diff options
Diffstat (limited to 'collectors/python.d.plugin/postgres')
-rw-r--r-- | collectors/python.d.plugin/postgres/Makefile.inc | 13 | ||||
-rw-r--r-- | collectors/python.d.plugin/postgres/README.md | 68 | ||||
-rw-r--r-- | collectors/python.d.plugin/postgres/postgres.chart.py | 823 | ||||
-rw-r--r-- | collectors/python.d.plugin/postgres/postgres.conf | 124 |
4 files changed, 1028 insertions, 0 deletions
diff --git a/collectors/python.d.plugin/postgres/Makefile.inc b/collectors/python.d.plugin/postgres/Makefile.inc new file mode 100644 index 00000000..91a185cb --- /dev/null +++ b/collectors/python.d.plugin/postgres/Makefile.inc @@ -0,0 +1,13 @@ +# SPDX-License-Identifier: GPL-3.0-or-later + +# THIS IS NOT A COMPLETE Makefile +# IT IS INCLUDED BY ITS PARENT'S Makefile.am +# IT IS REQUIRED TO REFERENCE ALL FILES RELATIVE TO THE PARENT + +# install these files +dist_python_DATA += postgres/postgres.chart.py +dist_pythonconfig_DATA += postgres/postgres.conf + +# do not install these files, but include them in the distribution +dist_noinst_DATA += postgres/README.md postgres/Makefile.inc + diff --git a/collectors/python.d.plugin/postgres/README.md b/collectors/python.d.plugin/postgres/README.md new file mode 100644 index 00000000..e7b108d3 --- /dev/null +++ b/collectors/python.d.plugin/postgres/README.md @@ -0,0 +1,68 @@ +# postgres + +Module monitors one or more postgres servers. + +**Requirements:** + + * `python-psycopg2` package. You have to install it manually. + +Following charts are drawn: + +1. **Database size** MB + * size + +2. **Current Backend Processes** processes + * active + +3. **Write-Ahead Logging Statistics** files/s + * total + * ready + * done + +4. **Checkpoints** writes/s + * scheduled + * requested + +5. **Current connections to db** count + * connections + +6. **Tuples returned from db** tuples/s + * sequential + * bitmap + +7. **Tuple reads from db** reads/s + * disk + * cache + +8. **Transactions on db** transactions/s + * committed + * rolled back + +9. **Tuples written to db** writes/s + * inserted + * updated + * deleted + * conflicts + +10. **Locks on db** count per type + * locks + +### configuration + +```yaml +socket: + name : 'socket' + user : 'postgres' + database : 'postgres' + +tcp: + name : 'tcp' + user : 'postgres' + database : 'postgres' + host : 'localhost' + port : 5432 +``` + +When no configuration file is found, module tries to connect to TCP/IP socket: `localhost:5432`. + +--- diff --git a/collectors/python.d.plugin/postgres/postgres.chart.py b/collectors/python.d.plugin/postgres/postgres.chart.py new file mode 100644 index 00000000..7f43877c --- /dev/null +++ b/collectors/python.d.plugin/postgres/postgres.chart.py @@ -0,0 +1,823 @@ +# -*- coding: utf-8 -*- +# Description: example netdata python.d module +# Authors: facetoe, dangtranhoang +# SPDX-License-Identifier: GPL-3.0-or-later + +from copy import deepcopy + +try: + import psycopg2 + from psycopg2 import extensions + from psycopg2.extras import DictCursor + from psycopg2 import OperationalError + PSYCOPG2 = True +except ImportError: + PSYCOPG2 = False + +from bases.FrameworkServices.SimpleService import SimpleService + +# default module values +update_every = 1 +priority = 60000 +retries = 60 + +METRICS = { + 'DATABASE': [ + 'connections', + 'xact_commit', + 'xact_rollback', + 'blks_read', + 'blks_hit', + 'tup_returned', + 'tup_fetched', + 'tup_inserted', + 'tup_updated', + 'tup_deleted', + 'conflicts', + 'temp_files', + 'temp_bytes', + 'size' + ], + 'BACKENDS': [ + 'backends_active', + 'backends_idle' + ], + 'INDEX_STATS': [ + 'index_count', + 'index_size' + ], + 'TABLE_STATS': [ + 'table_size', + 'table_count' + ], + 'WAL': [ + 'written_wal', + 'recycled_wal', + 'total_wal' + ], + 'WAL_WRITES': [ + 'wal_writes' + ], + 'ARCHIVE': [ + 'ready_count', + 'done_count', + 'file_count' + ], + 'BGWRITER': [ + 'checkpoint_scheduled', + 'checkpoint_requested', + 'buffers_checkpoint', + 'buffers_clean', + 'maxwritten_clean', + 'buffers_backend', + 'buffers_alloc', + 'buffers_backend_fsync' + ], + 'LOCKS': [ + 'ExclusiveLock', + 'RowShareLock', + 'SIReadLock', + 'ShareUpdateExclusiveLock', + 'AccessExclusiveLock', + 'AccessShareLock', + 'ShareRowExclusiveLock', + 'ShareLock', + 'RowExclusiveLock' + ], + 'AUTOVACUUM': [ + 'analyze', + 'vacuum_analyze', + 'vacuum', + 'vacuum_freeze', + 'brin_summarize' + ], + 'STANDBY_DELTA': [ + 'sent_delta', + 'write_delta', + 'flush_delta', + 'replay_delta' + ], + 'REPSLOT_FILES': [ + 'replslot_wal_keep', + 'replslot_files' + ] +} + +QUERIES = { + 'WAL': """ +SELECT + count(*) as total_wal, + count(*) FILTER (WHERE type = 'recycled') AS recycled_wal, + count(*) FILTER (WHERE type = 'written') AS written_wal +FROM + (SELECT + wal.name, + pg_{0}file_name( + CASE pg_is_in_recovery() + WHEN true THEN NULL + ELSE pg_current_{0}_{1}() + END ), + CASE + WHEN wal.name > pg_{0}file_name( + CASE pg_is_in_recovery() + WHEN true THEN NULL + ELSE pg_current_{0}_{1}() + END ) THEN 'recycled' + ELSE 'written' + END AS type + FROM pg_catalog.pg_ls_dir('pg_{0}') AS wal(name) + WHERE name ~ '^[0-9A-F]{{24}}$' + ORDER BY + (pg_stat_file('pg_{0}/'||name)).modification, + wal.name DESC) sub; +""", + 'ARCHIVE': """ +SELECT + CAST(COUNT(*) AS INT) AS file_count, + CAST(COALESCE(SUM(CAST(archive_file ~ $r$\.ready$$r$ as INT)),0) AS INT) AS ready_count, + CAST(COALESCE(SUM(CAST(archive_file ~ $r$\.done$$r$ AS INT)),0) AS INT) AS done_count +FROM + pg_catalog.pg_ls_dir('pg_{0}/archive_status') AS archive_files (archive_file); +""", + 'BACKENDS': """ +SELECT + count(*) - (SELECT count(*) + FROM pg_stat_activity + WHERE state = 'idle') + AS backends_active, + (SELECT count(*) + FROM pg_stat_activity + WHERE state = 'idle') + AS backends_idle +FROM pg_stat_activity; +""", + 'TABLE_STATS': """ +SELECT + ((sum(relpages) * 8) * 1024) AS table_size, + count(1) AS table_count +FROM pg_class +WHERE relkind IN ('r', 't'); +""", + 'INDEX_STATS': """ +SELECT + ((sum(relpages) * 8) * 1024) AS index_size, + count(1) AS index_count +FROM pg_class +WHERE relkind = 'i'; +""", + 'DATABASE': """ +SELECT + datname AS database_name, + numbackends AS connections, + xact_commit AS xact_commit, + xact_rollback AS xact_rollback, + blks_read AS blks_read, + blks_hit AS blks_hit, + tup_returned AS tup_returned, + tup_fetched AS tup_fetched, + tup_inserted AS tup_inserted, + tup_updated AS tup_updated, + tup_deleted AS tup_deleted, + conflicts AS conflicts, + pg_database_size(datname) AS size, + temp_files AS temp_files, + temp_bytes AS temp_bytes +FROM pg_stat_database +WHERE datname IN %(databases)s ; +""", + 'BGWRITER': """ +SELECT + checkpoints_timed AS checkpoint_scheduled, + checkpoints_req AS checkpoint_requested, + buffers_checkpoint * current_setting('block_size')::numeric buffers_checkpoint, + buffers_clean * current_setting('block_size')::numeric buffers_clean, + maxwritten_clean, + buffers_backend * current_setting('block_size')::numeric buffers_backend, + buffers_alloc * current_setting('block_size')::numeric buffers_alloc, + buffers_backend_fsync +FROM pg_stat_bgwriter; +""", + 'LOCKS': """ +SELECT + pg_database.datname as database_name, + mode, + count(mode) AS locks_count +FROM pg_locks +INNER JOIN pg_database + ON pg_database.oid = pg_locks.database +GROUP BY datname, mode +ORDER BY datname, mode; +""", + 'FIND_DATABASES': """ +SELECT + datname +FROM pg_stat_database +WHERE + has_database_privilege( + (SELECT current_user), datname, 'connect') + AND NOT datname ~* '^template\d '; +""", + 'FIND_STANDBY': """ +SELECT + application_name +FROM pg_stat_replication +WHERE application_name IS NOT NULL +GROUP BY application_name; +""", + 'FIND_REPLICATION_SLOT': """ +SELECT slot_name +FROM pg_replication_slots; +""", + 'STANDBY_DELTA': """ +SELECT + application_name, + pg_{0}_{1}_diff( + CASE pg_is_in_recovery() + WHEN true THEN pg_last_{0}_receive_{1}() + ELSE pg_current_{0}_{1}() + END, + sent_{1}) AS sent_delta, + pg_{0}_{1}_diff( + CASE pg_is_in_recovery() + WHEN true THEN pg_last_{0}_receive_{1}() + ELSE pg_current_{0}_{1}() + END, + write_{1}) AS write_delta, + pg_{0}_{1}_diff( + CASE pg_is_in_recovery() + WHEN true THEN pg_last_{0}_receive_{1}() + ELSE pg_current_{0}_{1}() + END, + flush_{1}) AS flush_delta, + pg_{0}_{1}_diff( + CASE pg_is_in_recovery() + WHEN true THEN pg_last_{0}_receive_{1}() + ELSE pg_current_{0}_{1}() + END, + replay_{1}) AS replay_delta +FROM pg_stat_replication +WHERE application_name IS NOT NULL; +""", + 'REPSLOT_FILES': """ +WITH wal_size AS ( + SELECT + current_setting('wal_block_size')::INT * setting::INT AS val + FROM pg_settings + WHERE name = 'wal_segment_size' + ) +SELECT + slot_name, + slot_type, + replslot_wal_keep, + count(slot_file) AS replslot_files +FROM + (SELECT + slot.slot_name, + CASE + WHEN slot_file <> 'state' THEN 1 + END AS slot_file , + slot_type, + COALESCE ( + floor( + (pg_wal_lsn_diff(pg_current_wal_lsn (),slot.restart_lsn) + - (pg_walfile_name_offset (restart_lsn)).file_offset) / (s.val) + ),0) AS replslot_wal_keep + FROM pg_replication_slots slot + LEFT JOIN ( + SELECT + slot2.slot_name, + pg_ls_dir('pg_replslot/' || slot2.slot_name) AS slot_file + FROM pg_replication_slots slot2 + ) files (slot_name, slot_file) + ON slot.slot_name = files.slot_name + CROSS JOIN wal_size s + ) AS d +GROUP BY + slot_name, + slot_type, + replslot_wal_keep; +""", + 'IF_SUPERUSER': """ +SELECT current_setting('is_superuser') = 'on' AS is_superuser; +""", + 'DETECT_SERVER_VERSION': """ +SHOW server_version_num; +""", + 'AUTOVACUUM': """ +SELECT + count(*) FILTER (WHERE query LIKE 'autovacuum: ANALYZE%%') AS analyze, + count(*) FILTER (WHERE query LIKE 'autovacuum: VACUUM ANALYZE%%') AS vacuum_analyze, + count(*) FILTER (WHERE query LIKE 'autovacuum: VACUUM%%' + AND query NOT LIKE 'autovacuum: VACUUM ANALYZE%%' + AND query NOT LIKE '%%to prevent wraparound%%') AS vacuum, + count(*) FILTER (WHERE query LIKE '%%to prevent wraparound%%') AS vacuum_freeze, + count(*) FILTER (WHERE query LIKE 'autovacuum: BRIN summarize%%') AS brin_summarize +FROM pg_stat_activity +WHERE query NOT LIKE '%%pg_stat_activity%%'; +""", + 'DIFF_LSN': """ +SELECT + pg_{0}_{1}_diff( + CASE pg_is_in_recovery() + WHEN true THEN pg_last_{0}_receive_{1}() + ELSE pg_current_{0}_{1}() + END, + '0/0') as wal_writes ; +""" +} + + +QUERY_STATS = { + QUERIES['DATABASE']: METRICS['DATABASE'], + QUERIES['BACKENDS']: METRICS['BACKENDS'], + QUERIES['LOCKS']: METRICS['LOCKS'] +} + +ORDER = [ + 'db_stat_temp_files', + 'db_stat_temp_bytes', + 'db_stat_blks', + 'db_stat_tuple_returned', + 'db_stat_tuple_write', + 'db_stat_transactions', + 'db_stat_connections', + 'database_size', + 'backend_process', + 'index_count', + 'index_size', + 'table_count', + 'table_size', + 'wal', + 'wal_writes', + 'archive_wal', + 'checkpointer', + 'stat_bgwriter_alloc', + 'stat_bgwriter_checkpoint', + 'stat_bgwriter_backend', + 'stat_bgwriter_backend_fsync', + 'stat_bgwriter_bgwriter', + 'stat_bgwriter_maxwritten', + 'replication_slot', + 'standby_delta', + 'autovacuum' +] + +CHARTS = { + 'db_stat_transactions': { + 'options': [None, 'Transactions on db', 'transactions/s', 'db statistics', 'postgres.db_stat_transactions', + 'line'], + 'lines': [ + ['xact_commit', 'committed', 'incremental'], + ['xact_rollback', 'rolled back', 'incremental'] + ] + }, + 'db_stat_connections': { + 'options': [None, 'Current connections to db', 'count', 'db statistics', 'postgres.db_stat_connections', + 'line'], + 'lines': [ + ['connections', 'connections', 'absolute'] + ] + }, + 'db_stat_blks': { + 'options': [None, 'Disk blocks reads from db', 'reads/s', 'db statistics', 'postgres.db_stat_blks', 'line'], + 'lines': [ + ['blks_read', 'disk', 'incremental'], + ['blks_hit', 'cache', 'incremental'] + ] + }, + 'db_stat_tuple_returned': { + 'options': [None, 'Tuples returned from db', 'tuples/s', 'db statistics', 'postgres.db_stat_tuple_returned', + 'line'], + 'lines': [ + ['tup_returned', 'sequential', 'incremental'], + ['tup_fetched', 'bitmap', 'incremental'] + ] + }, + 'db_stat_tuple_write': { + 'options': [None, 'Tuples written to db', 'writes/s', 'db statistics', 'postgres.db_stat_tuple_write', 'line'], + 'lines': [ + ['tup_inserted', 'inserted', 'incremental'], + ['tup_updated', 'updated', 'incremental'], + ['tup_deleted', 'deleted', 'incremental'], + ['conflicts', 'conflicts', 'incremental'] + ] + }, + 'db_stat_temp_bytes': { + 'options': [None, 'Temp files written to disk', 'KB/s', 'db statistics', 'postgres.db_stat_temp_bytes', + 'line'], + 'lines': [ + ['temp_bytes', 'size', 'incremental', 1, 1024] + ] + }, + 'db_stat_temp_files': { + 'options': [None, 'Temp files written to disk', 'files', 'db statistics', 'postgres.db_stat_temp_files', + 'line'], + 'lines': [ + ['temp_files', 'files', 'incremental'] + ] + }, + 'database_size': { + 'options': [None, 'Database size', 'MB', 'database size', 'postgres.db_size', 'stacked'], + 'lines': [ + ] + }, + 'backend_process': { + 'options': [None, 'Current Backend Processes', 'processes', 'backend processes', 'postgres.backend_process', + 'line'], + 'lines': [ + ['backends_active', 'active', 'absolute'], + ['backends_idle', 'idle', 'absolute'] + ] + }, + 'index_count': { + 'options': [None, 'Total indexes', 'index', 'indexes', 'postgres.index_count', 'line'], + 'lines': [ + ['index_count', 'total', 'absolute'] + ] + }, + 'index_size': { + 'options': [None, 'Indexes size', 'MB', 'indexes', 'postgres.index_size', 'line'], + 'lines': [ + ['index_size', 'size', 'absolute', 1, 1024 * 1024] + ] + }, + 'table_count': { + 'options': [None, 'Total Tables', 'tables', 'tables', 'postgres.table_count', 'line'], + 'lines': [ + ['table_count', 'total', 'absolute'] + ] + }, + 'table_size': { + 'options': [None, 'Tables size', 'MB', 'tables', 'postgres.table_size', 'line'], + 'lines': [ + ['table_size', 'size', 'absolute', 1, 1024 * 1024] + ] + }, + 'wal': { + 'options': [None, 'Write-Ahead Logs', 'files', 'wal', 'postgres.wal', 'line'], + 'lines': [ + ['written_wal', 'written', 'absolute'], + ['recycled_wal', 'recycled', 'absolute'], + ['total_wal', 'total', 'absolute'] + ] + }, + 'wal_writes': { + 'options': [None, 'Write-Ahead Logs', 'kilobytes/s', 'wal_writes', 'postgres.wal_writes', 'line'], + 'lines': [ + ['wal_writes', 'writes', 'incremental', 1, 1024] + ] + }, + 'archive_wal': { + 'options': [None, 'Archive Write-Ahead Logs', 'files/s', 'archive wal', 'postgres.archive_wal', 'line'], + 'lines': [ + ['file_count', 'total', 'incremental'], + ['ready_count', 'ready', 'incremental'], + ['done_count', 'done', 'incremental'] + ] + }, + 'checkpointer': { + 'options': [None, 'Checkpoints', 'writes', 'checkpointer', 'postgres.checkpointer', 'line'], + 'lines': [ + ['checkpoint_scheduled', 'scheduled', 'incremental'], + ['checkpoint_requested', 'requested', 'incremental'] + ] + }, + 'stat_bgwriter_alloc': { + 'options': [None, 'Buffers allocated', 'kilobytes/s', 'bgwriter', 'postgres.stat_bgwriter_alloc', 'line'], + 'lines': [ + ['buffers_alloc', 'alloc', 'incremental', 1, 1024] + ] + }, + 'stat_bgwriter_checkpoint': { + 