gitops/patroni
1
0
Fork 0
mirror of https://github.com/outbackdingo/patroni.git synced 2026-08-25 14:53:37 +00:00
Code Issues Packages Projects Releases Wiki Activity

Start slave with the correct recovery.conf on governor start

Browse Source 
Also governor is able to pick up already running master and slave
instances without restarting them.
In case is you have a lock and postgres is not running behaviour remains
the same: it will start master in read only mode and then promote if it
still has the lock.
This commit is contained in:
Alexander Kukushkin
2015-05-11 16:50:21 +02:00
parent eb1721f65f
commit e2faf641d5
3 changed files with 51 additions and 56 deletions
Download Patch File Download Diff File Expand all files Collapse all files
governor.py
+6 -16
View File
@@ -1,11 +1,10 @@
#!/usr/bin/env python
import sys
import yaml
import time
import urllib2
import atexit
import logging
import sys
import time
import yaml
from helpers.etcd import Etcd
from helpers.postgresql import Postgresql
@@ -30,14 +29,9 @@ def stop_postgresql():
atexit.register(stop_postgresql)
# wait for etcd to be available
etcd_ready = False
while not etcd_ready:
try:
etcd.touch_member(postgresql.name, postgresql.connection_string)
etcd_ready = True
except urllib2.URLError:
logging.info("waiting on etcd")
time.sleep(5)
while not etcd.touch_member(postgresql.name, postgresql.connection_string):
logging.info("waiting on etcd")
time.sleep(5)
# is data directory empty?
if postgresql.data_directory_empty():
@@ -60,11 +54,7 @@ if postgresql.data_directory_empty():
synced_from_leader = True
else:
time.sleep(5)
else:
postgresql.write_recovery_conf(None)
postgresql.start()
while True:
logging.info(ha.run_cycle())
time.sleep(config["loop_wait"])
helpers/etcd.py
+1 -1
View File
@@ -103,7 +103,7 @@ class Etcd:
raise CurrentLeaderError("Etcd is not responding properly")
def touch_member(self, member, connection_string):
self.put_client_path('/members/' + member, value=connection_string)
return self.put_client_path('/members/' + member, value=connection_string)
def take_leader(self, value):
return self.put_client_path('/leader', value=value, ttl=self.ttl)
helpers/ha.py
+44 -39
View File
@@ -44,55 +44,60 @@ class Ha:
def run_cycle(self):
try:
if self.state_handler.is_healthy():
self.load_cluster_from_etcd()
if self.is_unlocked():
if self.state_handler.is_healthiest_node(self.cluster.members):
if self.acquire_lock():
if not self.state_handler.is_leader():
self.state_handler.promote()
return "promoted self to leader by acquiring session lock"
return "acquired session lock as a leader"
else:
self.load_cluster_from_etcd()
if self.state_handler.is_leader():
self.demote()
return "demoted self due after trying and failing to obtain lock"
else:
self.follow_the_leader()
return "following new leader after trying and failing to obtain lock"
self.load_cluster_from_etcd()
if self.is_unlocked():
if not self.state_handler.is_healthy():
return 'no action. not healthy enough to do anything.'
elif self.state_handler.is_healthiest_node(self.cluster.members):
if self.acquire_lock():
if not self.state_handler.is_leader():
self.state_handler.promote()
return "promoted self to leader by acquiring session lock"
return "acquired session lock as a leader"
else:
self.load_cluster_from_etcd()
if self.state_handler.is_leader():
self.demote()
return "demoting self because i am not the healthiest node"
return "demoted self due after trying and failing to obtain lock"
else:
self.follow_the_leader()
return "following a different leader because i am not the healthiest node"
return "following new leader after trying and failing to obtain lock"
else:
if self.has_lock() and self.update_lock():
try:
if not self.state_handler.is_leader():
self.state_handler.promote()
return "promoted self to leader because i had the session lock"
else:
return "no action. i am the leader with the lock"
finally:
# create replication slots
self.state_handler.create_replication_slots([m.hostname for m in self.cluster.members])
self.load_cluster_from_etcd()
if self.state_handler.is_leader():
self.demote()
return "demoting self because i am not the healthiest node"
else:
logger.info("does not have lock")
if self.state_handler.is_leader():
self.demote()
return "demoting self because i do not have the lock and i was a leader"
else:
self.follow_the_leader()
return "no action. i am a secondary and i am following a leader"
self.follow_the_leader()
return "following a different leader because i am not the healthiest node"
else:
if not self.state_handler.is_running(): # XXX is_running == is_healthy
if self.has_lock() and not self.state_handler.is_healthy():
self.state_handler.write_recovery_conf(None)
self.state_handler.start()
return "postgresql was stopped. starting again."
return "no action. not healthy enough to do anything."
self.load_cluster_from_etcd()
if self.has_lock() and self.update_lock():
try:
if not self.state_handler.is_leader():
self.state_handler.promote()
return "promoted self to leader because i had the session lock"
else:
return "no action. i am the leader with the lock"
finally:
# create replication slots
self.state_handler.create_replication_slots([m.hostname for m in self.cluster.members])
else:
logger.info("does not have lock")
if not self.state_handler.is_healthy():
self.state_handler.write_recovery_conf(self.cluster.leader)
self.state_handler.start()
return 'starting as a secondary'
elif self.state_handler.is_leader():
self.demote()
return "demoting self because i do not have the lock and i was a leader"
else:
self.follow_the_leader()
return "no action. i am a secondary and i am following a leader"
except EtcdError:
logger.error("Error communicating with Etcd")
except OperationalError: