def __init__(self, config):
self.nap_time = config['loop_wait']
self.postgresql = Postgresql(config['postgresql'])
self.ha = Ha(self.postgresql, self.get_dcs(self.postgresql.name, config))
host, port = config['restapi']['listen'].split(':')
self.api = RestApiServer(self, config['restapi'])
self.skydns2 = config.get('skydns2')
self.next_run = time.time()
self.shutdown_member_ttl = 300
def set_up(self):
self.p = MockPostgresql()
with patch.object(Client, 'machines') as mock_machines:
mock_machines.__get__ = Mock(
return_value=['http://remotehost:2379'])
self.e = Etcd('foo', {
'ttl': 30,
'host': 'remotehost:2379',
'scope': 'test'
})
self.e.client.read = etcd_read
self.e.client.write = etcd_write
self.ha = Ha(self.p, self.e)
self.ha.load_cluster_from_dcs()
self.ha.cluster = get_unlocked_cluster()
self.ha.load_cluster_from_dcs = nop
def run(config):
etcd = Etcd(config["etcd"])
postgresql = Postgresql(config["postgresql"])
ha = Ha(postgresql, etcd)
atexit.register(stop_postgresql, postgresql)
logging.info("Governor Starting up")
# is data directory empty?
if postgresql.data_directory_empty():
logging.info("Governor Starting up: Empty Data Dir")
# racing to initialize
wait_for_etcd("cannot initialize member without ETCD", etcd,
postgresql)
if etcd.race("/initialize", postgresql.name) or not etcd.members():
logging.info(
"Governor Starting up: Initialisation Race ... WON!!!")
logging.info("Governor Starting up: Initialise Postgres")
postgresql.initialize()
logging.info("Governor Starting up: Initialise Complete")
etcd.take_leader(postgresql.name)
logging.info("Governor Starting up: Starting Postgres")
postgresql.start()
else:
logging.info("Governor Starting up: Initialisation Race ... LOST")
logging.info("Governor Starting up: Sync Postgres from Leader")
synced_from_leader = False
while not synced_from_leader:
leader = etcd.current_leader()
if not leader:
time.sleep(5)
continue
if postgresql.sync_from_leader(leader):
logging.info("Governor Starting up: Sync Completed")
postgresql.write_recovery_conf(leader)
logging.info("Governor Starting up: Starting Postgres")
postgresql.start()
synced_from_leader = True
else:
time.sleep(5)
else:
logging.info("Governor Starting up: Existing Data Dir")
postgresql.follow_no_leader()
logging.info("Governor Starting up: Starting Postgres")
postgresql.start()
wait_for_etcd(
"running in readonly mode; cannot participate in cluster HA without etcd",
etcd, postgresql)
logging.info("Governor Running: Starting Running Loop")
while True:
try:
ha.run_cycle()
# create replication slots
if postgresql.is_leader():
logging.info("Governor Running: I am the Leader")
for node in etcd.get_client_path(
"/members?recursive=true").get("node",
{}).get("nodes", []):
member = node["key"].split('/')[-1]
if member != postgresql.name:
postgresql.query(
"DO LANGUAGE plpgsql $$DECLARE somevar VARCHAR; BEGIN SELECT slot_name INTO somevar FROM pg_replication_slots WHERE slot_name = '%(slot)s' LIMIT 1; IF NOT FOUND THEN PERFORM pg_create_physical_replication_slot('%(slot)s'); END IF; END$$;"
% {"slot": member})
etcd.touch_member(postgresql.name, postgresql.connection_string)
time.sleep(config["loop_wait"])
except urllib2.URLError:
logging.info(
"Lost connection to etcd, setting no leader and waiting on etcd"
)
postgresql.follow_no_leader()
wait_for_etcd(
"running in readonly mode; cannot participate in cluster HA without etcd",
etcd, postgresql)
if os.getenv('GOVERNOR_POSTGRESQL_READ_ONLY_PORT'):
config['postgresql']['read_only_port'] = os.getenv(
