#.........这里部分代码省略.........

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]
    def test_reconciler_move_object_twice(self):
        # select some policies
        old_policy = random.choice(ENABLED_POLICIES)
        new_policy = random.choice([p for p in ENABLED_POLICIES
                                    if p != old_policy])

        # setup a split brain
        self.brain.stop_handoff_half()
        # get old_policy on two primaries
        self.brain.put_container(policy_index=int(old_policy))
        self.brain.start_handoff_half()
        self.brain.stop_primary_half()
        # force a recreate on handoffs
        self.brain.put_container(policy_index=int(old_policy))
        self.brain.delete_container()
        self.brain.put_container(policy_index=int(new_policy))
        self.brain.put_object()  # populate memcache with new_policy
        self.brain.start_primary_half()

        # at this point two primaries have old policy
        container_part, container_nodes = self.container_ring.get_nodes(
            self.account, self.container_name)
        head_responses = []
        for node in container_nodes:
            metadata = direct_client.direct_head_container(
                node, container_part, self.account, self.container_name)
            head_responses.append((node, metadata))
        old_container_node_ids = [
            node['id'] for node, metadata in head_responses
            if int(old_policy) ==
            int(metadata['X-Backend-Storage-Policy-Index'])]
        self.assertEqual(2, len(old_container_node_ids))

        # hopefully memcache still has the new policy cached
        self.brain.put_object(headers={'x-object-meta-test': 'custom-meta'},
                              contents='VERIFY')
        # double-check object correctly written to new policy
        conf_files = []
        for server in Manager(['container-reconciler']).servers:
            conf_files.extend(server.conf_files())
        conf_file = conf_files[0]
        int_client = InternalClient(conf_file, 'probe-test', 3)
        int_client.get_object_metadata(
            self.account, self.container_name, self.object_name,
            headers={'X-Backend-Storage-Policy-Index': int(new_policy)})
        int_client.get_object_metadata(
            self.account, self.container_name, self.object_name,
            acceptable_statuses=(4,),
            headers={'X-Backend-Storage-Policy-Index': int(old_policy)})

        # shutdown the containers that know about the new policy
        self.brain.stop_handoff_half()

        # and get rows enqueued from old nodes
        for server_type in ('container-replicator', 'container-updater'):
            server = Manager([server_type])
            tuple(server.once(number=n + 1) for n in old_container_node_ids)

        # verify entry in the queue for the "misplaced" new_policy
        for container in int_client.iter_containers('.misplaced_objects'):
            for obj in int_client.iter_objects('.misplaced_objects',
                                               container['name']):
                expected = '%d:/%s/%s/%s' % (new_policy, self.account,
                                             self.container_name,
                                             self.object_name)
                self.assertEqual(obj['name'], expected)

        Manager(['container-reconciler']).once()

        # verify object in old_policy
        int_client.get_object_metadata(
            self.account, self.container_name, self.object_name,
            headers={'X-Backend-Storage-Policy-Index': int(old_policy)})

        # verify object is *not* in new_policy
        int_client.get_object_metadata(
            self.account, self.container_name, self.object_name,
            acceptable_statuses=(4,),
            headers={'X-Backend-Storage-Policy-Index': int(new_policy)})

        self.get_to_final_state()

        # verify entry in the queue
        for container in int_client.iter_containers('.misplaced_objects'):
            for obj in int_client.iter_objects('.misplaced_objects',
                                               container['name']):
                expected = '%d:/%s/%s/%s' % (old_policy, self.account,
                                             self.container_name,
                                             self.object_name)
                self.assertEqual(obj['name'], expected)

        Manager(['container-reconciler']).once()

        # and now it flops back
        int_client.get_object_metadata(
            self.account, self.container_name, self.object_name,
            headers={'X-Backend-Storage-Policy-Index': int(new_policy)})
        int_client.get_object_metadata(
            self.account, self.container_name, self.object_name,
            acceptable_statuses=(4,),
#.........这里部分代码省略.........

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]
class ObjectExpirer(Daemon):
    """
    Daemon that queries the internal hidden expiring_objects_account to
    discover objects that need to be deleted.

    :param conf: The daemon configuration.
    """

    def __init__(self, conf):
        self.conf = conf
        self.logger = get_logger(conf, log_route='object-expirer')
        self.interval = int(conf.get('interval') or 300)
        self.expiring_objects_account = \
            (conf.get('auto_create_account_prefix') or '.') + \
            'expiring_objects'
        conf_path = conf.get('__file__') or '/etc/swift/object-expirer.conf'
        request_tries = int(conf.get('request_tries') or 3)
        self.swift = InternalClient(conf_path,
                                    'Swift Object Expirer',
                                    request_tries)
        self.report_interval = int(conf.get('report_interval') or 300)
        self.report_first_time = self.report_last_time = time()
        self.report_objects = 0
        self.recon_cache_path = conf.get('recon_cache_path',
                                         '/var/cache/swift')
        self.rcache = join(self.recon_cache_path, 'object.recon')

    def report(self, final=False):
        """
        Emits a log line report of the progress so far, or the final progress
        is final=True.

        :param final: Set to True for the last report once the expiration pass
                      has completed.
        """
        if final:
            elapsed = time() - self.report_first_time
            self.logger.info(_('Pass completed in %ds; %d objects expired') %
                             (elapsed, self.report_objects))
            dump_recon_cache({'object_expiration_pass': elapsed,
                              'expired_last_pass': self.report_objects},
                             self.rcache, self.logger)
        elif time() - self.report_last_time >= self.report_interval:
            elapsed = time() - self.report_first_time
            self.logger.info(_('Pass so far %ds; %d objects expired') %
                             (elapsed, self.report_objects))
            self.report_last_time = time()

    def run_once(self, *args, **kwargs):
        """
        Executes a single pass, looking for objects to expire.

        :param args: Extra args to fulfill the Daemon interface; this daemon
                     has no additional args.
        :param kwargs: Extra keyword args to fulfill the Daemon interface; this
                       daemon has no additional keyword args.
        """
        self.report_first_time = self.report_last_time = time()
        self.report_objects = 0
        try:
            self.logger.debug(_('Run begin'))
            containers, objects = \
                self.swift.get_account_info(self.expiring_objects_account)
            self.logger.info(_('Pass beginning; %s possible containers; %s '
                               'possible objects') % (containers, objects))
            for c in self.swift.iter_containers(self.expiring_objects_account):
                container = c['name']
                timestamp = int(container)
                if timestamp > int(time()):
                    break
                for o in self.swift.iter_objects(self.expiring_objects_account,
                                                 container):
                    obj = o['name']
                    timestamp, actual_obj = obj.split('-', 1)
                    timestamp = int(timestamp)
                    if timestamp > int(time()):
                        break
                    start_time = time()
                    try:
                        self.delete_actual_object(actual_obj, timestamp)
                        self.swift.delete_object(self.expiring_objects_account,
                                                 container, obj)
                        self.report_objects += 1
                        self.logger.increment('objects')
                    except (Exception, Timeout), err:
                        self.logger.increment('errors')
                        self.logger.exception(
                            _('Exception while deleting object %s %s %s') %
                            (container, obj, str(err)))
                    self.logger.timing_since('timing', start_time)
                    self.report()
                try:
                    self.swift.delete_container(
                        self.expiring_objects_account,
                        container,
                        acceptable_statuses=(2, HTTP_NOT_FOUND, HTTP_CONFLICT))
                except (Exception, Timeout), err:
                    self.logger.exception(
                        _('Exception while deleting container %s %s') %
                        (container, str(err)))
#.........这里部分代码省略.........

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]
class TestObjectExpirer(unittest.TestCase):

    def setUp(self):
        if len(POLICIES) < 2:
            raise SkipTest('Need more than one policy')

        self.expirer = Manager(['object-expirer'])
        self.expirer.start()
        err = self.expirer.stop()
        if err:
            raise SkipTest('Unable to verify object-expirer service')

        conf_files = []
        for server in self.expirer.servers:
            conf_files.extend(server.conf_files())
        conf_file = conf_files[0]
        self.client = InternalClient(conf_file, 'probe-test', 3)

        (self.pids, self.port2server, self.account_ring, self.container_ring,
         self.object_ring, self.policy, self.url, self.token,
         self.account, self.configs) = reset_environment()
        self.container_name = 'container-%s' % uuid.uuid4()
        self.object_name = 'object-%s' % uuid.uuid4()
        self.brain = BrainSplitter(self.url, self.token, self.container_name,
                                   self.object_name)

    def test_expirer_object_split_brain(self):
        old_policy = random.choice(list(POLICIES))
        wrong_policy = random.choice([p for p in POLICIES if p != old_policy])
        # create an expiring object and a container with the wrong policy
        self.brain.stop_primary_half()
        self.brain.put_container(int(old_policy))
        self.brain.put_object(headers={'X-Delete-After': 2})
        # get the object timestamp
        metadata = self.client.get_object_metadata(
            self.account, self.container_name, self.object_name,
            headers={'X-Backend-Storage-Policy-Index': int(old_policy)})
        create_timestamp = Timestamp(metadata['x-timestamp'])
        self.brain.start_primary_half()
        # get the expiring object updates in their queue, while we have all
        # the servers up
        Manager(['object-updater']).once()
        self.brain.stop_handoff_half()
        self.brain.put_container(int(wrong_policy))
        # don't start handoff servers, only wrong policy is available