'options': [None, 'Buffers written during checkpoints', 'kilobytes/s', 'bgwriter', + 'postgres.stat_bgwriter_checkpoint', 'line'], + 'lines': [ + ['buffers_checkpoint', 'checkpoint', 'incremental', 1, 1024] + ] + }, + 'stat_bgwriter_backend': { + 'options': [None, 'Buffers written directly by a backend', 'kilobytes/s', 'bgwriter', + 'postgres.stat_bgwriter_backend', 'line'], + 'lines': [ + ['buffers_backend', 'backend', 'incremental', 1, 1024] + ] + }, + 'stat_bgwriter_backend_fsync': { + 'options': [None, 'Fsync by backend', 'times', 'bgwriter', 'postgres.stat_bgwriter_backend_fsync', 'line'], + 'lines': [ + ['buffers_backend_fsync', 'backend fsync', 'incremental'] + ] + }, + 'stat_bgwriter_bgwriter': { + 'options': [None, 'Buffers written by the background writer', 'kilobytes/s', 'bgwriter', + 'postgres.bgwriter_bgwriter', 'line'], + 'lines': [ + ['buffers_clean', 'clean', 'incremental', 1, 1024] + ] + }, + 'stat_bgwriter_maxwritten': { + 'options': [None, 'Too many buffers written', 'times', 'bgwriter', 'postgres.stat_bgwriter_maxwritten', + 'line'], + 'lines': [ + ['maxwritten_clean', 'maxwritten', 'incremental'] + ] + }, + 'autovacuum': { + 'options': [None, 'Autovacuum workers', 'workers', 'autovacuum', 'postgres.autovacuum', 'line'], + 'lines': [ + ['analyze', 'analyze', 'absolute'], + ['vacuum', 'vacuum', 'absolute'], + ['vacuum_analyze', 'vacuum analyze', 'absolute'], + ['vacuum_freeze', 'vacuum freeze', 'absolute'], + ['brin_summarize', 'brin summarize', 'absolute'] + ] + }, + 'standby_delta': { + 'options': [None, 'Standby delta', 'kilobytes', 'replication delta', 'postgres.standby_delta', 'line'], + 'lines': [ + ['sent_delta', 'sent delta', 'absolute', 1, 1024], + ['write_delta', 'write delta', 'absolute', 1, 1024], + ['flush_delta', 'flush delta', 'absolute', 1, 1024], + ['replay_delta', 'replay delta', 'absolute', 1, 1024] + ] + }, + 'replication_slot': { + 'options': [None, 'Replication slot files', 'files', 'replication slot', 'postgres.replication_slot', 'line'], + 'lines': [ + ['replslot_wal_keep', 'wal keeped', 'absolute'], + ['replslot_files', 'pg_replslot files', 'absolute'] + ] + } +} + + +class Service(SimpleService): + def __init__(self, configuration=None, name=None): + SimpleService.__init__(self, configuration=configuration, name=name) + self.order = ORDER[:] + self.definitions = deepcopy(CHARTS) + self.table_stats = configuration.pop('table_stats', False) + self.index_stats = configuration.pop('index_stats', False) + self.database_poll = configuration.pop('database_poll', None) + self.configuration = configuration + self.connection = False + self.server_version = None + self.data = dict() + self.locks_zeroed = dict() + self.databases = list() + self.secondaries = list() + self.replication_slots = list() + self.queries = QUERY_STATS.copy() + + def _connect(self): + params = dict(user='postgres', + database=None, + password=None, + host=None, + port=5432) + params.update(self.configuration) + + if not self.connection: + try: + self.connection = psycopg2.connect(**params) + self.connection.set_isolation_level(extensions.ISOLATION_LEVEL_AUTOCOMMIT) + self.connection.set_session(readonly=True) + except OperationalError as error: + return False, str(error) + return True, True + + def check(self): + if not PSYCOPG2: + self.error('\'python-psycopg2\' module is needed to use postgres.chart.py') + return False + result, error = self._connect() + if not result: + conf = dict((k, (lambda k, v: v if k != 'password' else '*****')(k, v)) + for k, v in self.configuration.items()) + self.error('Failed to connect to %s. Error: %s' % (str(conf), error)) + return False + try: + cursor = self.connection.cursor() + self.databases = discover_databases_(cursor, QUERIES['FIND_DATABASES']) + is_superuser = check_if_superuser_(cursor, QUERIES['IF_SUPERUSER']) + self.secondaries = discover_secondaries_(cursor, QUERIES['FIND_STANDBY']) + self.server_version = detect_server_version(cursor, QUERIES['DETECT_SERVER_VERSION']) + if self.server_version >= 94000: + self.replication_slots = discover_replication_slots_(cursor, QUERIES['FIND_REPLICATION_SLOT']) + cursor.close() + + if self.database_poll and isinstance(self.database_poll, str): + self.databases = [dbase for dbase in self.databases if dbase in self.database_poll.split()] \ + or self.databases + + self.locks_zeroed = populate_lock_types(self.databases) + self.add_additional_queries_(is_superuser) + self.create_dynamic_charts_() + return True + except Exception as error: + self.error(str(error)) + return False + + def add_additional_queries_(self, is_superuser): + + if self.server_version >= 100000: + wal = 'wal' + lsn = 'lsn' + else: + wal = 'xlog' + lsn = 'location' + self.queries[QUERIES['BGWRITER']] = METRICS['BGWRITER'] + self.queries[QUERIES['DIFF_LSN'].format(wal, lsn)] = METRICS['WAL_WRITES'] + self.queries[QUERIES['STANDBY_DELTA'].format(wal, lsn)] = METRICS['STANDBY_DELTA'] + + if self.index_stats: + self.queries[QUERIES['INDEX_STATS']] = METRICS['INDEX_STATS'] + if self.table_stats: + self.queries[QUERIES['TABLE_STATS']] = METRICS['TABLE_STATS'] + if is_superuser: + self.queries[QUERIES['ARCHIVE'].format(wal)] = METRICS['ARCHIVE'] + if self.server_version >= 90400: + self.queries[QUERIES['WAL'].format(wal, lsn)] = METRICS['WAL'] + if self.server_version >= 100000: + self.queries[QUERIES['REPSLOT_FILES']] = METRICS['REPSLOT_FILES'] + if self.server_version >= 90400: + self.queries[QUERIES['AUTOVACUUM']] = METRICS['AUTOVACUUM'] + + def create_dynamic_charts_(self): + + for database_name in self.databases[::-1]: + self.definitions['database_size']['lines'].append( + [database_name + '_size', database_name, 'absolute', 1, 1024 * 1024]) + for chart_name in [name for name in self.order if name.startswith('db_stat')]: + add_database_stat_chart_(order=self.order, definitions=self.definitions, + name=chart_name, database_name=database_name) + + add_database_lock_chart_(order=self.order, definitions=self.definitions, database_name=database_name) + + for application_name in self.secondaries[::-1]: + add_replication_delta_chart_( + order=self.order, + definitions=self.definitions, + name='standby_delta', + application_name=application_name) + + for slot_name in self.replication_slots[::-1]: + add_replication_slot_chart_( + order=self.order, + definitions=self.definitions, + name='replication_slot', + slot_name=slot_name) + + def _get_data(self): + result, _ = self._connect() + if result: + cursor = self.connection.cursor(cursor_factory=DictCursor) + try: + self.data.update(self.locks_zeroed) + for query, metrics in self.queries.items(): + self.query_stats_(cursor, query, metrics) + + except OperationalError: + self.connection = False + cursor.close() + return None + else: + cursor.close() + return self.data + else: + return None + + def query_stats_(self, cursor, query, metrics): + cursor.execute(query, dict(databases=tuple(self.databases))) + for row in cursor: + for metric in metrics: + if 'database_name' in row: + dimension_id = '_'.join([row['database_name'], metric]) + elif 'application_name' in row: + dimension_id = '_'.join([row['application_name'], metric]) + elif 'slot_name' in row: + dimension_id = '_'.join([row['slot_name'], metric]) + else: + dimension_id = metric + if metric in row: + if row[metric] is not None: + self.data[dimension_id] = int(row[metric]) + elif 'locks_count' in row: + self.data[dimension_id] = row['locks_count'] if metric == row['mode'] else 0 + + +def discover_databases_(cursor, query): + cursor.execute(query) + result = list() + for db in [database[0] for database in cursor]: + if db not in result: + result.append(db) + return result + + +def discover_secondaries_(cursor, query): + cursor.execute(query) + result = list() + for sc in [standby[0] for standby in cursor]: + if sc not in result: + result.append(sc) + return result + + +def discover_replication_slots_(cursor, query): + cursor.execute(query) + result = list() + for slot in [replication_slot[0] for replication_slot in cursor]: + if slot not in result: + result.append(slot) + return result + + +def check_if_superuser_(cursor, query): + cursor.execute(query) + return cursor.fetchone()[0] + + +def detect_server_version(cursor, query): + cursor.execute(query) + return int(cursor.fetchone()[0]) + + +def populate_lock_types(databases): + result = dict() + for database in databases: + for lock_type in METRICS['LOCKS']: + key = '_'.join([database, lock_type]) + result[key] = 0 + + return result + + +def add_database_lock_chart_(order, definitions, database_name): + def create_lines(database): + result = list() + for lock_type in METRICS['LOCKS']: + dimension_id = '_'.join([database, lock_type]) + result.append([dimension_id, lock_type, 'absolute']) + return result + + chart_name = database_name + '_locks' + order.insert(-1, chart_name) + definitions[chart_name] = { + 'options': + [None, 'Locks on db: ' + database_name, 'locks', 'db ' + database_name, 'postgres.db_locks', 'line'], + 'lines': create_lines(database_name) + } + + +def add_database_stat_chart_(order, definitions, name, database_name): + def create_lines(database, lines): + result = list() + for line in lines: + new_line = ['_'.join([database, line[0]])] + line[1:] + result.append(new_line) + return result + + chart_template = CHARTS[name] + chart_name = '_'.join([database_name, name]) + order.insert(0, chart_name) + name, title, units, _, context, chart_type = chart_template['options'] + definitions[chart_name] = { + 'options': [name, title + ': ' + database_name, units, 'db ' + database_name, context, chart_type], + 'lines': create_lines(database_name, chart_template['lines'])} + + +def add_replication_delta_chart_(order, definitions, name, application_name): + def create_lines(standby, lines): + result = list() + for line in lines: + new_line = ['_'.join([standby, line[0]])] + line[1:] + result.append(new_line) + return result + + chart_template = CHARTS[name] + chart_name = '_'.join([application_name, name]) + position = order.index('database_size') + order.insert(position, chart_name) + name, title, units, family, context, chart_type = chart_template['options'] + definitions[chart_name] = { + 'options': [name, title + ': ' + application_name, units, 'replication delta', context, chart_type], + 'lines': create_lines(application_name, chart_template['lines'])} + + +def add_replication_slot_chart_(order, definitions, name, slot_name): + def create_lines(slot, lines): + result = list() + for line in lines: + new_line = ['_'.join([slot, line[0]])] + line[1:] + result.append(new_line) + return result + + chart_template = CHARTS[name] + chart_name = '_'.join([slot_name, name]) + position = order.index('database_size') + order.insert(position, chart_name) + name, title, units, family, context, chart_type = chart_template['options'] + definitions[chart_name] = { + 'options': [name, title + ': ' + slot_name, units, 'replication slot files', context, chart_type], + 'lines': create_lines(slot_name, chart_template['lines'])} diff --git a/collectors/python.d.plugin/postgres/postgres.conf b/collectors/python.d.plugin/postgres/postgres.conf new file mode 100644 index 00000000..b69ca371 --- /dev/null +++ b/collectors/python.d.plugin/postgres/postgres.conf @@ -0,0 +1,124 @@ +# netdata python.d.plugin configuration