import unittest import os from time import sleep from .helpers.ptrack_helpers import ProbackupTest, ProbackupException import shutil module_name = 'backup' class BackupTest(ProbackupTest, unittest.TestCase): # @unittest.skip("skip") # @unittest.expectedFailure # PGPRO-707 def test_backup_modes_archive(self): """standart backup modes with ARCHIVE WAL method""" fname = self.id().split('.')[3] node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), initdb_params=['--data-checksums'], pg_options={ 'ptrack_enable': 'on'}) backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() backup_id = self.backup_node(backup_dir, 'node', node) show_backup = self.show_pb(backup_dir, 'node')[0] self.assertEqual(show_backup['status'], "OK") self.assertEqual(show_backup['backup-mode'], "FULL") # postmaster.pid and postmaster.opts shouldn't be copied excluded = True db_dir = os.path.join( backup_dir, "backups", 'node', backup_id, "database") for f in os.listdir(db_dir): if ( os.path.isfile(os.path.join(db_dir, f)) and ( f == "postmaster.pid" or f == "postmaster.opts" ) ): excluded = False self.assertEqual(excluded, True) # page backup mode page_backup_id = self.backup_node( backup_dir, 'node', node, backup_type="page") # print self.show_pb(node) show_backup = self.show_pb(backup_dir, 'node')[1] self.assertEqual(show_backup['status'], "OK") self.assertEqual(show_backup['backup-mode'], "PAGE") # Check parent backup self.assertEqual( backup_id, self.show_pb( backup_dir, 'node', backup_id=show_backup['id'])["parent-backup-id"]) # ptrack backup mode self.backup_node(backup_dir, 'node', node, backup_type="ptrack") show_backup = self.show_pb(backup_dir, 'node')[2] self.assertEqual(show_backup['status'], "OK") self.assertEqual(show_backup['backup-mode'], "PTRACK") # Check parent backup self.assertEqual( page_backup_id, self.show_pb( backup_dir, 'node', backup_id=show_backup['id'])["parent-backup-id"]) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_smooth_checkpoint(self): """full backup with smooth checkpoint""" fname = self.id().split('.')[3] node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), initdb_params=['--data-checksums']) backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() self.backup_node( backup_dir, 'node', node, options=["-C"]) self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") node.stop() # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_incremental_backup_without_full(self): """page-level backup without validated full backup""" fname = self.id().split('.')[3] node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), initdb_params=['--data-checksums'], pg_options={'ptrack_enable': 'on'}) backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() try: self.backup_node(backup_dir, 'node', node, backup_type="page") # we should die here because exception is what we expect to happen self.assertEqual( 1, 0, "Expecting Error because page backup should not be possible " "without valid full backup.\n Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( "ERROR: Valid backup on current timeline is not found. " "Create new FULL backup before an incremental one.", e.message, "\n Unexpected Error Message: {0}\n CMD: {1}".format( repr(e.message), self.cmd)) sleep(1) try: self.backup_node(backup_dir, 'node', node, backup_type="ptrack") # we should die here because exception is what we expect to happen self.assertEqual( 1, 0, "Expecting Error because page backup should not be possible " "without valid full backup.\n Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( "ERROR: Valid backup on current timeline is not found. " "Create new FULL backup before an incremental one.", e.message, "\n Unexpected Error Message: {0}\n CMD: {1}".format( repr(e.message), self.cmd)) self.assertEqual( self.show_pb(backup_dir, 'node')[0]['status'], "ERROR") # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_incremental_backup_corrupt_full(self): """page-level backup with corrupted full backup""" fname = self.id().split('.')[3] node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), initdb_params=['--data-checksums'], pg_options={'ptrack_enable': 'on'}) backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() backup_id = self.backup_node(backup_dir, 'node', node) file = os.path.join( backup_dir, "backups", "node", backup_id, "database", "postgresql.conf") os.remove(file) try: self.validate_pb(backup_dir, 'node') # we should die here because exception is what we expect to happen self.assertEqual( 1, 0, "Expecting Error because of validation of corrupted backup.\n" " Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: self.assertTrue( "INFO: Validate backups of the instance 'node'" in e.message and "WARNING: Backup file".format( file) in e.message and "is not found".format(file) in e.message and "WARNING: Backup {0} data files are corrupted".format( backup_id) in e.message and "WARNING: Some backups are not valid" in e.message, "\n Unexpected Error Message: {0}\n CMD: {1}".format( repr(e.message), self.cmd)) try: self.backup_node(backup_dir, 'node', node, backup_type="page") # we should die here because exception is what we expect to happen self.assertEqual( 1, 0, "Expecting Error because page backup should not be possible " "without valid full backup.\n Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( "ERROR: Valid backup on current timeline is not found. " "Create new FULL backup before an incremental one.", e.message, "\n Unexpected Error Message: {0}\n CMD: {1}".format( repr(e.message), self.cmd)) self.assertEqual( self.show_pb(backup_dir, 'node', backup_id)['status'], "CORRUPT") self.assertEqual( self.show_pb(backup_dir, 'node')[1]['status'], "ERROR") # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_ptrack_threads(self): """ptrack multi thread backup mode""" fname = self.id().split('.')[3] node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), initdb_params=['--data-checksums'], pg_options={'ptrack_enable': 'on'}) backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() self.backup_node( backup_dir, 'node', node, backup_type="full", options=["-j", "4"]) self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") self.backup_node( backup_dir, 'node', node, backup_type="ptrack", options=["-j", "4"]) self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_ptrack_threads_stream(self): """ptrack multi thread backup mode and stream""" fname = self.id().split('.')[3] node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'ptrack_enable': 'on'}) backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() self.backup_node( backup_dir, 'node', node, backup_type="full", options=["-j", "4", "--stream"]) self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") self.backup_node( backup_dir, 'node', node, backup_type="ptrack", options=["-j", "4", "--stream"]) self.assertEqual(self.show_pb(backup_dir, 'node')[1]['status'], "OK") # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_page_corruption_heal_via_ptrack_1(self): """make node, corrupt some page, check that backup failed""" fname = self.id().split('.')[3] node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() self.backup_node( backup_dir, 'node', node, backup_type="full", options=["-j", "4", "--stream"]) node.safe_psql( "postgres", "create table t_heap as select 1 as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,1000) i") node.safe_psql( "postgres", "CHECKPOINT;") heap_path = node.safe_psql( "postgres", "select pg_relation_filepath('t_heap')").rstrip() with open(os.path.join(node.data_dir, heap_path), "rb+", 0) as f: f.seek(9000) f.write(b"bla") f.flush() f.close self.backup_node( backup_dir, 'node', node, backup_type="full", options=["-j", "4", "--stream", "--log-level-file=verbose"]) # open log file and check with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: log_content = f.read() self.assertIn('block 1, try to fetch via SQL', log_content) self.assertIn('SELECT pg_catalog.pg_ptrack_get_block', log_content) f.close self.assertTrue( self.show_pb(backup_dir, 'node')[1]['status'] == 'OK', "Backup Status should be OK") # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_page_corruption_heal_via_ptrack_2(self): """make node, corrupt some page, check that backup failed""" fname = self.id().split('.')[3] node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() self.backup_node( backup_dir, 'node', node, backup_type="full", options=["-j", "4", "--stream"]) node.safe_psql( "postgres", "create table t_heap as select 1 as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,1000) i") node.safe_psql( "postgres", "CHECKPOINT;") heap_path = node.safe_psql( "postgres", "select pg_relation_filepath('t_heap')").rstrip() node.stop() with open(os.path.join(node.data_dir, heap_path), "rb+", 0) as f: f.seek(9000) f.write(b"bla") f.flush() f.close node.slow_start() try: self.backup_node( backup_dir, 'node', node, backup_type="full", options=["-j", "4", "--stream", '--log-level-console=LOG']) # we should die here because exception is what we expect to happen self.assertEqual( 1, 0, "Expecting Error because of page " "corruption in PostgreSQL instance.\n" " Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: if self.remote: self.assertTrue( "WARNING: File" in e.message and "try to fetch via SQL" in e.message and "WARNING: page verification failed, " "calculated checksum" in e.message and "ERROR: query failed: " "ERROR: invalid page in block" in e.message and "query was: SELECT pg_catalog.pg_ptrack_get_block_2" in e.message, "\n Unexpected Error Message: {0}\n CMD: {1}".format( repr(e.message), self.cmd)) else: self.assertTrue( "LOG: File" in e.message and "blknum" in e.message and "have wrong checksum" in e.message and "try to fetch via SQL" in e.message and "WARNING: page verification failed, " "calculated checksum" in e.message and "ERROR: query failed: " "ERROR: invalid page in block" in e.message and "query was: SELECT pg_catalog.pg_ptrack_get_block_2" in e.message, "\n Unexpected Error Message: {0}\n CMD: {1}".format( repr(e.message), self.cmd)) self.assertTrue( self.show_pb(backup_dir, 'node')[1]['status'] == 'ERROR', "Backup Status should be ERROR") # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_backup_detect_corruption(self): """make node, corrupt some page, check that backup failed""" fname = self.id().split('.')[3] node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() node.safe_psql( "postgres", "create table t_heap as select 1 as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,1000) i") node.safe_psql( "postgres", "CHECKPOINT;") heap_path = node.safe_psql( "postgres", "select pg_relation_filepath('t_heap')").rstrip() with open(os.path.join(node.data_dir, heap_path), "rb+", 0) as f: f.seek(9000) f.write(b"bla") f.flush() f.close try: self.backup_node( backup_dir, 'node', node, backup_type="full", options=["-j", "4", "--stream"]) # we should die here because exception is what we expect to happen self.assertEqual( 1, 0, "Expecting Error because tablespace mapping is incorrect" "\n Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( 'WARNING: Corruption detected in file', e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) self.assertIn( 'ERROR: Data file corruption', e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_tablespace_in_pgdata_pgpro_1376(self): """PGPRO-1376 """ fname = self.id().split('.')[3] node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() self.create_tblspace_in_node( node, 'tblspace1', tblspc_path=( os.path.join( node.data_dir, 'somedirectory', '100500')) ) self.create_tblspace_in_node( node, 'tblspace2', tblspc_path=(os.path.join(node.data_dir)) ) node.safe_psql( "postgres", "create table t_heap1 tablespace tblspace1 as select 1 as id, " "md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,1000) i") node.safe_psql( "postgres", "create table t_heap2 tablespace tblspace2 as select 1 as id, " "md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,1000) i") backup_id_1 = self.backup_node( backup_dir, 'node', node, backup_type="full", options=["-j", "4", "--stream"]) node.safe_psql( "postgres", "drop table t_heap2") node.safe_psql( "postgres", "drop tablespace tblspace2") self.backup_node( backup_dir, 'node', node, backup_type="full", options=["-j", "4", "--stream"]) pgdata = self.pgdata_content(node.data_dir) relfilenode = node.safe_psql( "postgres", "select 't_heap1'::regclass::oid" ).rstrip() list = [] for root, dirs, files in os.walk(os.path.join( backup_dir, 'backups', 'node', backup_id_1)): for file in files: if file == relfilenode: path = os.path.join(root, file) list = list + [path] # We expect that relfilenode occures only once if len(list) > 1: message = "" for string in list: message = message + string + "\n" self.assertEqual( 1, 0, "Following file copied twice by backup:\n {0}".format( message) ) node.cleanup() self.restore_node( backup_dir, 'node', node, options=["-j", "4"]) if self.paranoia: pgdata_restored = self.pgdata_content(node.data_dir) self.compare_pgdata(pgdata, pgdata_restored) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_basic_tablespace_handling(self): """ make node, take full backup, check that restore with tablespace mapping will end with error, take page backup, check that restore with tablespace mapping will end with success """ fname = self.id().split('.')[3] node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() self.backup_node( backup_dir, 'node', node, backup_type="full", options=["-j", "4", "--stream"]) tblspace1_old_path = self.get_tblspace_path(node, 'tblspace1_old') tblspace2_old_path = self.get_tblspace_path(node, 'tblspace2_old') self.create_tblspace_in_node( node, 'some_lame_tablespace') self.create_tblspace_in_node( node, 'tblspace1', tblspc_path=tblspace1_old_path) self.create_tblspace_in_node( node, 'tblspace2', tblspc_path=tblspace2_old_path) node.safe_psql( "postgres", "create table t_heap_lame tablespace some_lame_tablespace " "as select 1 as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,1000) i") node.safe_psql( "postgres", "create table t_heap2 tablespace tblspace2 as select 1 as id, " "md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,1000) i") tblspace1_new_path = self.get_tblspace_path(node, 'tblspace1_new') tblspace2_new_path = self.get_tblspace_path(node, 'tblspace2_new') node_restored = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node_restored')) node_restored.cleanup() try: self.restore_node( backup_dir, 'node', node_restored, options=[ "-j", "4", "-T", "{0}={1}".format( tblspace1_old_path, tblspace1_new_path), "-T", "{0}={1}".format( tblspace2_old_path, tblspace2_new_path)]) # we should die here because exception is what we expect to happen self.assertEqual( 1, 0, "Expecting Error because tablespace mapping is incorrect" "\n Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: self.assertTrue( 'ERROR: --tablespace-mapping option' in e.message and 'have an entry in tablespace_map file' in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) node.safe_psql( "postgres", "drop table t_heap_lame") node.safe_psql( "postgres", "drop tablespace some_lame_tablespace") self.backup_node( backup_dir, 'node', node, backup_type="delta", options=["-j", "4", "--stream"]) self.restore_node( backup_dir, 'node', node_restored, options=[ "-j", "4", "-T", "{0}={1}".format( tblspace1_old_path, tblspace1_new_path), "-T", "{0}={1}".format( tblspace2_old_path, tblspace2_new_path)]) if self.paranoia: pgdata = self.pgdata_content(node.data_dir) if self.paranoia: pgdata_restored = self.pgdata_content(node_restored.data_dir) self.compare_pgdata(pgdata, pgdata_restored) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_tablespace_handling_1(self): """ make node with tablespace A, take full backup, check that restore with tablespace mapping of tablespace B will end with error """ fname = self.id().split('.')[3] node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() tblspace1_old_path = self.get_tblspace_path(node, 'tblspace1_old') tblspace2_old_path = self.get_tblspace_path(node, 'tblspace2_old') tblspace_new_path = self.get_tblspace_path(node, 'tblspace_new') self.create_tblspace_in_node( node, 'tblspace1', tblspc_path=tblspace1_old_path) self.backup_node( backup_dir, 'node', node, backup_type="full", options=["-j", "4", "--stream"]) node_restored = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node_restored')) node_restored.cleanup() try: self.restore_node( backup_dir, 'node', node_restored, options=[ "-j", "4", "-T", "{0}={1}".format( tblspace2_old_path, tblspace_new_path)]) # we should die here because exception is what we expect to happen self.assertEqual( 1, 0, "Expecting Error because tablespace mapping is incorrect" "\n Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: self.assertTrue( 'ERROR: --tablespace-mapping option' in e.message and 'have an entry in tablespace_map file' in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_tablespace_handling_2(self): """ make node without tablespaces, take full backup, check that restore with tablespace mapping will end with error """ fname = self.id().split('.')[3] node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() tblspace1_old_path = self.get_tblspace_path(node, 'tblspace1_old') tblspace_new_path = self.get_tblspace_path(node, 'tblspace_new') self.backup_node( backup_dir, 'node', node, backup_type="full", options=["-j", "4", "--stream"]) node_restored = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node_restored')) node_restored.cleanup() try: self.restore_node( backup_dir, 'node', node_restored, options=[ "-j", "4", "-T", "{0}={1}".format( tblspace1_old_path, tblspace_new_path)]) # we should die here because exception is what we expect to happen self.assertEqual( 1, 0, "Expecting Error because tablespace mapping is incorrect" "\n Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: self.assertTrue( 'ERROR: --tablespace-mapping option' in e.message and 'have an entry in tablespace_map file' in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_drop_rel_during_backup_delta(self): """""" fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() node.safe_psql( "postgres", "create table t_heap as select i" " as id from generate_series(0,100) i") relative_path = node.safe_psql( "postgres", "select pg_relation_filepath('t_heap')").rstrip() absolute_path = os.path.join(node.data_dir, relative_path) # FULL backup self.backup_node(backup_dir, 'node', node, options=['--stream']) # DELTA backup gdb = self.backup_node( backup_dir, 'node', node, backup_type='delta', gdb=True, options=['--log-level-file=verbose']) gdb.set_breakpoint('backup_files') gdb.run_until_break() # REMOVE file node.safe_psql( "postgres", "DROP TABLE t_heap") node.safe_psql( "postgres", "CHECKPOINT") # File removed, we can proceed with backup gdb.continue_execution_until_exit() pgdata = self.pgdata_content(node.data_dir) with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: log_content = f.read() self.assertTrue( 'LOG: File "{0}" is not found'.format(absolute_path) in log_content, 'File "{0}" should be deleted but it`s not'.format(absolute_path)) node.cleanup() self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) # Physical comparison pgdata_restored = self.pgdata_content(node.data_dir) self.compare_pgdata(pgdata, pgdata_restored) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_drop_rel_during_backup_page(self): """""" fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() node.safe_psql( "postgres", "create table t_heap as select i" " as id from generate_series(0,100) i") relative_path = node.safe_psql( "postgres", "select pg_relation_filepath('t_heap')").rstrip() absolute_path = os.path.join(node.data_dir, relative_path) # FULL backup self.backup_node(backup_dir, 'node', node, options=['--stream']) # PAGE backup gdb = self.backup_node( backup_dir, 'node', node, backup_type='page', gdb=True, options=['--log-level-file=verbose']) gdb.set_breakpoint('backup_files') gdb.run_until_break() # REMOVE file os.remove(absolute_path) # File removed, we can proceed with backup gdb.continue_execution_until_exit() pgdata = self.pgdata_content(node.data_dir) with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: log_content = f.read() self.assertTrue( 'LOG: File "{0}" is not found'.format(absolute_path) in log_content, 'File "{0}" should be deleted but it`s not'.format(absolute_path)) node.cleanup() self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) # Physical comparison pgdata_restored = self.pgdata_content(node.data_dir) self.compare_pgdata(pgdata, pgdata_restored) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_drop_rel_during_backup_ptrack(self): """""" fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'ptrack_enable': 'on'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() node.safe_psql( "postgres", "create table t_heap as select i" " as id from generate_series(0,100) i") relative_path = node.safe_psql( "postgres", "select pg_relation_filepath('t_heap')").rstrip() absolute_path = os.path.join(node.data_dir, relative_path) # FULL backup self.backup_node(backup_dir, 'node', node, options=['--stream']) # PTRACK backup gdb = self.backup_node( backup_dir, 'node', node, backup_type='ptrack', gdb=True, options=['--log-level-file=verbose']) gdb.set_breakpoint('backup_files') gdb.run_until_break() # REMOVE file os.remove(absolute_path) # File removed, we can proceed with backup gdb.continue_execution_until_exit() pgdata = self.pgdata_content(node.data_dir) with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: log_content = f.read() self.assertTrue( 'LOG: File "{0}" is not found'.format(absolute_path) in log_content, 'File "{0}" should be deleted but it`s not'.format(absolute_path)) node.cleanup() self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) # Physical comparison pgdata_restored = self.pgdata_content(node.data_dir) self.compare_pgdata(pgdata, pgdata_restored) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_persistent_slot_for_stream_backup(self): """""" fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'max_wal_size': '40MB'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() node.safe_psql( "postgres", "SELECT pg_create_physical_replication_slot('slot_1')") # FULL backup self.backup_node( backup_dir, 'node', node, options=['--stream', '--slot=slot_1']) # FULL backup self.backup_node( backup_dir, 'node', node, options=['--stream', '--slot=slot_1']) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_basic_temp_slot_for_stream_backup(self): """""" fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'max_wal_size': '40MB'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() # FULL backup self.backup_node( backup_dir, 'node', node, options=['--stream', '--temp-slot']) if self.get_version(node) < self.version_to_num('10.0'): return unittest.skip('You need PostgreSQL >= 10 for this test') else: pg_receivexlog_path = self.get_bin_path('pg_receivewal') # FULL backup self.backup_node( backup_dir, 'node', node, options=['--stream', '--slot=slot_1', '--temp-slot']) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_backup_concurrent_drop_table(self): """""" fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() node.pgbench_init(scale=1) # FULL backup gdb = self.backup_node( backup_dir, 'node', node, options=['--stream', '--compress'], gdb=True) gdb.set_breakpoint('backup_data_file') gdb.run_until_break() node.safe_psql( 'postgres', 'DROP TABLE pgbench_accounts') # do checkpoint to guarantee filenode removal node.safe_psql( 'postgres', 'CHECKPOINT') gdb.remove_all_breakpoints() gdb.continue_execution_until_exit() show_backup = self.show_pb(backup_dir, 'node')[0] self.assertEqual(show_backup['status'], "OK") # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_pg_11_adjusted_wal_segment_size(self): """""" if self.pg_config_version < self.version_to_num('11.0'): return unittest.skip('You need PostgreSQL >= 11 for this test') fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=[ '--data-checksums', '--wal-segsize=64'], pg_options={ 'min_wal_size': '128MB', 'autovacuum': 'off'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() node.pgbench_init(scale=5) # FULL STREAM backup self.backup_node( backup_dir, 'node', node, options=['--stream']) pgbench = node.pgbench(options=['-T', '5', '-c', '2']) pgbench.wait() # PAGE STREAM backup self.backup_node( backup_dir, 'node', node, backup_type='page', options=['--stream']) pgbench = node.pgbench(options=['-T', '5', '-c', '2']) pgbench.wait() # DELTA STREAM backup self.backup_node( backup_dir, 'node', node, backup_type='delta', options=['--stream']) pgbench = node.pgbench(options=['-T', '5', '-c', '2']) pgbench.wait() # FULL ARCHIVE backup self.backup_node(backup_dir, 'node', node) pgbench = node.pgbench(options=['-T', '5', '-c', '2']) pgbench.wait() # PAGE ARCHIVE backup self.backup_node(backup_dir, 'node', node, backup_type='page') pgbench = node.pgbench(options=['-T', '5', '-c', '2']) pgbench.wait() # DELTA ARCHIVE backup backup_id = self.backup_node(backup_dir, 'node', node, backup_type='delta') pgdata = self.pgdata_content(node.data_dir) # delete output = self.delete_pb( backup_dir, 'node', options=[ '--expired', '--delete-wal', '--retention-redundancy=1']) # validate self.validate_pb(backup_dir) # merge self.merge_backup(backup_dir, 'node', backup_id=backup_id) # restore node.cleanup() self.restore_node( backup_dir, 'node', node, backup_id=backup_id) pgdata_restored = self.pgdata_content(node.data_dir) self.compare_pgdata(pgdata, pgdata_restored) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_sigint_handling(self): """""" fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() # FULL backup gdb = self.backup_node( backup_dir, 'node', node, gdb=True, options=['--stream', '--log-level-file=verbose']) gdb.set_breakpoint('copy_file') gdb.run_until_break() gdb.continue_execution_until_break(20) gdb.remove_all_breakpoints() gdb._execute('signal SIGINT') gdb.continue_execution_until_error() backup_id = self.show_pb(backup_dir, 'node')[0]['id'] self.assertEqual( 'ERROR', self.show_pb(backup_dir, 'node', backup_id)['status'], 'Backup STATUS should be "ERROR"') # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_sigterm_handling(self): """""" fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() # FULL backup gdb = self.backup_node( backup_dir, 'node', node, gdb=True, options=['--stream', '--log-level-file=verbose']) gdb.set_breakpoint('copy_file') gdb.run_until_break() gdb.continue_execution_until_break(20) gdb.remove_all_breakpoints() gdb._execute('signal SIGTERM') gdb.continue_execution_until_error() backup_id = self.show_pb(backup_dir, 'node')[0]['id'] self.assertEqual( 'ERROR', self.show_pb(backup_dir, 'node', backup_id)['status'], 'Backup STATUS should be "ERROR"') # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_sigquit_handling(self): """""" fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() # FULL backup gdb = self.backup_node( backup_dir, 'node', node, gdb=True, options=['--stream', '--log-level-file=verbose']) gdb.set_breakpoint('copy_file') gdb.run_until_break() gdb.continue_execution_until_break(20) gdb.remove_all_breakpoints() gdb._execute('signal SIGQUIT') gdb.continue_execution_until_error() backup_id = self.show_pb(backup_dir, 'node')[0]['id'] self.assertEqual( 'ERROR', self.show_pb(backup_dir, 'node', backup_id)['status'], 'Backup STATUS should be "ERROR"') # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_drop_table(self): """""" fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() connect_1 = node.connect("postgres") connect_1.execute( "create table t_heap as select i" " as id from generate_series(0,100) i") connect_1.commit() connect_2 = node.connect("postgres") connect_2.execute("SELECT * FROM t_heap") connect_2.commit() # DROP table connect_2.execute("DROP TABLE t_heap") connect_2.commit() # FULL backup self.backup_node( backup_dir, 'node', node, options=['--stream']) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_basic_missing_file_permissions(self): """""" fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() relative_path = node.safe_psql( "postgres", "select pg_relation_filepath('pg_class')").rstrip() full_path = os.path.join(node.data_dir, relative_path) os.chmod(full_path, 000) try: # FULL backup self.backup_node( backup_dir, 'node', node, options=['--stream']) # we should die here because exception is what we expect to happen self.assertEqual( 1, 0, "Expecting Error because of missing permissions" "\n Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( 'ERROR: cannot open file', e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) os.chmod(full_path, 700) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_basic_missing_dir_permissions(self): """""" fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() full_path = os.path.join(node.data_dir, 'pg_twophase') os.chmod(full_path, 000) try: # FULL backup self.backup_node( backup_dir, 'node', node, options=['--stream']) # we should die here because exception is what we expect to happen self.assertEqual( 1, 0, "Expecting Error because of missing permissions" "\n Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( 'ERROR: Cannot open directory', e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) os.chmod(full_path, 700) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_backup_with_least_privileges_role(self): """""" fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, # pg_options={'ptrack_enable': 'on'}, initdb_params=['--data-checksums'], ) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() node.safe_psql( 'postgres', 'CREATE DATABASE backupdb') node.safe_psql( 'backupdb', "REVOKE TEMPORARY ON DATABASE backupdb FROM PUBLIC;" "REVOKE ALL on SCHEMA public from PUBLIC; " "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " "REVOKE ALL on SCHEMA pg_catalog from PUBLIC; " "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " "REVOKE ALL on SCHEMA information_schema from PUBLIC; " "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " "CREATE ROLE backup WITH LOGIN REPLICATION; " "GRANT CONNECT ON DATABASE backupdb to backup; " "GRANT USAGE ON SCHEMA pg_catalog TO backup; " "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " # for partial restore, checkdb and ptrack "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean, boolean) TO backup; " # for exclusive backup for PG 9.5 and ptrack "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup() TO backup; " "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" ) # ptrack functions # for fname in [ # 'oideq(oid, oid)', # 'ptrack_version()', # 'pg_ptrack_clear()', # 'pg_ptrack_control_lsn()', # 'pg_ptrack_get_and_clear_db(oid, oid)', # 'pg_ptrack_get_and_clear(oid, oid)', # 'pg_ptrack_get_block_2(oid, oid, oid, bigint)']: # try: # node.safe_psql( # "backupdb", # "GRANT EXECUTE ON FUNCTION pg_catalog.{0} " # "TO backup".format(fname)) # except: # pass # FULL backup self.backup_node( backup_dir, 'node', node, datname='backupdb', options=['--stream', '-U', 'backup']) self.backup_node( backup_dir, 'node', node, datname='backupdb', options=['-U', 'backup']) # PAGE self.backup_node( backup_dir, 'node', node, backup_type='page', datname='backupdb', options=['-U', 'backup']) self.backup_node( backup_dir, 'node', node, backup_type='page', datname='backupdb', options=['--stream', '-U', 'backup']) # DELTA self.backup_node( backup_dir, 'node', node, backup_type='delta', datname='backupdb', options=['-U', 'backup']) self.backup_node( backup_dir, 'node', node, backup_type='delta', datname='backupdb', options=['--stream', '-U', 'backup']) # PTRACK # self.backup_node( # backup_dir, 'node', node, backup_type='ptrack', # datname='backupdb', options=['-U', 'backup']) # self.backup_node( # backup_dir, 'node', node, backup_type='ptrack', # datname='backupdb', options=['--stream', '-U', 'backup']) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_parent_choosing(self): """ PAGE3 <- RUNNING(parent should be FULL) PAGE2 <- OK PAGE1 <- CORRUPT FULL """ fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() full_id = self.backup_node(backup_dir, 'node', node) # PAGE1 page1_id = self.backup_node( backup_dir, 'node', node, backup_type='page') # PAGE2 page2_id = self.backup_node( backup_dir, 'node', node, backup_type='page') # Change PAGE1 to ERROR self.change_backup_status(backup_dir, 'node', page1_id, 'ERROR') # PAGE3 page3_id = self.backup_node( backup_dir, 'node', node, backup_type='page', options=['--log-level-file=LOG']) log_file_path = os.path.join(backup_dir, 'log', 'pg_probackup.log') with open(log_file_path) as f: log_file_content = f.read() self.assertIn( "WARNING: Backup {0} has invalid parent: {1}. " "Cannot be a parent".format(page2_id, page1_id), log_file_content) self.assertIn( "WARNING: Backup {0} has status: ERROR. " "Cannot be a parent".format(page1_id), log_file_content) self.assertIn( "Parent backup: {0}".format(full_id), log_file_content) self.assertEqual( self.show_pb( backup_dir, 'node', backup_id=page3_id)['parent-backup-id'], full_id) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_parent_choosing_1(self): """ PAGE3 <- RUNNING(parent should be FULL) PAGE2 <- OK PAGE1 <- (missing) FULL """ fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() full_id = self.backup_node(backup_dir, 'node', node) # PAGE1 page1_id = self.backup_node( backup_dir, 'node', node, backup_type='page') # PAGE2 page2_id = self.backup_node( backup_dir, 'node', node, backup_type='page') # Delete PAGE1 shutil.rmtree( os.path.join(backup_dir, 'backups', 'node', page1_id)) # PAGE3 page3_id = self.backup_node( backup_dir, 'node', node, backup_type='page', options=['--log-level-file=LOG']) log_file_path = os.path.join(backup_dir, 'log', 'pg_probackup.log') with open(log_file_path) as f: log_file_content = f.read() self.assertIn( "WARNING: Backup {0} has missing parent: {1}. " "Cannot be a parent".format(page2_id, page1_id), log_file_content) self.assertIn( "Parent backup: {0}".format(full_id), log_file_content) self.assertEqual( self.show_pb( backup_dir, 'node', backup_id=page3_id)['parent-backup-id'], full_id) # Clean after yourself self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_parent_choosing_2(self): """ PAGE3 <- RUNNING(backup should fail) PAGE2 <- OK PAGE1 <- OK FULL <- (missing) """ fname = self.id().split('.')[3] backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') node = self.make_simple_node( base_dir=os.path.join(module_name, fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() full_id = self.backup_node(backup_dir, 'node', node) # PAGE1 page1_id = self.backup_node( backup_dir, 'node', node, backup_type='page') # PAGE2 page2_id = self.backup_node( backup_dir, 'node', node, backup_type='page') # Delete FULL shutil.rmtree( os.path.join(backup_dir, 'backups', 'node', full_id)) # PAGE3 try: self.backup_node( backup_dir, 'node', node, backup_type='page', options=['--log-level-file=LOG']) # we should die here because exception is what we expect to happen self.assertEqual( 1, 0, "Expecting Error because FULL backup is missing" "\n Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( 'WARNING: Failed to find a valid backup chain', e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) self.assertIn( 'ERROR: Valid backup on current timeline is not found. ' 'Create new FULL backup before an incremental one.', e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) self.assertEqual( self.show_pb( backup_dir, 'node')[2]['status'], 'ERROR') # Clean after yourself self.del_test_dir(module_name, fname)