        # make sure auto-created containers get in the account listing
        Manager(['container-updater']).once()
        # this guy should no-op since it's unable to expire the object
        self.expirer.once()

        self.brain.start_handoff_half()
        get_to_final_state()

        # validate object is expired
        found_in_policy = None
        metadata = self.client.get_object_metadata(
            self.account, self.container_name, self.object_name,
            acceptable_statuses=(4,),
            headers={'X-Backend-Storage-Policy-Index': int(old_policy)})
        self.assert_('x-backend-timestamp' in metadata)
        self.assertEqual(Timestamp(metadata['x-backend-timestamp']),
                         create_timestamp)

        # but it is still in the listing
        for obj in self.client.iter_objects(self.account,
                                            self.container_name):
            if self.object_name == obj['name']:
                break
        else:
            self.fail('Did not find listing for %s' % self.object_name)

        # clear proxy cache
        client.post_container(self.url, self.token, self.container_name, {})
        # run the expirier again after replication
        self.expirer.once()

        # object is not in the listing
        for obj in self.client.iter_objects(self.account,
                                            self.container_name):
            if self.object_name == obj['name']:
                self.fail('Found listing for %s' % self.object_name)

        # and validate object is tombstoned
        found_in_policy = None
        for policy in POLICIES:
            metadata = self.client.get_object_metadata(
                self.account, self.container_name, self.object_name,
                acceptable_statuses=(4,),
                headers={'X-Backend-Storage-Policy-Index': int(policy)})
            if 'x-backend-timestamp' in metadata:
                if found_in_policy:
                    self.fail('found object in %s and also %s' %
                              (found_in_policy, policy))
                found_in_policy = policy
                self.assert_('x-backend-timestamp' in metadata)
                self.assert_(Timestamp(metadata['x-backend-timestamp']) >
                             create_timestamp)

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]
class ObjectExpirer(Daemon):
    """
    Daemon that queries the internal hidden expiring_objects_account to
    discover objects that need to be deleted.

    :param conf: The daemon configuration.
    """

    def __init__(self, conf):
        self.conf = conf
        self.logger = get_logger(conf, log_route='object-expirer')
        self.interval = int(conf.get('interval') or 300)
        self.expiring_objects_account = \
            (conf.get('auto_create_account_prefix') or '.') + \
            (conf.get('expiring_objects_account_name') or 'expiring_objects')
        conf_path = conf.get('__file__') or '/etc/swift/object-expirer.conf'
        request_tries = int(conf.get('request_tries') or 3)
        self.swift = InternalClient(conf_path,
                                    'Swift Object Expirer',
                                    request_tries)
        self.report_interval = int(conf.get('report_interval') or 300)
        self.report_first_time = self.report_last_time = time()
        self.report_objects = 0
        self.recon_cache_path = conf.get('recon_cache_path',
                                         '/var/cache/swift')
        self.rcache = join(self.recon_cache_path, 'object.recon')
        self.concurrency = int(conf.get('concurrency', 1))
        if self.concurrency < 1:
            raise ValueError("concurrency must be set to at least 1")
        self.processes = int(self.conf.get('processes', 0))
        self.process = int(self.conf.get('process', 0))

    def report(self, final=False):
        """
        Emits a log line report of the progress so far, or the final progress
        is final=True.

        :param final: Set to True for the last report once the expiration pass
                      has completed.
        """
        if final:
            elapsed = time() - self.report_first_time
            self.logger.info(_('Pass completed in %ds; %d objects expired') %
                             (elapsed, self.report_objects))
            dump_recon_cache({'object_expiration_pass': elapsed,
                              'expired_last_pass': self.report_objects},
                             self.rcache, self.logger)
        elif time() - self.report_last_time >= self.report_interval:
            elapsed = time() - self.report_first_time
            self.logger.info(_('Pass so far %ds; %d objects expired') %
                             (elapsed, self.report_objects))
            self.report_last_time = time()

    def run_once(self, *args, **kwargs):
        """
        Executes a single pass, looking for objects to expire.