for postgresql +# +# This file is in YaML format. Generally the format is: +# +# name: value +# +# There are 2 sections: +# - global variables +# - one or more JOBS +# +# JOBS allow you to collect values from multiple sources. +# Each source will have its own set of charts. +# +# JOB parameters have to be indented (using spaces only, example below). + +# ---------------------------------------------------------------------- +# Global Variables +# These variables set the defaults for all JOBs, however each JOB +# may define its own, overriding the defaults. + +# update_every sets the default data collection frequency. +# If unset, the python.d.plugin default is used. +# update_every: 1 + +# priority controls the order of charts at the netdata dashboard. +# Lower numbers move the charts towards the top of the page. +# If unset, the default for python.d.plugin is used. +# priority: 60000 + +# retries sets the number of retries to be made in case of failures. +# If unset, the default for python.d.plugin is used. +# Attempts to restore the service are made once every update_every +# and only if the module has collected values in the past. +# retries: 60 + +# autodetection_retry sets the job re-check interval in seconds. +# The job is not deleted if check fails. +# Attempts to start the job are made once every autodetection_retry. +# This feature is disabled by default. +# autodetection_retry: 0 + +# ---------------------------------------------------------------------- +# JOBS (data collection sources) +# +# The default JOBS share the same *name*. JOBS with the same name +# are mutually exclusive. Only one of them will be allowed running at +# any time. This allows autodetection to try several alternatives and +# pick the one that works. +# +# Any number of jobs is supported. +# +# All python.d.plugin JOBS (for all its modules) support a set of +# predefined parameters. These are: +# +# job_name: +# name: myname # the JOB's name as it will appear at the +# # dashboard (by default is the job_name) +# # JOBs sharing a name are mutually exclusive +# update_every: 1 # the JOB's data collection frequency +# priority: 60000 # the JOB's order on the dashboard +# retries: 60 # the JOB's number of restoration attempts +# autodetection_retry: 0 # the JOB's re-check interval in seconds +# +# A single connection is required in order to pull statistics. +# +# Connections can be configured with the following options: +# +# database : 'example_db_name' +# user : 'example_user' +# password : 'example_pass' +# host : 'localhost' +# port : 5432 +# +# Additionally, the following options allow selective disabling of charts +# +# table_stats : false +# index_stats : false +# database_poll : 'dbase_name1 dbase_name2' # poll only specified databases (all other will be excluded from charts) +# +# Postgres permissions are configured at its pg_hba.conf file. You can +# "trust" local clients to allow netdata to connect, or you can create +# a postgres user for netdata and add its password below to allow +# netdata connect. +# +# Postgres supported versions are : +# - 9.3 (without autovacuum) +# - 9.4 +# - 9.5 +# - 9.6 +# - 10 +# +# Superuser access is needed for theses charts: +# Write-Ahead Logs +# Archive Write-Ahead Logs +# +# Autovacuum charts is allowed since Postgres 9.4 +# ---------------------------------------------------------------------- + +socket: + name : 'local' + user : 'postgres' + database : 'postgres' + +tcp: + name : 'local' + database : 'postgres' + user : 'postgres' + host : 'localhost' + port : 5432 + +tcpipv4: + name : 'local' + database : 'postgres' + user : 'postgres' + host : '127.0.0.1' + port : 5432 + +tcpipv6: + name : 'local' + database : 'postgres' + user : 'postgres' + host : '::1' + port : 5432 + |