diff --git a/swh/archiver/director.py b/swh/archiver/director.py index d4e3c24..a9f630f 100644 --- a/swh/archiver/director.py +++ b/swh/archiver/director.py @@ -1,296 +1,287 @@ # Copyright (C) 2015-2016 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import logging import click - -from datetime import datetime +import time from swh.core import hashutil, config from swh.objstorage import PathSlicingObjStorage from swh.objstorage.api.client import RemoteObjStorage from swh.scheduler.celery_backend.config import app from . import tasks # NOQA from .storage import ArchiverStorage DEFAULT_CONFIG = { 'objstorage_type': ('str', 'local_storage'), 'objstorage_path': ('str', '/tmp/swh-storage/objects'), 'objstorage_slicing': ('str', '0:2/2:4/4:6'), 'objstorage_url': ('str', 'http://localhost:5003/'), 'batch_max_size': ('int', 50), 'archival_max_age': ('int', 3600), 'retention_policy': ('int', 2), 'asynchronous': ('bool', True), 'dbconn': ('str', 'dbname=softwareheritage-archiver-dev user=guest') } task_name = 'swh.storage.archiver.tasks.SWHArchiverTask' logger = logging.getLogger() class ArchiverDirector(): """Process the files in order to know which one is needed as backup. The archiver director processes the files in the local storage in order to know which one needs archival and it delegates this task to archiver workers. Attributes: master_objstorage: the local storage of the master server. master_objstorage_args (dict): arguments of the master objstorage initialization. archiver_storage: a wrapper for archiver db operations. db_conn_archiver: Either a libpq connection string, or a psycopg2 connection for the archiver db. slave_objstorages: Iterable of remote obj storages to the slaves servers used for backup. config: Archiver_configuration. A dictionary that must contain the following keys: objstorage_type (str): type of objstorage used (local_storage or remote_storage). If the storage is local, the arguments keys must be present objstorage_path (str): master's objstorage path objstorage_slicing (str): masters's objstorage slicing Otherwise, if it's a remote objstorage, the keys must be: objstorage_url (str): url of the remote objstorage batch_max_size (int): The number of content items that can be given to the same archiver worker. archival_max_age (int): Delay given to the worker to copy all the files in a given batch. retention_policy (int): Required number of copies for the content to be considered safe. asynchronous (boolean): Indicate whenever the archival should run in asynchronous mode or not. """ def __init__(self, db_conn_archiver, config): """ Constructor of the archiver director. Args: db_conn_archiver: Either a libpq connection string, or a psycopg2 connection for the archiver db. config: Archiver_configuration. A dictionary that must contain the following keys: objstorage_type (str): type of objstorage used (local_objstorage or remote_objstorage). If the storage is local, the arguments keys must be present objstorage_path (str): master's objstorage path objstorage_slicing (str): masters's objstorage slicing Otherwise, if it's a remote objstorage, the keys must be: objstorage_url (str): url of the remote objstorage batch_max_size (int): The number of content items that can be given to the same archiver worker. archival_max_age (int): Delay given to the worker to copy all the files in a given batch. retention_policy (int): Required number of copies for the content to be considered safe. asynchronous (boolean): Indicate whenever the archival should run in asynchronous mode or not. """ # Get the slave storages self.db_conn_archiver = db_conn_archiver self.archiver_storage = ArchiverStorage(db_conn_archiver) self.slave_objstorages = { id: url for id, url in self.archiver_storage.archive_ls() } # Check that there is enough backup servers for the retention policy if config['retention_policy'] > len(self.slave_objstorages) + 1: raise ValueError( "Can't have a retention policy of %d with %d backup servers" % (config['retention_policy'], len(self.slave_objstorages)) ) # Get the master storage that contains content to be archived if config['objstorage_type'] == 'local_objstorage': master_objstorage_args = { 'root': config['objstorage_path'], 'slicing': config['objstorage_slicing'] } master_objstorage = PathSlicingObjStorage( **master_objstorage_args ) elif config['objstorage_type'] == 