async def test_master_start(app):
plugins = setup_plugins(app,
dcs_get_database_identifier='1234',
dcs_lock=True,
pg_replication_role='master',
pg_get_database_identifier='1234')
def start_monitoring():
app.unhealthy('test_monitor', 'Waiting for first check')
plugins.start_monitoring.side_effect = start_monitoring
# sync startup
timeout = app.initialize()
assert plugins.mock_calls == [
call.initialize(),
call.get_my_id(),
# compare our id with the id in the DCS
call.dcs_get_database_identifier(),
call.pg_get_database_identifier(),
# check if I am a replica
call.pg_replication_role(),
# no, so check if there is a master
call.dcs_lock('master'),
# no master, so sure the DB is running
call.pg_start(),
# start monitoring
call.start_monitoring(),
call.dcs_watch(
app.master_lock_changed,
app._notify_state,
app._notify_conn_info,
),
call.get_conn_info(),
# set our first state
call.dcs_set_state({
'host': '127.0.0.1',
'replication_role': 'master',
'health_problems': {'test_monitor':
{'can_be_replica': False, 'reason': 'Waiting for first check'}}})
]
# Carry on running afterwards
assert timeout == None
assert app.health_problems == {'test_monitor': {'can_be_replica': False, 'reason': 'Waiting for first check'}}
# Our test monitor becomes healthy
plugins.reset_mock()
app.healthy('test_monitor')
assert plugins.mock_calls == [
call.dcs_set_state({
'host': '127.0.0.1',
'replication_role': 'master',
'health_problems': {}}),
call.pg_replication_role(),
call.dcs_lock('master'),
call.dcs_set_conn_info({'host': '127.0.0.1'}),
]
def test_replica_start(app):
plugins = setup_plugins(app,
dcs_get_database_identifier='1234',
dcs_lock=True,
pg_am_i_replica=True,
pg_get_database_identifier='1234')
app._conn_info['a'] = 'b'
def start_monitoring():
app.unhealthy('test_monitor', 'Waiting for first check')
plugins.start_monitoring.side_effect = start_monitoring
# sync startup
timeout = app.initialize()
assert plugins.mock_calls == [
call.initialize(),
call.get_my_id(),
# compare our id with the id in the DCS
call.dcs_get_database_identifier(),
call.pg_get_database_identifier(),
# check if I am a replica
call.pg_am_i_replica(),
# not master, so sure the DB is running
call.pg_start(),
# start monitoring
call.start_monitoring(),
call.dcs_watch(conn_info=None, state=None),
# setup our connection info
call.get_conn_info(),
# set our first state
call.dcs_set_state({
'a': 'b',
'host': '127.0.0.1',
'health_problems': {'test_monitor':
{'can_be_replica': False, 'reason': 'Waiting for first check'}},
})
]
# Carry on running afterwards
assert timeout == None
assert app.health_problems == {'test_monitor': {'can_be_replica': False, 'reason': 'Waiting for first check'}}
# Our test monitor becomes healthy
plugins.reset_mock()
app.healthy('test_monitor')
assert plugins.mock_calls == [
call.dcs_set_state({'health_problems': {},
'a': 'b',
'host': '127.0.0.1',
}),
call.pg_am_i_replica(),
call.dcs_set_conn_info({'a': 'b', 'host': '127.0.0.1'}),
]
async def test_replica_reaction_to_master_lock_change(app):
plugins = setup_plugins(app,
pg_get_timeline=42,
pg_replication_role='replica')
assert app.initialize() == None
plugins.reset_mock()
# if the lock changes owner to someone else, carry on trucking
plugins.reset_mock()
app.master_lock_changed('someone else')
assert app._plugins.mock_calls == [
call.pg_replication_role(),
]
assert app._master_lock_owner == 'someone else'
# if the lock is owned by us, er, we stop replication and become the master
plugins.reset_mock()
plugins.pg_replication_role.side_effect = ['replica', 'master']
app.master_lock_changed(app.my_id)
assert app._plugins.mock_calls == [
call.pg_replication_role(),
call.pg_stop_replication(),
call.pg_replication_role(),
call.pg_get_timeline(),
call.dcs_set_timeline(42),
call.dcs_set_state({
'health_problems': {},
'replication_role': 'master',
'host': '127.0.0.1'}),
]
assert app._master_lock_owner == app.my_id
def test_replica_slightly_sick(app):
plugins = setup_plugins(app,
pg_am_i_replica=True)
app.initialize()
plugins.reset_mock()
app.unhealthy('boom', 'It went Boom', can_be_replica=True)
assert plugins.mock_calls == [
call.dcs_set_state({'host': '127.0.0.1', 'health_problems': {'boom': {'reason': 'It went Boom', 'can_be_replica': True}}}),
call.pg_am_i_replica(),
]
async def test_master_unhealthy(app):
plugins = setup_plugins(app,
pg_am_i_replica=False)
app.initialize()
plugins.reset_mock()
app.unhealthy('boom', 'It went Boom', can_be_replica=True)
assert plugins.mock_calls == [
call.dcs_set_state({'host': '127.0.0.1', 'health_problems': {'boom': {'reason': 'It went Boom', 'can_be_replica': True}}}),
call.pg_am_i_replica(),
call.dcs_delete_conn_info(),
]
async def test_replica_reaction_to_master_lock_change(app):
plugins = setup_plugins(app,
pg_get_timeline=42,
pg_replication_role='replica')
assert app.initialize() == None
plugins.reset_mock()
# if the lock changes owner to someone else, carry on trucking
plugins.reset_mock()
app.master_lock_changed('someone else')
assert plugins.mock_calls == [
call.pg_replication_role(),
call.master_lock_changed('someone else')
]
assert app._master_lock_owner == 'someone else'
# if the lock is owned by us, er, we stop replication and become the master
plugins.reset_mock()
plugins.pg_replication_role.side_effect = ['replica', 'master']
app.master_lock_changed(app.my_id)
assert plugins.mock_calls == [
call.pg_replication_role(),
call.dcs_set_state({
'replication_role': 'taking-over',
'willing': None,
'health_problems': {},
'host': '127.0.0.1'
}),
call.pg_stop_replication(),
call.pg_replication_role(),
call.pg_get_timeline(),
call.dcs_set_timeline(42),
call.dcs_set_state({
'health_problems': {},
'replication_role': 'master',
'willing': None,
'host': '127.0.0.1'
}),
call.master_lock_changed('42')
]
assert app._master_lock_owner == app.my_id
def test_replica_slightly_sick(app):
plugins = setup_plugins(app,
pg_replication_role='replica')
app.initialize()
plugins.reset_mock()
app.unhealthy('boom', 'It went Boom', can_be_replica=True)
assert plugins.mock_calls == [
call.dcs_set_state({
'host': '127.0.0.1',
'replication_role': 'replica',
'willing': None, # I am not going to participate in master elections
'health_problems': {'boom': {'reason': 'It went Boom', 'can_be_replica': True}}}),
call.pg_replication_role(),
]
def test_replica_unhealthy(app):
plugins = setup_plugins(app,
pg_replication_role='replica')
app.initialize()
plugins.reset_mock()
app.unhealthy('boom', 'It went Boom')
assert plugins.mock_calls == [
call.dcs_set_state({
'host': '127.0.0.1',
'replication_role': 'replica',
'health_problems': {'boom': {'reason': 'It went Boom', 'can_be_replica': False}}}),
call.pg_replication_role(),
call.dcs_delete_conn_info(),
]
async def test_master_unhealthy(app):
plugins = setup_plugins(app, pg_replication_role='master')
app.initialize()
plugins.reset_mock()
app.unhealthy('boom', 'It went Boom', can_be_replica=True)
assert plugins.mock_calls == [
call.dcs_set_state({
'host': '127.0.0.1',
'replication_role': 'master',
'health_problems': {
'boom': {
'reason': 'It went Boom',
'can_be_replica': True
}
}
}),
call.pg_replication_role(),
call.dcs_delete_conn_info(),
]
plugins.reset_mock()
# now we should have _handle_unhealthy_master running
with patch('asyncio.sleep') as sleep, patch(
'zgres.deadman.App._stop') as exit, patch(
'time.sleep') as blocking_sleep:
sleeper = FakeSleeper()
sleep.side_effect = sleeper
exit.side_effect = lambda: sleeper.finish()
# there is no replica, so we just sleep and ping the
# DCS to find a willing replica
states = [iter([])]
plugins.dcs_list_state.side_effect = states
await sleeper.next()
assert plugins.mock_calls == [call.dcs_list_state()]
