gitops/patroni
1
0
Fork 0
mirror of https://github.com/outbackdingo/patroni.git synced 2026-08-26 15:40:21 +00:00
Code Issues Packages Projects Releases Wiki Activity

Bugfix. It wasn't possible to start the old cluster if all its members were shutdown.

Browse Source 
Basically this is rollback to the original decision tree with the small exception:
1 - If leader is defined and it's not me - then slave would be started immidiately
with the correct recovery conf.
2 - If leader is defined in and it's my host, then it will start instance in
readonly (but without primary_conninfo in recovery.conf)
3 - And the third case - if the leader is not defined - it also will start
instance in readonly, without primary_conninfo.
After performing 2 or 3 it will perform usual decision tree.
This commit is contained in:
Alexander Kukushkin
2015-05-12 10:19:45 +02:00
parent be7677a1b8
commit 9dcbc75fd2
1 changed files with 25 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files
helpers/ha.py
+25 -27
View File
@@ -45,65 +45,63 @@ class Ha:
def run_cycle(self):
try:
self.load_cluster_from_etcd()
if not self.state_handler.is_healthy():
has_lock = self.has_lock()
self.state_handler.write_recovery_conf(None if has_lock else self.cluster.leader)
self.state_handler.start()
if not has_lock:
return 'started as a secondary'
logging.info('started as readonly because i had the session 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.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"
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"
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"
return 'following new leader after trying and failing to obtain lock'
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 'demoting self because i am not the healthiest node'
else:
self.follow_the_leader()
return "following a different leader because i am not the healthiest node"
return 'following a different leader because i am not the healthiest node'
else:
if self.has_lock() and not self.state_handler.is_healthy():
self.state_handler.write_recovery_conf(None)
self.state_handler.start()
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"
return 'promoted self to leader because i had the session lock'
else:
return "no action. i am the leader with the lock"
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():
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"
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"
return 'no action. i am a secondary and i am following a leader'
except EtcdError:
logger.error("Error communicating with Etcd")
logger.error('Error communicating with Etcd')
except OperationalError:
logger.error("Error communicating with Postgresql. Will try again.")
logger.error('Error communicating with Postgresql. Will try again')
except HealthiestMemberError:
logger.error("failed to determine healthiest member fromt etcd")
logger.error('failed to determine healthiest member fromt etcd')
def run(self):
while True: