psycopg2/tests/test_replication.py

277 lines
8.8 KiB
Python
Raw Permalink Normal View History

2015-10-15 19:01:43 +03:00
#!/usr/bin/env python
# test_replication.py - unit test for replication protocol
#
# Copyright (C) 2015-2019 Daniele Varrazzo <daniele.varrazzo@gmail.com>
# Copyright (C) 2020-2021 The Psycopg Team
2015-10-15 19:01:43 +03:00
#
# psycopg2 is free software: you can redistribute it and/or modify it
# under the terms of the GNU Lesser General Public License as published
# by the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# In addition, as a special exception, the copyright holders give
# permission to link this program with the OpenSSL library (or with
# modified versions of OpenSSL that use the same license as OpenSSL),
# and distribute linked combinations including the two.
#
# You must obey the GNU Lesser General Public License in all respects for
# all of the code used other than OpenSSL.
#
# psycopg2 is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public
# License for more details.
import time
from select import select
2015-10-15 19:01:43 +03:00
import psycopg2
from psycopg2 import sql
2016-10-11 02:10:53 +03:00
from psycopg2.extras import (
PhysicalReplicationConnection, LogicalReplicationConnection, StopReplication)
2015-10-15 19:01:43 +03:00
from . import testconfig
import unittest
from .testutils import ConnectingTestCase
from .testutils import skip_before_postgres, skip_if_green
2016-12-25 21:00:30 +03:00
skip_repl_if_green = skip_if_green("replication not supported in green mode")
2015-10-15 19:01:43 +03:00
class ReplicationTestCase(ConnectingTestCase):
def setUp(self):
super().setUp()
2015-10-19 18:02:18 +03:00
self.slot = testconfig.repl_slot
2015-10-15 19:01:43 +03:00
self._slots = []
def tearDown(self):
# first close all connections, as they might keep the slot(s) active
super().tearDown()
2015-10-15 19:01:43 +03:00
time.sleep(0.025) # sometimes the slot is still active, wait a little
2015-10-15 19:01:43 +03:00
if self._slots:
kill_conn = self.connect()
2015-10-15 19:01:43 +03:00
if kill_conn:
kill_cur = kill_conn.cursor()
for slot in self._slots:
kill_cur.execute("SELECT pg_drop_replication_slot(%s)", (slot,))
kill_conn.commit()
2015-10-15 19:01:43 +03:00
kill_conn.close()
2015-10-19 18:02:18 +03:00
def create_replication_slot(self, cur, slot_name=testconfig.repl_slot, **kwargs):
2015-10-15 19:01:43 +03:00
cur.create_replication_slot(slot_name, **kwargs)
self._slots.append(slot_name)
2015-10-19 18:02:18 +03:00
def drop_replication_slot(self, cur, slot_name=testconfig.repl_slot):
2015-10-15 19:01:43 +03:00
cur.drop_replication_slot(slot_name)
self._slots.remove(slot_name)
2015-10-19 18:02:18 +03:00
# generate some events for our replication stream
def make_replication_events(self):
conn = self.connect()
2016-10-11 02:10:53 +03:00
if conn is None:
return
2015-10-19 18:02:18 +03:00
cur = conn.cursor()
try:
cur.execute("DROP TABLE dummy1")
except psycopg2.ProgrammingError:
conn.rollback()
2016-10-11 02:10:53 +03:00
cur.execute(
"CREATE TABLE dummy1 AS SELECT * FROM generate_series(1, 5) AS id")
2015-10-19 18:02:18 +03:00
conn.commit()
2015-10-15 19:01:43 +03:00
class ReplicationTest(ReplicationTestCase):
@skip_before_postgres(9, 0)
def test_physical_replication_connection(self):
conn = self.repl_connect(connection_factory=PhysicalReplicationConnection)
2016-10-11 02:10:53 +03:00
if conn is None:
return
2015-10-15 19:01:43 +03:00
cur = conn.cursor()
cur.execute("IDENTIFY_SYSTEM")
cur.fetchall()
2016-12-25 19:46:11 +03:00
@skip_before_postgres(9, 0)
def test_datestyle(self):
if testconfig.repl_dsn is None:
return self.skipTest("replication tests disabled by default")
conn = self.repl_connect(
dsn=testconfig.repl_dsn, options='-cdatestyle=german',
connection_factory=PhysicalReplicationConnection)
if conn is None:
return
cur = conn.cursor()
cur.execute("IDENTIFY_SYSTEM")
cur.fetchall()
2015-10-15 19:01:43 +03:00
@skip_before_postgres(9, 4)
def test_logical_replication_connection(self):