        :param args: Extra args to fulfill the Daemon interface; this daemon
                     has no additional args.
        :param kwargs: Extra keyword args to fulfill the Daemon interface; this
                       daemon accepts processes and process keyword args.
                       These will override the values from the config file if
                       provided.
        """
        processes, process = self.get_process_values(kwargs)
        pool = GreenPool(self.concurrency)
        containers_to_delete = []
        self.report_first_time = self.report_last_time = time()
        self.report_objects = 0
        try:
            self.logger.debug(_('Run begin'))
            containers, objects = \
                self.swift.get_account_info(self.expiring_objects_account)
            self.logger.info(_('Pass beginning; %s possible containers; %s '
                               'possible objects') % (containers, objects))
            for c in self.swift.iter_containers(self.expiring_objects_account):
                container = c['name']
                timestamp = int(container)
                if timestamp > int(time()):
                    break
                containers_to_delete.append(container)
                for o in self.swift.iter_objects(self.expiring_objects_account,
                                                 container):
                    obj = o['name'].encode('utf8')
                    if processes > 0:
                        obj_process = int(
                            hashlib.md5('%s/%s' % (container, obj)).
                            hexdigest(), 16)
                        if obj_process % processes != process:
                            continue
                    timestamp, actual_obj = obj.split('-', 1)
                    timestamp = int(timestamp)
                    if timestamp > int(time()):
                        break
                    pool.spawn_n(
                        self.delete_object, actual_obj, timestamp,
                        container, obj)
            pool.waitall()
            for container in containers_to_delete:
                try:
#.........这里部分代码省略.........

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]
class TestObjectExpirer(ReplProbeTest):

    def setUp(self):
        if len(ENABLED_POLICIES) < 2:
            raise SkipTest('Need more than one policy')

        self.expirer = Manager(['object-expirer'])
        self.expirer.start()
        err = self.expirer.stop()
        if err:
            raise SkipTest('Unable to verify object-expirer service')

        conf_files = []
        for server in self.expirer.servers:
            conf_files.extend(server.conf_files())
        conf_file = conf_files[0]
        self.client = InternalClient(conf_file, 'probe-test', 3)

        super(TestObjectExpirer, self).setUp()
        self.container_name = 'container-%s' % uuid.uuid4()
        self.object_name = 'object-%s' % uuid.uuid4()
        self.brain = BrainSplitter(self.url, self.token, self.container_name,
                                   self.object_name)

    def test_expirer_object_split_brain(self):
        old_policy = random.choice(ENABLED_POLICIES)
        wrong_policy = random.choice([p for p in ENABLED_POLICIES
                                      if p != old_policy])
        # create an expiring object and a container with the wrong policy
        self.brain.stop_primary_half()
        self.brain.put_container(int(old_policy))
        self.brain.put_object(headers={'X-Delete-After': 2})
        # get the object timestamp
        metadata = self.client.get_object_metadata(
            self.account, self.container_name, self.object_name,
            headers={'X-Backend-Storage-Policy-Index': int(old_policy)})
        create_timestamp = Timestamp(metadata['x-timestamp'])
        self.brain.start_primary_half()
        # get the expiring object updates in their queue, while we have all
        # the servers up
        Manager(['object-updater']).once()
        self.brain.stop_handoff_half()
        self.brain.put_container(int(wrong_policy))
        # don't start handoff servers, only wrong policy is available

        # make sure auto-created containers get in the account listing
        Manager(['container-updater']).once()
        # this guy should no-op since it's unable to expire the object
        self.expirer.once()

        self.brain.start_handoff_half()
        self.get_to_final_state()

        # validate object is expired
        found_in_policy = None
        metadata = self.client.get_object_metadata(
            self.account, self.container_name, self.object_name,
            acceptable_statuses=(4,),
            headers={'X-Backend-Storage-Policy-Index': int(old_policy)})
        self.assertTrue('x-backend-timestamp' in metadata)
        self.assertEqual(Timestamp(metadata['x-backend-timestamp']),
                         create_timestamp)

        # but it is still in the listing
        for obj in self.client.iter_objects(self.account,
                                            self.container_name):
            if self.object_name == obj['name']:
                break
        else:
            self.fail('Did not find listing for %s' % self.object_name)

        # clear proxy cache
        client.post_container(self.url, self.token, self.container_name, {})
        # run the expirier again after replication
        self.expirer.once()

        # object is not in the listing
        for obj in self.client.iter_objects(self.account,
                                            self.container_name):
            if self.object_name == obj['name']:
                self.fail('Found listing for %s' % self.object_name)

        # and validate object is tombstoned
        found_in_policy = None
        for policy in ENABLED_POLICIES:
            metadata = self.client.get_object_metadata(
                self.account, self.container_name, self.object_name,
                acceptable_statuses=(4,),
                headers={'X-Backend-Storage-Policy-Index': int(policy)})
            if 'x-backend-timestamp' in metadata:
                if found_in_policy:
                    self.fail('found object in %s and also %s' %
                              (found_in_policy, policy))
                found_in_policy = policy
                self.assertTrue('x-backend-timestamp' in metadata)
                self.assertTrue(Timestamp(metadata['x-backend-timestamp']) >
                                create_timestamp)

    def test_expirer_object_should_not_be_expired(self):
        obj_brain = BrainSplitter(self.url, self.token, self.container_name,
#.........这里部分代码省略.........