'remote_objstorage': master_objstorage_args = {'base_url': config['objstorage_url']} master_objstorage = RemoteObjStorage(**master_objstorage_args) else: raise ValueError( 'Unknow objstorage class `%s`' % config['objstorage_type'] ) self.master_objstorage = master_objstorage self.master_objstorage_args = master_objstorage_args # Keep the full configuration self.config = config def run(self): """ Run the archiver director. The archiver director will check all the contents of the archiver database and do the required backup jobs. """ if self.config['asynchronous']: run_fn = self.run_async_worker else: run_fn = self.run_sync_worker for batch in self.get_unarchived_content(): run_fn(batch) def run_async_worker(self, batch): """ Produce a worker that will be added to the task queue. """ task = app.tasks[task_name] task.delay(batch, archiver_args=self.db_conn_archiver, master_objstorage_args=self.master_objstorage_args, slave_objstorages=self.slave_objstorages, config=self.config) def run_sync_worker(self, batch): """ Run synchronously a worker on the given batch. """ task = app.tasks[task_name] task(batch, archiver_args=self.db_conn_archiver, master_objstorage_args=self.master_objstorage_args, slave_objstorages=self.slave_objstorages, config=self.config) def get_unarchived_content(self): """ Get contents that need to be archived. Yields: A batch of contents. Batches are dictionaries which associates a content id to the data about servers that contains it or not. {'id1': {'present': [('slave1', 'slave1_url')], 'missing': [('slave2', 'slave2_url'), ('slave3', 'slave3_url')] }, 'id2': {'present': [], 'missing': [ ('slave1', 'slave1_url'), ('slave2', 'slave2_url'), ('slave3', 'slave3_url') ]} } Where keys (idX) are sha1 of the content and (slaveX, slaveX_url) are ids and urls of the storage slaves. At least all the content that don't have enough copies on the backups servers are distributed into these batches. """ - # Get the data about each content referenced into the archiver. - missing_copy = {} - for content_id in self.archiver_storage.content_archive_ls(): - db_content_id = '\\x' + hashutil.hash_to_hex(content_id[0]) - - # Fetch the datas about archival status of the content - backups = self.archiver_storage.content_archive_get( - content=db_content_id - ) - for _content_id, server_id, status, mtime in backups: - virtual_status = self.get_virtual_status(status, mtime) - server_data = (server_id, self.slave_objstorages[server_id]) - - missing_copy.setdefault( - db_content_id, - {'present': [], 'missing': []} - ).setdefault(virtual_status, []).append(server_data) - - # Check the content before archival. - try: - self.master_objstorage.check(content_id[0]) - except Exception as e: - # Exception can be Error or ObjNotFoundError. - logger.error(e) - # TODO Do something to restore the content? - - if len(missing_copy) >= self.config['batch_max_size']: - yield missing_copy - missing_copy = {} - - if len(missing_copy) > 0: - yield missing_copy + contents = {} + # Get the archives + archives = [(k, v) for k, v in self.archiver_storage.archive_ls()] + # Get all the contents referenced into the archiver tables + for content_id, copies in self.archiver_storage.content_archive_get(): + content_id = r'\x' + hashutil.hash_to_hex(content_id) + data = {'present': [], 'missing': []} + # For each archive server, check the current content status + for archive_id, archive_url in archives: + if archive_id not in copies: + data['missing'].append((archive_id, archive_url)) + else: + copy = copies[archive_id] + vstatus = self.get_virtual_status(copy['status'], + copy['mtime']) + data[vstatus].append((archive_id, archive_url)) + + contents[content_id] = data + + if len(contents) >= self.config['batch_max_size']: + yield contents + contents = {} + + if len(contents) > 0: + yield contents def get_virtual_status(self, status, mtime): """ Compute the virtual presence of a content. If the status is ongoing but the time is not elasped, the archiver consider it will be present in the futur, and so consider it as present. However, if the time is elasped, the copy may have failed, so consider the content as missing. Arguments: status (string): One of ('present', 'missing', 'ongoing'). The status of the content. mtime (datetime): Time at which the content have been updated for the last time. Returns: The virtual status of the studied content, which is 'present' or 'missing'. Raises: ValueError: if the status is not one 'present', 'missing' or 'ongoing' """ if status in ('present', 'missing'): return status # If the status is 'ongoing' but there is still time, another worker # may still be on the task. if status == 'ongoing': - mtime = mtime.replace(tzinfo=None) - elapsed = (datetime.now() - mtime).total_seconds() + elapsed = int(time.time()) - mtime if elapsed <= self.config['archival_max_age']: return 'present' else: return 'missing' else: raise ValueError("status must be either 'present', 'missing' " "or 'ongoing'") @click.command() @click.argument('config-path', required=1) @click.option('--dbconn', default=DEFAULT_CONFIG['dbconn'][1], help="Connection string for the archiver database") @click.option('--async/--sync', default=DEFAULT_CONFIG['asynchronous'][1], help="Indicates if the archiver should run asynchronously") def launch(config_path, dbconn, async): # The configuration have following priority : # command line > file config > default config cl_config = { 'dbconn': dbconn, 'asynchronous': async } conf = config.read(config_path, DEFAULT_CONFIG) conf.update(cl_config) # Create connection data and run the archiver. archiver = ArchiverDirector(conf['dbconn'], conf) - logger.info("Starting an archival at", datetime.now()) + logger.info("Starting an archival at", time.time()) archiver.run() if __name__ == '__main__': launch() diff --git a/swh/archiver/storage.py b/swh/archiver/storage.py index 028304c..73586e0 100644 --- a/swh/archiver/storage.py +++ b/swh/archiver/storage.py @@ -1,89 +1,82 @@ # Copyright (C) 2016 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import psycopg2 +import time -from ..common import db_transaction_generator +from ..common import db_transaction_generator, db_transaction from ..db import Db from ..exc import StorageDBError class ArchiverStorage(): """SWH Archiver storage proxy, encompassing DB """ def __init__(self, db_conn): """ Args: db_conn: either a libpq connection string, or a psycopg2 connection """ try: if isinstance(db_conn, psycopg2.extensions.connection): self.db = Db(db_conn) else: self.db = Db.connect(db_conn) except psycopg2.OperationalError as e: raise StorageDBError(e) @db_transaction_generator def archive_ls(self, cur=None): """ Get all the archives registered on the server. Yields: a tuple (server_id, server_url) for each archive server. """ yield from self.db.archive_ls(cur) @db_transaction_generator - def content_archive_ls(self, cur=None): - """ Get the archival status of the content - - Get an iterable over all the content that is referenced - in a backup server. - - Yields: - the sha1 of each content referenced at least one time - in the database of archiveal status. - - """ - yield from self.db.content_archive_ls(cur) - - @db_transaction_generator - def content_archive_get(self, content=None, archive=None, cur=None): + def content_archive_get(self, content=None, cur=None): """ Get the archival status of a content in a specific server. Retreive from the database the archival status of the given content in the given archive server. Args: content: the sha1 of the content. May be None for any id. archive: the database id of the server we're looking into may be None for any server. Yields: A tuple (content_id, server_id, archival status, mtime, tzinfo). """ - yield from self.db.content_archive_get(content, archive, cur) + yield from self.db.content_archive_get(content, cur) - @db_transaction_generator + @db_transaction def content_archive_update(self, content_id, archive_id, new_status=None, cur=None): - """ Update the status of a archive content and set it's mtime to now() + """ Update the status of an archive content and set its mtime to - Change the last modification time of an archived content and change - its status to the given one. + Change the mtime of an archived content for the given archive and set + it's mtime to the current time. Args: - content_id (string): The content id. - archive_id (string): The id of the concerned archive. - new_status (string): One of missing, ongoing or present, this - status will replace the previous one. If not given, the - function only changes the mtime of the content. + content_id (str): content sha1 + archive_id (str): name of the archive + new_status (str): one of 'missing', 'present' or 'ongoing'. + this status will replace the previous one. If not given, + the function only change the mtime of the content for the + given archive. """ - yield from self.db.content_archive_update(content_id, - archive_id, - new_status, - cur) + # FIXME check how to alter direclty the json object with postgres + # Get the data and alter it + copies = self.db.content_archive_get(content_id)['copies'] + if new_status is not None: + copies[archive_id]['status'] = new_status + copies[archive_id]['mtime'] = int(time.time()) + + # Then save the new data + self.db.content_archive_update(content_id, copies) diff --git a/swh/archiver/tests/test_archiver.py b/swh/archiver/tests/test_archiver.py index acd8f35..333f40b 100644 --- a/swh/archiver/tests/test_archiver.py +++ b/swh/archiver/tests/test_archiver.py @@ -1,275 +1,280 @@ # Copyright (C) 2015 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import tempfile import unittest import os +import time +import json from nose.tools import istest from nose.plugins.attrib import attr -from datetime import datetime, timedelta from swh.core import hashutil from swh.core.tests.db_testing import DbsTestFixture from server_testing import ServerTestFixture from swh.storage import Storage from swh.storage.archiver import ArchiverDirector, ArchiverWorker from swh.objstorage.exc import ObjNotFoundError from swh.objstorage.api.client import RemoteObjStorage from swh.objstorage.api.server import app TEST_DIR = os.path.dirname(os.path.abspath(__file__)) TEST_DATA_DIR = os.path.join(TEST_DIR, '../../../../swh-storage-testdata') @attr('db') class TestArchiver(DbsTestFixture, ServerTestFixture, unittest.TestCase): """ Test the objstorage archiver. """ TEST_DB_NAMES = [ 'softwareheritage-test', 'softwareheritage-archiver-test', ] TEST_DB_DUMPS = [ os.path.join(TEST_DATA_DIR, 'dumps/swh.dump'), os.path.join(TEST_DATA_DIR, 'dumps/swh-archiver.dump'), ] TEST_DB_DUMP_TYPES = [ 'pg_dump', 'pg_dump', ] def setUp(self): # Launch the backup server self.backup_objroot = tempfile.mkdtemp(prefix='remote') self.config = { 'storage_base': self.backup_objroot, 'storage_slicing': '0:2/2:4/4:6' } self.app = app super().setUp() # Retrieve connection (depends on the order in TEST_DB_NAMES) self.conn_storage = self.conns[0] # db connection to storage self.conn = self.conns[1] # archiver db's connection self.cursor = self.cursors[1] # a reader storage to check content has been archived self.remote_objstorage = RemoteObjStorage(self.url()) # Create the local storage. self.objroot = tempfile.mkdtemp(prefix='local') # a writer storage to store content before archiving self.storage = Storage(self.conn_storage, self.objroot) # Initializes and fill the tables. self.initialize_tables() # Create the archiver self.archiver = self.__create_director() self.storage_data = ('banco', 'http://localhost:%s/' % self.port) def tearDown(self): self.empty_tables() super().tearDown() def initialize_tables(self): """ Initializes the database with a sample of items. """ # Add an archive (update existing one for technical reason, # altering enum cannot run in a transaction...) self.cursor.execute("""UPDATE archive SET url='{}' WHERE id='banco' """.format(self.url())) self.conn.commit() def empty_tables(self): # Remove all content self.cursor.execute('DELETE FROM content_archive') self.conn.commit() - def __add_content(self, content_data, status='missing', date='now()'): - # Add the content + def __add_content(self, content_data, status='missing', date=None): + # Add the content to the storage content = hashutil.hashdata(content_data) content.update({'data': content_data}) self.storage.content_add([content]) # Then update database content_id = r'\x' + hashutil.hash_to_hex(content['sha1']) + copies = {'banco': { + 'status': status, + 'mtime': date or int(time.time()) # if date is None, use now() + }} self.cursor.execute("""INSERT INTO content_archive - VALUES('%s'::sha1, 'banco', '%s', %s) - """ % (content_id, status, date)) + VALUES('%s'::sha1, '%s') + """ % (content_id, json.dumps(copies))) return content['sha1'] def __get_missing(self): self.cursor.execute("""SELECT content_id FROM content_archive WHERE status='missing'""") return self.cursor.fetchall() def __create_director(self, batch_size=5000, archival_max_age=3600, retention_policy=1, asynchronous=False): config = { 'objstorage_type': 'local_objstorage', 'objstorage_path': self.objroot, 'objstorage_slicing': '0:2/2:4/4:6', 'batch_max_size': batch_size, 'archival_max_age': archival_max_age, 'retention_policy': retention_policy, 'asynchronous': asynchronous # Avoid depending on queue for tests. } director = ArchiverDirector(db_conn_archiver=self.conn, config=config) return director def __create_worker(self, batch={}, config={}): mobjstorage_args = self.archiver.master_objstorage_args if not config: config = self.archiver.config return ArchiverWorker(batch, archiver_args=self.conn, master_objstorage_args=mobjstorage_args, slave_objstorages=[self.storage_data], config=config) # Integration test - + @istest def archive_missing_content(self): """ Run archiver on a missing content should archive it. """ content_data = b'archive_missing_content' content_id = self.__add_content(content_data) # before, the content should not be there try: self.remote_objstorage.content_get(content_id) - except: + except ObjNotFoundError: pass + else: + self.fail('Content should not be present before archival') self.archiver.run() # now the content should be present on remote objstorage remote_data = self.remote_objstorage.content_get(content_id) - # After the run, the content should be archived after the archiver run. self.assertEquals(content_data, remote_data) @istest def archive_present_content(self): """ A content that is not 'missing' shouldn't be archived. """ id = self.__add_content(b'archive_present_content', status='present') # After the run, the content should NOT be in the archive.* self.archiver.run() with self.assertRaises(ObjNotFoundError): self.remote_objstorage.content_get(id) @istest def archive_already_enough(self): """ A content missing with enough copies shouldn't be archived. """ id = self.__add_content(b'archive_alread_enough') director = self.__create_director(retention_policy=0) director.run() with self.assertRaises(ObjNotFoundError): self.remote_objstorage.content_get(id) # Unit test for ArchiverDirector def vstatus(self, status, mtime): return self.archiver.get_virtual_status(status, mtime) @istest def vstatus_present(self): self.assertEquals( self.vstatus('present', None), 'present' ) @istest def vstatus_missing(self): self.assertEquals( self.vstatus('missing', None), 'missing' ) @istest def vstatus_ongoing_remaining(self): - current_time = datetime.now() self.assertEquals( - self.vstatus('ongoing', current_time), + self.vstatus('ongoing', int(time.time())), 'present' ) @istest def vstatus_ongoing_elapsed(self): - past_time = datetime.now() - timedelta( - seconds=self.archiver.config['archival_max_age'] + 1 + past_time = ( + int(time.time()) - self.archiver.config['archival_max_age'] - 1 ) self.assertEquals( self.vstatus('ongoing', past_time), 'missing' ) # Unit tests for archive worker @istest def need_archival_missing(self): """ A content should still need archival when it is missing. """ id = self.__add_content(b'need_archival_missing', status='missing') id = r'\x' + hashutil.hash_to_hex(id) worker = self.__create_worker() self.assertEqual(worker.need_archival(id, self.storage_data), True) @istest def need_archival_present(self): """ A content should still need archival when it is missing """ id = self.__add_content(b'need_archival_missing', status='present') id = r'\x' + hashutil.hash_to_hex(id) worker = self.__create_worker() self.assertEqual(worker.need_archival(id, self.storage_data), False) @istest def need_archival_ongoing_remaining(self): """ An ongoing archival with remaining time shouldnt need archival. """ id = self.__add_content(b'need_archival_ongoing_remaining', - status='ongoing', date="'%s'" % datetime.now()) + status='ongoing') id = r'\x' + hashutil.hash_to_hex(id) worker = self.__create_worker() self.assertEqual(worker.need_archival(id, self.storage_data), False) @istest def need_archival_ongoing_elasped(self): """ An ongoing archival with elapsed time should be scheduled again. """ id = self.__add_content( b'archive_ongoing_elapsed', status='ongoing', - date="'%s'" % (datetime.now() - timedelta( - seconds=self.archiver.config['archival_max_age'] + 1 - )) + date=( + int(time.time()) - self.archiver.config['archival_max_age'] - 1 + ) ) id = r'\x' + hashutil.hash_to_hex(id) worker = self.__create_worker() self.assertEqual(worker.need_archival(id, self.storage_data), True) @istest def content_sorting_by_archiver(self): """ Check that the content is correctly sorted. """ batch = { 'id1': { 'present': [('slave1', 'slave1_url')], 'missing': [] }, 'id2': { 'present': [], 'missing': [('slave1', 'slave1_url')] } } worker = self.__create_worker(batch=batch) mapping = worker.sort_content_by_archive() self.assertNotIn('id1', mapping[('slave1', 'slave1_url')]) self.assertIn('id2', mapping[('slave1', 'slave1_url')]) diff --git a/swh/archiver/worker.py b/swh/archiver/worker.py index e1220f1..3f93615 100644 --- a/swh/archiver/worker.py +++ b/swh/archiver/worker.py @@ -1,251 +1,250 @@ # Copyright (C) 2015 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import random import logging - -from datetime import datetime +import time from swh.objstorage import PathSlicingObjStorage from swh.objstorage.api.client import RemoteObjStorage from .storage import ArchiverStorage from .copier import ArchiverCopier logger = logging.getLogger() class ArchiverWorker(): """ Do the required backups on a given batch of contents. Process the content of a content batch in order to do the needed backups on the slaves servers. Attributes: batch: The content this worker has to archive, which is a dictionary that associates a content's sha1 id to the list of servers where the content is present or missing (see ArchiverDirector::get_unarchived_content). master_objstorage_args: The connection argument to initialize the master storage with the db connection url & the object storage path. slave_objstorages: A map that associates server_id to the remote server config: Archiver_configuration. A dictionary that must contains the following keys. objstorage_path (string): the path of the objstorage of the master. batch_max_size (int): The number of content items that can be given to the same archiver worker. archival_max_age (int): Delay given to the worker to copy all the files in a given batch. retention_policy (int): Required number of copies for the content to be considered safe. asynchronous (boolean): Indicate whenever the archival should run in asynchronous mode or not. """ def __init__(self, batch, archiver_args, master_objstorage_args, slave_objstorages, config): """ Constructor of the ArchiverWorker class. Args: batch: A batch of content, which is a dictionary that associates a content's sha1 id to the list of servers where the content is present. archiver_args: The archiver's arguments to establish connection to db. master_objstorage_args: The master storage arguments. slave_objstorages: A map that associates server_id to the remote server. config: Archiver_configuration. A dictionary that must contains the following keys. objstorage_path (string): the path of the objstorage of the master. batch_max_size (int): The number of content items that can be given to the same archiver worker. archival_max_age (int): Delay given to the worker to copy all the files in a given batch. retention_policy (int): Required number of copies for the content to be considered safe. asynchronous (boolean): Indicate whenever the archival should run in asynchronous mode or not. """ self.batch = batch self.archiver_storage = ArchiverStorage(archiver_args) self.slave_objstorages = slave_objstorages self.config = config if config['objstorage_type'] == 'local_objstorage': master_objstorage = PathSlicingObjStorage(**master_objstorage_args) else: master_objstorage = RemoteObjStorage(**master_objstorage_args) self.master_objstorage = master_objstorage def _choose_backup_servers(self, allowed_storage, backup_number): """ Choose the slave servers for archival. Choose the given amount of servers among those which don't already contain a copy of the content. Args: allowed_storage: servers when the content is not already present. backup_number (int): The number of servers we have to choose in order to fullfill the objective. """ # In case there is not enough backup servers to get all the backups # we need, just do our best. # Such situation should not happen. backup_number = min(backup_number, len(allowed_storage)) # TODO Find a better (or a good) policy to choose the backup servers. # The random choice should be equivalently distributed between # servers for a great amount of data, but don't take care of servers # capacities. return random.sample(allowed_storage, backup_number) def _get_archival_status(self, content_id, server): """ Get the archival status of the required content. Attributes: content_id (string): Sha1 of the content. server: Tuple (archive_id, archive_url) of the archive server. Returns: A