conn = self.repl_connect(connection_factory=LogicalReplicationConnection)
2016-10-11 02:10:53 +03:00
if conn is None:
return
2015-10-15 19:01:43 +03:00
cur = conn.cursor()
cur.execute("IDENTIFY_SYSTEM")
cur.fetchall()
2016-10-11 02:10:53 +03:00
@skip_before_postgres(9, 4) # slots require 9.4
2015-10-15 19:01:43 +03:00
def test_create_replication_slot(self):
conn = self.repl_connect(connection_factory=PhysicalReplicationConnection)
2016-10-11 02:10:53 +03:00
if conn is None:
return
2015-10-15 19:01:43 +03:00
cur = conn.cursor()
2015-10-19 18:02:18 +03:00
self.create_replication_slot(cur)
2016-10-11 02:10:53 +03:00
self.assertRaises(
psycopg2.ProgrammingError, self.create_replication_slot, cur)
2015-10-15 19:01:43 +03:00
2016-10-11 02:10:53 +03:00
@skip_before_postgres(9, 4) # slots require 9.4
2016-12-25 21:00:30 +03:00
@skip_repl_if_green
2015-10-16 17:36:03 +03:00
def test_start_on_missing_replication_slot(self):
conn = self.repl_connect(connection_factory=PhysicalReplicationConnection)
2016-10-11 02:10:53 +03:00
if conn is None:
return
2015-10-16 17:36:03 +03:00
cur = conn.cursor()
2016-10-11 02:10:53 +03:00
self.assertRaises(psycopg2.ProgrammingError,
cur.start_replication, self.slot)
2015-10-16 17:36:03 +03:00
2015-10-19 18:02:18 +03:00
self.create_replication_slot(cur)
cur.start_replication(self.slot)
2015-10-16 17:36:03 +03:00
@skip_before_postgres(9, 4) # slots require 9.4
@skip_repl_if_green
def test_start_replication_expert_sql(self):
conn = self.repl_connect(connection_factory=LogicalReplicationConnection)
if conn is None:
return
cur = conn.cursor()
self.create_replication_slot(cur, output_plugin='test_decoding')
cur.start_replication_expert(
sql.SQL("START_REPLICATION SLOT {slot} LOGICAL 0/00000000").format(
slot=sql.Identifier(self.slot)))
2016-10-11 02:10:53 +03:00
@skip_before_postgres(9, 4) # slots require 9.4
2016-12-25 21:00:30 +03:00
@skip_repl_if_green
def test_start_and_recover_from_error(self):
conn = self.repl_connect(connection_factory=LogicalReplicationConnection)
2016-10-11 02:10:53 +03:00
if conn is None:
return
cur = conn.cursor()
self.create_replication_slot(cur, output_plugin='test_decoding')
Smart replication feedback This commit makes psycopg2 responsible for sending the status update (feedback) messages to the server regardless of whether a synchronous or asynchronous connection is used. Feedback is sent every *status_update* (default value is 10) seconds, which could be configured by passing a corresponding parameter to the `start_replication()` or `start_replication_expert()` methods. The actual feedback message is sent by the `pq_read_replication_message()` when the *status_update* timeout is reached. The default behavior of the `send_feedback()` method is changed. It doesn't send a feedback message on every call anymore but just updates internal structures. There is still a way to *force* sending a message if *force* or *reply* parameters are set. The new approach has certain advantages: 1. The client can simply call the `send_feedback()` for every processed message and the library will take care of not overwhelming the server. Actually, in the synchronous mode it is even mandatory to confirm every processed message. 2. The library tracks internally the pointer of the last received message which is not keepalive. If the client confirmed the last message and after that server sends only keepalives with increasing *wal_end*, the library can safely move forward *flush* position to the *wal_end* and later automatically report it to the server. Reporting of the *wal_end* received from keepalive messages is very important. Not doing so casing: 1. Excessive disk usage, because the replication slot prevents from WAL being cleaned up. 2. The smart and fast shutdown of the server could last indefinitely because walsender waits until the client report *flush* position equal to the *wal_end*. This implementation is only extending the existing API and therefore should not break any of the existing code.