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]

#.........这里部分代码省略.........
            processes = int(kwargs['processes'])
        else:
            processes = self.processes

        if kwargs.get('process') is not None:
            process = int(kwargs['process'])
        else:
            process = self.process

        if process < 0:
            raise ValueError(
                'process must be an integer greater than or equal to 0')

        if processes < 0:
            raise ValueError(
                'processes must be an integer greater than or equal to 0')

        if processes and process >= processes:
            raise ValueError(
                'process must be less than or equal to processes')

        return processes, process

    def aggregate_container(self, container):
        start_time = time()
        try:
            objs_to_delete = list()
            bytes_recvs = dict()
            bytes_sents = dict()

            ts, tenant_id, account = container.split('_', 2)
            ts = int(float(ts))

            for o in self.swift.iter_objects(self.sample_account, container):
                name = o['name']
                objs_to_delete.append(name)
                ts, bytes_rv, bytes_st, trans_id, client_ip = name.split('/')
                bill_type = self.get_billtype_by_client_ip(client_ip, ts)
                bytes_recvs[bill_type] = bytes_recvs.get(bill_type,
                                                         0) + int(bytes_rv)
                bytes_sents[bill_type] = bytes_sents.get(bill_type,
                                                         0) + int(bytes_st)
                self.report_objects += 1

            for o in objs_to_delete:
                self.swift.delete_object(self.sample_account, container, o)

            for bill_type, bt_rv in bytes_recvs.items():
                t_object = 'transfer/%d/%d/%d_%d_%d' % (ts, bill_type, bt_rv,
                                                        bytes_sents[bill_type],
                                                        self.report_objects)
                self._hidden_update(tenant_id, t_object)
        except (Exception, Timeout) as err:
            self.logger.increment('errors')
            self.logger.exception(
                _('Exception while aggregating sample %s %s') %
                (container, str(err)))

        self.logger.timing_since('timing', start_time)
        self.report()

    def account_info(self, tenant_id, timestamp):
        path = '/v1/%s/%s?prefix=usage/%d&limit=1' % (self.aggregate_account,
                                                      tenant_id, timestamp)
        resp = self.swift.make_request('GET', path, {}, (2,))
        if len(resp.body) == 0:

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]
class ObjectRestorer(Daemon):
    """
    Daemon that queries the internal hidden expiring_objects_account to
    discover objects that need to be deleted.

    :param conf: The daemon configuration.
    """

    def __init__(self, conf):
        self.conf = conf
        self.container_ring = Ring('/etc/swift', ring_name='container')
        self.logger = get_logger(conf, log_route='object-restorer')
        self.logger.set_statsd_prefix('s3-object-restorer')
        self.interval = int(conf.get('interval') or 300)
        self.restoring_object_account = '.s3_restoring_objects'
        self.expiring_restored_account = '.s3_expiring_restored_objects'
        self.glacier_account_prefix = '.glacier_'
        self.todo_container = 'todo'
        self.restoring_container = 'restoring'
        conf_path = '/etc/swift/s3-object-restorer.conf'
        request_tries = int(conf.get('request_tries') or 3)
        self.glacier = self._init_glacier()
        self.glacier_tmpdir = conf.get('temp_path', '/var/cache/s3/')
        self.swift = InternalClient(conf_path,
                                    'Swift Object Restorer',
                                    request_tries)
        self.report_interval = int(conf.get('report_interval') or 300)
        self.report_first_time = self.report_last_time = time()
        self.report_objects = 0
        self.recon_cache_path = conf.get('recon_cache_path',
                                         '/var/cache/swift')
        self.rcache = join(self.recon_cache_path, 'object.recon')
        self.concurrency = int(conf.get('concurrency', 1))
        if self.concurrency < 1:
            raise ValueError("concurrency must be set to at least 1")
        self.processes = int(self.conf.get('processes', 0))
        self.process = int(self.conf.get('process', 0))
        self.client = Client(self.conf.get('sentry_sdn', ''))

    def _init_glacier(self):
        con = Layer2(region_name='ap-northeast-1')
        return con.get_vault('swift-s3-transition')

    def report(self, final=False):
        """
        Emits a log line report of the progress so far, or the final progress
        is final=True.