dictionary that contains all the required data : 'content_id', 'archive_id', 'status', and 'mtime' """ + archive = server[0] t, = list( - self.archiver_storage.content_archive_get(content_id, server[0]) + self.archiver_storage.content_archive_get(content_id) ) return { 'content_id': t[0], - 'archive_id': t[1], - 'status': t[2], - 'mtime': t[3] + 'archive_id': archive, + 'status': t[1][archive]['status'], + 'mtime': t[1][archive]['mtime'] } def _content_archive_update(self, content_id, archive_id, new_status=None): - """ Update the status of a archive content and set it's mtime to now() + """ Update the status of a archive content and set its mtime to now. Change the last modification time of an archived content and change its status to the given one. Args: content_id (string): The content id. archive_id (string): The id of the concerned archive. new_status (string): One of missing, ongoing or present, this status will replace the previous one. If not given, the function only changes the mtime of the content. """ self.archiver_storage.content_archive_update( content_id, archive_id, new_status ) def need_archival(self, content, destination): """ Indicates whenever a content need archivage. Filter function that returns True if a given content still require to be archived. Args: content (str): Sha1 of a content. destination: Tuple (archive id, archive url). """ archival_status = self._get_archival_status( content, destination ) status = archival_status['status'] mtime = archival_status['mtime'] # If the archive is already present, no need to backup. if status == 'present': return False # If the content is ongoing but still have time, there is # another worker working on this content. elif status == 'ongoing': - mtime = mtime.replace(tzinfo=None) - elapsed = (datetime.now() - mtime).total_seconds() + elapsed = int(time.time()) - mtime if elapsed <= self.config['archival_max_age']: return False return True def sort_content_by_archive(self): """ Create a map {archive_server -> list of content) Create a mapping that associate to a archive server all the contents that needs to be archived in it by the current worker. The map is in the form of : { (archive_1, archive_1_url): [content1, content2, content_3] (archive_2, archive_2_url): [content1, content3] } Returns: The created mapping. """ slaves_copy = {} for content_id in self.batch: # Choose some servers to upload the content among the missing ones. server_data = self.batch[content_id] nb_present = len(server_data['present']) nb_backup = self.config['retention_policy'] - nb_present backup_servers = self._choose_backup_servers( server_data['missing'], nb_backup ) # Fill the map destination -> content to upload for server in backup_servers: slaves_copy.setdefault(server, []).append(content_id) return slaves_copy def run(self): """ Do the task expected from the archiver worker. Process the content in the batch, ensure that the elements still need an archival, and spawn copiers to copy files in each destinations. """ # Get a map (archive -> [contents]) slaves_copy = self.sort_content_by_archive() # At this point, re-check the archival status in order to know if the # job have been done by another worker. for destination in slaves_copy: # list() is needed because filter's result will be consumed twice. slaves_copy[destination] = list(filter( lambda content_id: self.need_archival(content_id, destination), slaves_copy[destination] )) for content_id in slaves_copy[destination]: self._content_archive_update(content_id, destination[0], new_status='ongoing') # Spawn a copier for each destination for destination in slaves_copy: try: self.run_copier(destination, slaves_copy[destination]) except: logger.error('Unable to copy a batch to %s' % destination) def run_copier(self, destination, contents): """ Run a copier in order to archive the given contents Upload the given contents to the given archive. If the process fail, the whole content is considered uncopied and remains 'ongoing', waiting to be rescheduled as there is a delay. Attributes: destination: Tuple (archive_id, archive_url) of the destination. contents: List of contents to archive. """ ac = ArchiverCopier(destination, contents, self.master_objstorage) if ac.run(): # Once the archival complete, update the database. for content_id in contents: self._content_archive_update(content_id, destination[0], new_status='present')