# we add a willing replica
states = [iter([('other', {'willing': 1})])]
plugins.dcs_list_state.side_effect = states
plugins.reset_mock()
await sleeper.next()
assert plugins.mock_calls == [
call.dcs_list_state(),
call.pg_replication_role(),
call.pg_stop(),
call.dcs_disconnect()
]
async def test_master_unhealthy(app):
plugins = setup_plugins(app,
pg_replication_role='master')
app.initialize()
plugins.reset_mock()
app.unhealthy('boom', 'It went Boom', can_be_replica=True)
assert plugins.mock_calls == [
call.dcs_set_state({
'host': '127.0.0.1',
'replication_role': 'master',
'health_problems': {'boom': {'reason': 'It went Boom', 'can_be_replica': True}}}),
call.pg_replication_role(),
call.dcs_delete_conn_info(),
]
plugins.reset_mock()
# now we should have _handle_unhealthy_master running
with patch('asyncio.sleep') as sleep, patch('sys.exit') as exit, patch('time.sleep') as blocking_sleep:
sleeper = FakeSleeper()
sleep.side_effect = sleeper
exit.side_effect = lambda x: sleeper.finish()
# there is no replica, so we just sleep and ping the
# DCS to find a willing replica
states = [iter([])]
plugins.dcs_get_all_state.side_effect = states
plugins.willing_replicas.side_effect = [iter([])]
await sleeper.next()
assert plugins.mock_calls == [
call.dcs_get_all_state(),
call.willing_replicas(states[0])]
# we add a willing replica
states = [iter([])]
plugins.dcs_get_all_state.side_effect = states
plugins.willing_replicas.side_effect = [iter([('other', {})])]
plugins.reset_mock()
await sleeper.next()
assert plugins.mock_calls == [
call.dcs_get_all_state(),
call.willing_replicas(states[0]),
call.pg_replication_role(),
call.pg_stop(),
call.dcs_disconnect()
]
def test_replica_slightly_sick(app):
plugins = setup_plugins(app, pg_replication_role='replica')
app.initialize()
plugins.reset_mock()
app.unhealthy('boom', 'It went Boom', can_be_replica=True)
assert plugins.mock_calls == [
call.dcs_set_state({
'host': '127.0.0.1',
'replication_role': 'replica',
'willing':
None, # I am not going to participate in master elections
'health_problems': {
'boom': {
'reason': 'It went Boom',
'can_be_replica': True
}
}
}),
call.pg_replication_role(),
]
def test_replica_start(app):
plugins = setup_plugins(app,
dcs_get_database_identifier='1234',
dcs_lock=True,
pg_replication_role='replica',
pg_get_database_identifier='1234')
app._conn_info['a'] = 'b'
def start_monitoring():
app.unhealthy('test_monitor', 'Waiting for first check')
plugins.start_monitoring.side_effect = start_monitoring
# sync startup
timeout = app.initialize()
assert plugins.mock_calls == [
call.initialize(),
call.get_my_id(),
# compare our id with the id in the DCS
call.dcs_get_database_identifier(),
call.pg_get_database_identifier(),
# check if I am a replica
call.pg_replication_role(),
# not master, so sure the DB is running
call.pg_start(),
# start monitoring
call.start_monitoring(),
call.dcs_watch(
app.master_lock_changed,
app._notify_state,
app._notify_conn_info,
),
# setup our connection info
call.get_conn_info(),
# set our first state
call.dcs_set_state({
'a': 'b',
'host': '127.0.0.1',
'replication_role': 'replica',
'health_problems': {
'test_monitor': {
'can_be_replica': False,
'reason': 'Waiting for first check'
}
},
})
]
# Carry on running afterwards
assert timeout == None
assert app.health_problems == {
'test_monitor': {
'can_be_replica': False,
'reason': 'Waiting for first check'
}
}
# Our test monitor becomes healthy
plugins.reset_mock()
with patch('time.time') as mock_time:
app.healthy('test_monitor')
assert plugins.mock_calls == [
call.veto_takeover({
'health_problems': {},
'a': 'b',
'replication_role': 'replica',
'host': '127.0.0.1'
}),
call.dcs_set_state({
'health_problems': {},
'a': 'b',
'replication_role': 'replica',
'host': '127.0.0.1',
'willing': mock_time(),
}),
call.pg_replication_role(),
call.dcs_set_conn_info({
'a': 'b',
'host': '127.0.0.1'
}),
]