'GOVERNOR_POSTGRESQL_READ_ONLY_PORT')
if os.getenv('GOVERNOR_POSTGRESQL_DATA_DIR'):
config['postgresql']['data_dir'] = os.getenv(
'GOVERNOR_POSTGRESQL_DATA_DIR')
if os.getenv('GOVERNOR_POSTGRESQL_REPLICATION_NETWORK'):
config['postgresql']['replication']['network'] = os.getenv(
'GOVERNOR_POSTGRESQL_REPLICATION_NETWORK')
etcd = Etcd(config["etcd"])
postgresql = Postgresql(config["postgresql"])
ha = Ha(postgresql, etcd)
# leave things clean when shutting down, if possible
def shutdown(signal, frame):
logging.info("Governor Shutting Down: Received Shutdown Signal")
try:
if ha.has_lock():
logging.info("Governor Shutting Down: Abdicating Leadership")
etcd.abdicate(postgresql.name)
logging.info("Governor Shutting Down: Removing Membership")
etcd.delete_member(postgresql.name)
except:
logging.exception("Error during Abdication")
pass
def run(config):
etcd = Etcd(config["etcd"])
postgresql = Postgresql(config["postgresql"])
ha = Ha(postgresql, etcd)
atexit.register(stop_postgresql, postgresql)
signal.signal(signal.SIGTERM, signalhandler)
logging.info("Governor Starting up")
# is data directory empty?
if postgresql.data_directory_empty():
logging.info("Governor Starting up: Empty Data Dir")
# racing to initialize
wait_for_etcd("cannot initialize member without ETCD", etcd, postgresql)
if etcd.race("/initialize", postgresql.name):
logging.info("Governor Starting up: Initialisation Race ... WON!!!")
logging.info("Governor Starting up: Initialise Postgres")
postgresql.initialize()
logging.info("Governor Starting up: Initialise Complete")
etcd.take_leader(postgresql.name)
logging.info("Governor Starting up: Starting Postgres")
postgresql.start()
else:
logging.info("Governor Starting up: Initialisation Race ... LOST")
logging.info("Governor Starting up: Sync Postgres from Leader")
synced_from_leader = False
while not synced_from_leader:
leader = etcd.current_leader()
if not leader:
time.sleep(5)
continue
if postgresql.sync_from_leader(leader):
logging.info("Governor Starting up: Sync Completed")
postgresql.write_recovery_conf(leader)
logging.info("Governor Starting up: Starting Postgres")
postgresql.start()
synced_from_leader = True
else:
time.sleep(5)
else:
logging.info("Governor Starting up: Existing Data Dir")
postgresql.follow_no_leader()
logging.info("Governor Starting up: Starting Postgres")
postgresql.start()
wait_for_etcd("running in readonly mode; cannot participate in cluster HA without etcd", etcd, postgresql)
logging.info("Governor Running: Starting Running Loop")
while True:
try:
logging.info("Governor Running: %s" % ha.run_cycle())
# create replication slots
if postgresql.is_leader():
logging.info("Governor Running: I am the Leader")
for node in etcd.members():
member = node["hostname"]
if member != postgresql.name:
if postgresql.is_leader():
postgresql.ensure_replication_slot(
postgresql.replication_slot_name(member)
)
else:
postgresql.drop_replication_slot(
postgresql.replication_slot_name(member)
)
etcd.touch_member(postgresql.name, postgresql.connection_string)
time.sleep(config["loop_wait"])
except (urllib2.URLError, socket.timeout):
logging.info("Lost connection to etcd, setting no leader and waiting on etcd")
postgresql.follow_no_leader()
wait_for_etcd("running in readonly mode; cannot participate in cluster HA without etcd", etcd, postgresql)
def test_load_cluster_from_dcs(self):
ha = Ha(self.p, self.e)
ha.load_cluster_from_dcs()
self.e.get_cluster = get_unlocked_cluster
ha.load_cluster_from_dcs()
