diff --git a/doc/pgprobackup.xml b/doc/pgprobackup.xml index a2161f0c..91ec1253 100644 --- a/doc/pgprobackup.xml +++ b/doc/pgprobackup.xml @@ -4055,10 +4055,10 @@ pg_probackup archive-push -B backup_dir --instance - To speed up archiving, you can specify the option - to run archive-push on multiple threads. - If you provide the option, WAL files - will be copied in batches of the specified size. + To speed up archiving, you can specify the option + to copy WAL segments in batches of the specified size. + If option is used, then you can also specify + the option to copy the batch of WAL segments on multiple threads. WAL segments copied to the archive are synced to disk unless @@ -4096,10 +4096,10 @@ pg_probackup archive-get -B backup_dir --instance - To speed up recovery, you can specify the option - to run archive-get on multiple threads. - If you provide the option, WAL segments - will be copied in batches of the specified size. + To speed up recovery, you can specify the option + to copy WAL segments in batches of the specified size. + If option is used, then you can also specify + the option to copy the batch of WAL segments on multiple threads. diff --git a/src/archive.c b/src/archive.c index 1055948b..a712c102 100644 --- a/src/archive.c +++ b/src/archive.c @@ -1109,7 +1109,7 @@ do_archive_get(InstanceConfig *instance, const char *prefetch_dir_arg, { /* discard prefetch */ // n_fetched = 0; - rmtree(prefetch_dir, false); + pgut_rmtree(prefetch_dir, false, false); } } else diff --git a/src/backup.c b/src/backup.c index 3dfdeee9..0887a217 100644 --- a/src/backup.c +++ b/src/backup.c @@ -191,7 +191,7 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo, bool no_sync, bool { /* try to setup multi-timeline backup chain */ elog(WARNING, "Valid backup on current timeline %u is not found, " - "try to look up on previous timelines", + "trying to look up on previous timelines", current.tli); tli_list = catalog_get_timelines(&instance_config); diff --git a/src/delete.c b/src/delete.c index 5f422c30..a6263b63 100644 --- a/src/delete.c +++ b/src/delete.c @@ -985,7 +985,7 @@ do_delete_instance(void) parray_free(backup_list); /* Delete all wal files. */ - rmtree(arclog_path, false); + pgut_rmtree(arclog_path, false, true); /* Delete backup instance config file */ join_path_components(instance_config_path, backup_instance_path, BACKUP_CATALOG_CONF_FILE); diff --git a/src/pg_probackup.h b/src/pg_probackup.h index 9c590365..e202c39b 100644 --- a/src/pg_probackup.h +++ b/src/pg_probackup.h @@ -1038,6 +1038,8 @@ extern int fio_send_file(const char *from_fullpath, const char *to_fullpath, FIL extern void fio_list_dir(parray *files, const char *root, bool exclude, bool follow_symlink, bool add_root, bool backup_logs, int external_dir_num); +extern bool pgut_rmtree(const char *path, bool rmtopdir, bool strict); + /* return codes for fio_send_pages() and fio_send_file() */ #define SEND_OK (0) #define FILE_MISSING (-1) diff --git a/src/utils/pgut.c b/src/utils/pgut.c index b7599816..6d996f47 100644 --- a/src/utils/pgut.c +++ b/src/utils/pgut.c @@ -43,6 +43,9 @@ static void on_interrupt(void); static void on_cleanup(void); static pqsigfunc oldhandler = NULL; +static char ** pgut_pgfnames(const char *path, bool strict); +static void pgut_pgfnames_cleanup(char **filenames); + void discard_response(PGconn *conn); void @@ -1062,3 +1065,140 @@ discard_response(PGconn *conn) PQclear(res); } while (res); } + +/* + * pgfnames + * + * return a list of the names of objects in the argument directory. Caller + * must call pgfnames_cleanup later to free the memory allocated by this + * function. + */ +char ** +pgut_pgfnames(const char *path, bool strict) +{ + DIR *dir; + struct dirent *file; + char **filenames; + int numnames = 0; + int fnsize = 200; /* enough for many small dbs */ + + dir = opendir(path); + if (dir == NULL) + { + elog(strict ? ERROR : WARNING, "could not open directory \"%s\": %m", path); + return NULL; + } + + filenames = (char **) palloc(fnsize * sizeof(char *)); + + while (errno = 0, (file = readdir(dir)) != NULL) + { + if (strcmp(file->d_name, ".") != 0 && strcmp(file->d_name, "..") != 0) + { + if (numnames + 1 >= fnsize) + { + fnsize *= 2; + filenames = (char **) repalloc(filenames, + fnsize * sizeof(char *)); + } + filenames[numnames++] = pstrdup(file->d_name); + } + } + + if (errno) + { + elog(strict ? ERROR : WARNING, "could not read directory \"%s\": %m", path); + return NULL; + } + + filenames[numnames] = NULL; + + if (closedir(dir)) + { + elog(strict ? ERROR : WARNING, "could not close directory \"%s\": %m", path); + return NULL; + } + + return filenames; +} + +/* + * pgfnames_cleanup + * + * deallocate memory used for filenames + */ +void +pgut_pgfnames_cleanup(char **filenames) +{ + char **fn; + + for (fn = filenames; *fn; fn++) + pfree(*fn); + + pfree(filenames); +} + +/* Shamelessly stolen from commom/rmtree.c */ +bool +pgut_rmtree(const char *path, bool rmtopdir, bool strict) +{ + bool result = true; + char pathbuf[MAXPGPATH]; + char **filenames; + char **filename; + struct stat statbuf; + + /* + * we copy all the names out of the directory before we start modifying + * it. + */ + filenames = pgut_pgfnames(path, strict); + + if (filenames == NULL) + return false; + + /* now we have the names we can start removing things */ + for (filename = filenames; *filename; filename++) + { + snprintf(pathbuf, MAXPGPATH, "%s/%s", path, *filename); + + if (lstat(pathbuf, &statbuf) != 0) + { + elog(strict ? ERROR : WARNING, "could not stat file or directory \"%s\": %m", pathbuf); + result = false; + break; + } + + if (S_ISDIR(statbuf.st_mode)) + { + /* call ourselves recursively for a directory */ + if (!pgut_rmtree(pathbuf, true, strict)) + { + result = false; + break; + } + } + else + { + if (unlink(pathbuf) != 0) + { + elog(strict ? ERROR : WARNING, "could not remove file or directory \"%s\": %m", pathbuf); + result = false; + break; + } + } + } + + if (rmtopdir) + { + if (rmdir(path) != 0) + { + elog(strict ? ERROR : WARNING, "could not remove file or directory \"%s\": %m", path); + result = false; + } + } + + pgut_pgfnames_cleanup(filenames); + + return result; +} diff --git a/tests/backup.py b/tests/backup.py index 814bf684..d99cfe10 100644 --- a/tests/backup.py +++ b/tests/backup.py @@ -147,10 +147,9 @@ class BackupTest(ProbackupTest, unittest.TestCase): "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 1 is not found. " - "Create new FULL backup before an incremental one.", - e.message, + self.assertTrue( + "WARNING: Valid backup on current timeline 1 is not found" in e.message and + "ERROR: Create new full backup before an incremental one" in e.message, "\n Unexpected Error Message: {0}\n CMD: {1}".format( repr(e.message), self.cmd)) @@ -165,10 +164,9 @@ class BackupTest(ProbackupTest, unittest.TestCase): "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 1 is not found. " - "Create new FULL backup before an incremental one.", - e.message, + self.assertTrue( + "WARNING: Valid backup on current timeline 1 is not found" in e.message and + "ERROR: Create new full backup before an incremental one" in e.message, "\n Unexpected Error Message: {0}\n CMD: {1}".format( repr(e.message), self.cmd)) @@ -176,6 +174,10 @@ class BackupTest(ProbackupTest, unittest.TestCase): self.show_pb(backup_dir, 'node')[0]['status'], "ERROR") + self.assertEqual( + self.show_pb(backup_dir, 'node')[1]['status'], + "ERROR") + # Clean after yourself self.del_test_dir(module_name, fname) @@ -315,10 +317,8 @@ class BackupTest(ProbackupTest, unittest.TestCase): self.del_test_dir(module_name, fname) # @unittest.skip("skip") - def test_page_corruption_heal_via_ptrack_1(self): + def test_page_detect_corruption(self): """make node, corrupt some page, check that backup failed""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') fname = self.id().split('.')[3] node = self.make_simple_node( @@ -380,98 +380,6 @@ class BackupTest(ProbackupTest, unittest.TestCase): # 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""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - ptrack_enable=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() - - if node.major_version >= 12: - node.safe_psql( - "postgres", - "CREATE EXTENSION ptrack WITH SCHEMA pg_catalog") - - 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 shared buffer" 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" 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 shared buffer" 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" 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): @@ -495,6 +403,10 @@ class BackupTest(ProbackupTest, unittest.TestCase): "postgres", "create extension ptrack") + 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, " @@ -529,10 +441,6 @@ class BackupTest(ProbackupTest, unittest.TestCase): node.slow_start() - # self.backup_node( - # backup_dir, 