2019-05-06 11:27:44 +03:00
self.make_replication_events()
2016-10-11 02:10:53 +03:00
def consume(msg):
Smart replication feedback This commit makes psycopg2 responsible for sending the status update (feedback) messages to the server regardless of whether a synchronous or asynchronous connection is used. Feedback is sent every *status_update* (default value is 10) seconds, which could be configured by passing a corresponding parameter to the `start_replication()` or `start_replication_expert()` methods. The actual feedback message is sent by the `pq_read_replication_message()` when the *status_update* timeout is reached. The default behavior of the `send_feedback()` method is changed. It doesn't send a feedback message on every call anymore but just updates internal structures. There is still a way to *force* sending a message if *force* or *reply* parameters are set. The new approach has certain advantages: 1. The client can simply call the `send_feedback()` for every processed message and the library will take care of not overwhelming the server. Actually, in the synchronous mode it is even mandatory to confirm every processed message. 2. The library tracks internally the pointer of the last received message which is not keepalive. If the client confirmed the last message and after that server sends only keepalives with increasing *wal_end*, the library can safely move forward *flush* position to the *wal_end* and later automatically report it to the server. Reporting of the *wal_end* received from keepalive messages is very important. Not doing so casing: 1. Excessive disk usage, because the replication slot prevents from WAL being cleaned up. 2. The smart and fast shutdown of the server could last indefinitely because walsender waits until the client report *flush* position equal to the *wal_end*. This implementation is only extending the existing API and therefore should not break any of the existing code.
2019-05-06 11:27:44 +03:00
raise StopReplication()
with self.assertRaises(psycopg2.DataError):
# try with invalid options
cur.start_replication(
slot_name=self.slot, options={'invalid_param': 'value'})
cur.consume_stream(consume)
# try with correct command
cur.start_replication(slot_name=self.slot)
Smart replication feedback This commit makes psycopg2 responsible for sending the status update (feedback) messages to the server regardless of whether a synchronous or asynchronous connection is used. Feedback is sent every *status_update* (default value is 10) seconds, which could be configured by passing a corresponding parameter to the `start_replication()` or `start_replication_expert()` methods. The actual feedback message is sent by the `pq_read_replication_message()` when the *status_update* timeout is reached. The default behavior of the `send_feedback()` method is changed. It doesn't send a feedback message on every call anymore but just updates internal structures. There is still a way to *force* sending a message if *force* or *reply* parameters are set. The new approach has certain advantages: 1. The client can simply call the `send_feedback()` for every processed message and the library will take care of not overwhelming the server. Actually, in the synchronous mode it is even mandatory to confirm every processed message. 2. The library tracks internally the pointer of the last received message which is not keepalive. If the client confirmed the last message and after that server sends only keepalives with increasing *wal_end*, the library can safely move forward *flush* position to the *wal_end* and later automatically report it to the server. Reporting of the *wal_end* received from keepalive messages is very important. Not doing so casing: 1. Excessive disk usage, because the replication slot prevents from WAL being cleaned up. 2. The smart and fast shutdown of the server could last indefinitely because walsender waits until the client report *flush* position equal to the *wal_end*. This implementation is only extending the existing API and therefore should not break any of the existing code.
2019-05-06 11:27:44 +03:00
self.assertRaises(StopReplication, cur.consume_stream, consume)
@skip_before_postgres(9, 4) # slots require 9.4
@skip_repl_if_green
def test_keepalive(self):
conn = self.repl_connect(connection_factory=LogicalReplicationConnection)
if conn is None:
return
cur = conn.cursor()
self.create_replication_slot(cur, output_plugin='test_decoding')
self.make_replication_events()
cur.start_replication(self.slot)
def consume(msg):
raise StopReplication()
self.assertRaises(StopReplication,
cur.consume_stream, consume, keepalive_interval=2)
conn.close()
2016-10-11 02:10:53 +03:00
@skip_before_postgres(9, 4) # slots require 9.4
2016-12-25 21:00:30 +03:00
@skip_repl_if_green
def test_stop_replication(self):
conn = self.repl_connect(connection_factory=LogicalReplicationConnection)
2016-10-11 02:10:53 +03:00
if conn is None:
return
cur = conn.cursor()
2015-10-19 18:02:18 +03:00
self.create_replication_slot(cur, output_plugin='test_decoding')
2015-10-19 18:02:18 +03:00
self.make_replication_events()
2015-10-19 18:02:18 +03:00
cur.start_replication(self.slot)
2016-10-11 02:10:53 +03:00
def consume(msg):
raise StopReplication()
self.assertRaises(StopReplication, cur.consume_stream, consume)
2015-10-15 19:01:43 +03:00
class AsyncReplicationTest(ReplicationTestCase):
2016-10-11 02:10:53 +03:00
@skip_before_postgres(9, 4) # slots require 9.4
2016-12-25 21:00:30 +03:00
@skip_repl_if_green
2015-10-15 19:01:43 +03:00
def test_async_replication(self):
2016-10-11 02:10:53 +03:00
conn = self.repl_connect(
connection_factory=LogicalReplicationConnection, async_=1)
2016-10-11 02:10:53 +03:00
if conn is None:
return
2016-12-25 19:46:11 +03:00
2015-10-15 19:01:43 +03:00
cur = conn.cursor()
2015-10-19 18:02:18 +03:00
self.create_replication_slot(cur, output_plugin='test_decoding')
2015-10-15 19:01:43 +03:00
self.wait(cur)
2015-10-19 18:02:18 +03:00
cur.start_replication(self.slot)
2015-10-15 19:01:43 +03:00
self.wait(cur)
2015-10-19 18:02:18 +03:00
self.make_replication_events()
self.msg_count = 0
2016-10-11 02:10:53 +03:00
2015-10-19 18:02:18 +03:00
def consume(msg):
# just check the methods
2020-11-18 00:52:11 +03:00
f"{cur.io_timestamp}: {repr(msg)}"
f"{cur.feedback_timestamp}: {repr(msg)}"
f"{cur.wal_end}: {repr(msg)}"
2015-10-19 18:02:18 +03:00
self.msg_count += 1
if self.msg_count > 3:
cur.send_feedback(reply=True)
2015-10-19 18:02:18 +03:00
raise StopReplication()
cur.send_feedback(flush_lsn=msg.data_start)
# cannot be used in asynchronous mode
self.assertRaises(psycopg2.ProgrammingError, cur.consume_stream, consume)
2015-10-19 18:02:18 +03:00
def process_stream():
while True:
msg = cur.read_message()
2015-10-19 18:02:18 +03:00
if msg:
consume(msg)
else:
select([cur], [], [])
self.assertRaises(StopReplication, process_stream)
2015-10-15 19:01:43 +03:00
2016-10-11 02:10:53 +03:00
2015-10-15 19:01:43 +03:00
def test_suite():
return unittest.TestLoader().loadTestsFromName(__name__)
2016-10-11 02:10:53 +03:00
2015-10-15 19:01:43 +03:00
if __name__ == "__main__":
unittest.main()