        :param final: Set to True for the last report once the expiration pass
                      has completed.
        """
        if final:
            elapsed = time() - self.report_first_time
            self.logger.info(_('Pass completed in %ds; %d objects restored') %
                             (elapsed, self.report_objects))
            dump_recon_cache({'object_expiration_pass': elapsed,
                              'expired_last_pass': self.report_objects},
                             self.rcache, self.logger)
        elif time() - self.report_last_time >= self.report_interval:
            elapsed = time() - self.report_first_time
            self.logger.info(_('Pass so far %ds; %d objects restored') %
                             (elapsed, self.report_objects))
            self.report_last_time = time()

    def run_once(self, *args, **kwargs):
        """
        Executes a single pass, looking for objects to expire.

        :param args: Extra args to fulfill the Daemon interface; this daemon
                     has no additional args.
        :param kwargs: Extra keyword args to fulfill the Daemon interface; this
                       daemon accepts processes and process keyword args.
                       These will override the values from the config file if
                       provided.
        """
        processes, process = self.get_process_values(kwargs)
        pool = GreenPool(self.concurrency)
        self.report_first_time = self.report_last_time = time()
        self.report_objects = 0
        try:
            self.logger.debug(_('Run begin'))

            for o in self.swift.iter_objects(self.restoring_object_account,
                                             self.todo_container):
                obj = o['name'].encode('utf8')
                if processes > 0:
                    obj_process = int(
                        hashlib.md5('%s/%s' % (self.todo_container, obj)).
                        hexdigest(), 16)
                    if obj_process % processes != process:
                        continue
                pool.spawn_n(self.start_object_restoring, obj)

            pool.waitall()

            for o in self.swift.iter_objects(self.restoring_object_account,
                                             self.restoring_container):
                obj = o['name'].encode('utf8')
                if processes > 0:
                    obj_process = int(
#.........这里部分代码省略.........

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]
class ObjectExpirer(Daemon):
    def __init__(self, conf):
        super(ObjectExpirer, self).__init__(conf)
        self.conf = conf
        self.logger = get_logger(conf, log_route='s3-object-expirer')
        self.logger.set_statsd_prefix('s3-object-expirer')
        self.interval = int(conf.get('interval') or 300)
        self.s3_expiring_objects_account = \
            (conf.get('auto_create_account_prefix') or '.') + \
            (conf.get('expiring_objects_account_name') or
             's3_expiring_objects')
        conf_path = conf.get('__file__') or '/etc/swift/s3-object-expirer.conf'
        request_tries = int(conf.get('request_tries') or 3)
        self.swift = InternalClient(conf_path,
                                    'Swift Object Expirer',
                                    request_tries)
        self.glacier = self._init_glacier()
        self.glacier_account_prefix = '.glacier_'
        self.report_interval = int(conf.get('report_interval') or 300)
        self.report_first_time = self.report_last_time = time()
        self.report_objects = 0
        self.recon_cache_path = conf.get('recon_cache_path',
                                         '/var/cache/swift')
        self.rcache = join(self.recon_cache_path, 'object.recon')
        self.concurrency = int(conf.get('concurrency', 1))
        if self.concurrency < 1:
            raise ValueError("concurrency must be set to at least 1")
        self.processes = int(self.conf.get('processes', 0))
        self.process = int(self.conf.get('process', 0))
        self.client = Client(self.conf.get('sentry_sdn', ''))

    def _init_glacier(self):
        con = Layer2(region_name='ap-northeast-1')
        return con.get_vault('swift-s3-transition')

    def report(self, final=False):
        """
        Emits a log line report of the progress so far, or the final progress
        is final=True.

        :param final: Set to True for the last report once the expiration pass
                      has completed.
        """
        if final:
            elapsed = time() - self.report_first_time
            self.logger.info(_('Pass completed in %ds; %d objects expired') %
                             (elapsed, self.report_objects))
            dump_recon_cache({'object_expiration_pass': elapsed,
                              'expired_last_pass': self.report_objects},
                             self.rcache, self.logger)
        elif time() - self.report_last_time >= self.report_interval:
            elapsed = time() - self.report_first_time
            self.logger.info(_('Pass so far %ds; %d objects expired') %
                             (elapsed, self.report_objects))
            self.report_last_time = time()

    def run_once(self, *args, **kwargs):
        """
        Executes a single pass, looking for objects to expire.

        :param args: Extra args to fulfill the Daemon interface; this daemon
                     has no additional args.
        :param kwargs: Extra keyword args to fulfill the Daemon interface; this
                       daemon accepts processes and process keyword args.
                       These will override the values from the config file if
                       provided.
        """
        processes, process = self.get_process_values(kwargs)
        pool = GreenPool(self.concurrency)
        containers_to_delete = []
        self.report_first_time = self.report_last_time = time()
        self.report_objects = 0
        try:
            self.logger.debug(_('Run begin'))
            containers, objects = \
                self.swift.get_account_info(self.s3_expiring_objects_account)
            self.logger.info(_('Pass beginning; %s possible containers; %s '
                               'possible objects') % (containers, objects))

            for c in self.swift.iter_containers(self.
                                                s3_expiring_objects_account):
                container = c['name']
                timestamp = int(container)
                if timestamp > int(time()):
                    break
                containers_to_delete.append(container)
                for o in self.swift.iter_objects(self
                                                 .s3_expiring_objects_account,
                                                 container):
                    obj = o['name'].encode('utf8')
                    if processes > 0:
                        obj_process = int(
                            hashlib.md5('%s/%s' % (container, obj)).
                            hexdigest(), 16)
                        if obj_process % processes != process:
                            continue

                    pool.spawn_n(self.delete_object, container, obj)
            pool.waitall()
            for container in containers_to_delete:
#.........这里部分代码省略.........

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]
class TestObjectExpirer(ReplProbeTest):
    def setUp(self):
        if len(ENABLED_POLICIES) < 2:
            raise SkipTest("Need more than one policy")

        self.expirer = Manager(["object-expirer"])
        self.expirer.start()
        err = self.expirer.stop()
        if err:
            raise SkipTest("Unable to verify object-expirer service")

        conf_files = []
        for server in self.expirer.servers:
            conf_files.extend(server.conf_files())
        conf_file = conf_files[0]
        self.client = InternalClient(conf_file, "probe-test", 3)

        super(TestObjectExpirer, self).setUp()
        self.container_name = "container-%s" % uuid.uuid4()
        self.object_name = "object-%s" % uuid.uuid4()
        self.brain = BrainSplitter(self.url, self.token, self.container_name, self.object_name)

    def test_expirer_object_split_brain(self):
        old_policy = random.choice(ENABLED_POLICIES)
        wrong_policy = random.choice([p for p in ENABLED_POLICIES if p != old_policy])
        # create an expiring object and a container with the wrong policy
        self.brain.stop_primary_half()
        self.brain.put_container(int(old_policy))
        self.brain.put_object(headers={"X-Delete-After": 2})
        # get the object timestamp
        metadata = self.client.get_object_metadata(
            self.account,
            self.container_name,
            self.object_name,
            headers={"X-Backend-Storage-Policy-Index": int(old_policy)},
        )
        create_timestamp = Timestamp(metadata["x-timestamp"])
        self.brain.start_primary_half()
        # get the expiring object updates in their queue, while we have all
        # the servers up
        Manager(["object-updater"]).once()
        self.brain.stop_handoff_half()
        self.brain.put_container(int(wrong_policy))
        # don't start handoff servers, only wrong policy is available

        # make sure auto-created containers get in the account listing
        Manager(["container-updater"]).once()
        # this guy should no-op since it's unable to expire the object
        self.expirer.once()

        self.brain.start_handoff_half()
        self.get_to_final_state()

        # validate object is expired
        found_in_policy = None
        metadata = self.client.get_object_metadata(
            self.account,
            self.container_name,
            self.object_name,
            acceptable_statuses=(4,),
            headers={"X-Backend-Storage-Policy-Index": int(old_policy)},
        )
        self.assertTrue("x-backend-timestamp" in metadata)
        self.assertEqual(Timestamp(metadata["x-backend-timestamp"]), create_timestamp)

        # but it is still in the listing
        for obj in self.client.iter_objects(self.account, self.container_name):
            if self.object_name == obj["name"]:
                break
        else:
            self.fail("Did not find listing for %s" % self.object_name)

        # clear proxy cache
        client.post_container(self.url, self.token, self.container_name, {})
        # run the expirier again after replication
        self.expirer.once()

        # object is not in the listing
        for obj in self.client.iter_objects(self.account, self.container_name):
            if self.object_name == obj["name"]:
                self.fail("Found listing for %s" % self.object_name)