'node', node, - # backup_type="full", options=["-j", "4", "--stream"]) - try: self.backup_node( backup_dir, 'node', node, @@ -608,12 +516,11 @@ class BackupTest(ProbackupTest, unittest.TestCase): "\n Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: - self.assertTrue( - 'WARNING: page verification failed, ' - 'calculated checksum' in e.message and - 'ERROR: query failed: ERROR: ' - 'invalid page in block 1 of relation' in e.message and - 'ERROR: Data files transferring failed' in e.message, + self.assertIn( + 'ERROR: Corruption detected in file "{0}", block 1: ' + 'page verification failed, calculated checksum'.format( + heap_fullpath), + e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) diff --git a/tests/delta.py b/tests/delta.py index 4447b9a8..fdbaf127 100644 --- a/tests/delta.py +++ b/tests/delta.py @@ -1050,163 +1050,6 @@ class DeltaTest(ProbackupTest, unittest.TestCase): # Clean after yourself self.del_test_dir(module_name, fname) - # @unittest.skip("skip") - def test_delta_corruption_heal_via_ptrack(self): - """make node, corrupt some page, check that backup failed""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - 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() - - if node.major_version >= 12: - node.safe_psql( - "postgres", - "CREATE EXTENSION ptrack WITH SCHEMA pg_catalog") - - 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="delta", - 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 shared buffer', 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(self): - """make node, corrupt some page, check that backup failed""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - 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() - - if node.major_version >= 12: - node.safe_psql( - "postgres", - "CREATE EXTENSION ptrack WITH SCHEMA pg_catalog") - - 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="delta", - 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( - "LOG: File" in e.message and - "try to fetch via shared buffer" 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" in e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - else: - self.assertTrue( - "WARNING: File" in e.message and - "blknum" in e.message and - "have wrong checksum" in e.message and - "try to fetch via shared buffer" 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" 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) - def test_delta_nullified_heap_page_backup(self): """ make node, take full backup, nullify some heap block, diff --git a/tests/exclude.py b/tests/exclude.py index 0402c978..c9efe22a 100644 --- a/tests/exclude.py +++ b/tests/exclude.py @@ -153,3 +153,85 @@ class ExcludeTest(ProbackupTest, unittest.TestCase): # Clean after yourself self.del_test_dir(module_name, fname) + + # @unittest.skip("skip") + def test_exclude_log_dir(self): + """ + check that by default 'log' and 'pg_log' directories are not backed up + """ + 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={ + 'logging_collector': 'on', + 'log_filename': 'postgresql.log'}) + + 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=['--stream']) + + log_dir = node.safe_psql( + 'postgres', + 'show log_directory').rstrip() + + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, options=["-j", "4"]) + + # check that PGDATA/log or PGDATA/pg_log do not exists + path = os.path.join(node.data_dir, log_dir) + log_file = os.path.join(path, 'postgresql.log') + self.assertTrue(os.path.exists(path)) + self.assertFalse(os.path.exists(log_file)) + + # Clean after yourself + self.del_test_dir(module_name, fname) + + # @unittest.skip("skip") + def test_exclude_log_dir_1(self): + """ + check that "--backup-pg-log" works correctly + """ + 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={ + 'logging_collector': 'on', + 'log_filename': 'postgresql.log'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + log_dir = node.safe_psql( + 'postgres', + 'show log_directory').rstrip() + + self.backup_node( + backup_dir, 'node', node, + backup_type='full', options=['--stream', '--backup-pg-log']) + + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, options=["-j", "4"]) + + # check that PGDATA/log or PGDATA/pg_log do not exists + path = os.path.join(node.data_dir, log_dir) + log_file = os.path.join(path, 'postgresql.log') + self.assertTrue(os.path.exists(path)) + self.assertTrue(os.path.exists(log_file)) + + # Clean after yourself + self.del_test_dir(module_name, fname)