        # and validate object is tombstoned
        found_in_policy = None
        for policy in ENABLED_POLICIES:
            metadata = self.client.get_object_metadata(
                self.account,
                self.container_name,
                self.object_name,
                acceptable_statuses=(4,),
                headers={"X-Backend-Storage-Policy-Index": int(policy)},
            )
            if "x-backend-timestamp" in metadata:
                if found_in_policy:
                    self.fail("found object in %s and also %s" % (found_in_policy, policy))
                found_in_policy = policy
                self.assertTrue("x-backend-timestamp" in metadata)
                self.assertTrue(Timestamp(metadata["x-backend-timestamp"]) > create_timestamp)

# 需要导入模块: from swift.common.internal_client import InternalClient [as 别名]
# 或者: from swift.common.internal_client.InternalClient import iter_objects [as 别名]
class ObjectTransitor(Daemon):
    def __init__(self, conf):
        super(ObjectTransitor, self).__init__(conf)
        self.conf = conf
        self.logger = get_logger(conf, log_route='s3-object-transitor')
        self.logger.set_statsd_prefix('s3-object-transitor')
        self.interval = int(conf.get('interval') or 300)
        self.s3_tr_objects_account = \
            (conf.get('auto_create_account_prefix') or '.') + \
            (conf.get('expiring_objects_account_name') or
             's3_transitioning_objects')
        conf_path = conf.get('__file__') or \
            '/etc/swift/s3-object-transitor.conf'
        request_tries = int(conf.get('request_tries') or 3)
        self.swift = InternalClient(conf_path, 'Swift Object Transitor',
                                    request_tries)
        self.report_interval = int(conf.get('report_interval') or 300)
        self.report_first_time = self.report_last_time = time()
        self.report_objects = 0
        self.recon_cache_path = conf.get('recon_cache_path',
                                         '/var/cache/swift')
        self.rcache = join(self.recon_cache_path, 'object.recon')
        self.concurrency = int(conf.get('concurrency', 1))
        if self.concurrency < 1:
            raise ValueError("concurrency must be set to at least 1")
        self.processes = int(self.conf.get('processes', 0))
        self.process = int(self.conf.get('process', 0))
        self.client = Client(self.conf.get('sentry_sdn', ''))

    def report(self, final=False):
        """
        Emits a log line report of the progress so far, or the final progress
        is final=True.

        :param final: Set to True for the last report once the expiration pass
                      has completed.
        """
        if final:
            elapsed = time() - self.report_first_time
            self.logger.info(_('Pass completed in %ds; %d objects '
                               'transitioned') %
                             (elapsed, self.report_objects))
            dump_recon_cache({'object_transition_pass': elapsed,
                              'transitioned_last_pass': self.report_objects},
                             self.rcache, self.logger)
        elif time() - self.report_last_time >= self.report_interval:
            elapsed = time() - self.report_first_time
            self.logger.info(_('Pass so far %ds; %d objects transitioned') %
                             (elapsed, self.report_objects))
            self.report_last_time = time()

    def run_once(self, *args, **kwargs):
        """
        Executes a single pass, looking for objects to expire.

        :param args: Extra args to fulfill the Daemon interface; this daemon
                     has no additional args.
        :param kwargs: Extra keyword args to fulfill the Daemon interface; this
                       daemon accepts processes and process keyword args.
                       These will override the values from the config file if
                       provided.
        """
        processes, process = self.get_process_values(kwargs)
        pool = GreenPool(self.concurrency)
        containers_to_delete = []
        self.report_first_time = self.report_last_time = time()
        self.report_objects = 0
        try:
            self.logger.debug(_('Run begin'))
            containers, objects = \
                self.swift.get_account_info(self.s3_tr_objects_account)
            self.logger.info(_('Pass beginning; %s possible containers; %s '
                               'possible objects') % (containers, objects))

            for c in self.swift.iter_containers(self.s3_tr_objects_account):
                container = c['name']
                timestamp = int(container)
                if timestamp > int(time()):
                    break
                containers_to_delete.append(container)
                for o in self.swift.iter_objects(self.s3_tr_objects_account,
                                                 container):
                    obj = o['name'].encode('utf8')
                    if processes > 0:
                        obj_process = int(
                            hashlib.md5('%s/%s' % (container, obj)).
                            hexdigest(), 16)
                        if obj_process % processes != process:
                            continue

                    pool.spawn_n(self.transition_object, container, obj)
            pool.waitall()
            for container in containers_to_delete:
                try:
                    self.swift.delete_container(self.s3_tr_objects_account,
                                                container, (2, 4))
                except (Exception, Timeout) as err:
                    report_exception(self.logger,
                                     _('Exception while deleting container %s %s') %
                                     (container, str(err)), self.client)
#.........这里部分代码省略.........