Handle the case when initialize flag is not set and leader is present.

patroni/__init__.py

@@ -6,6 +6,7 @@ import yaml
 
 from patroni.api import RestApiServer
 from patroni.etcd import Etcd
+from patroni.exceptions import DCSError
 from patroni.ha import Ha
 from patroni.postgresql import Postgresql
 from patroni.utils import setup_signal_handlers, sleep, reap_children
@@ -59,21 +60,26 @@ class Patroni:
         # is data directory empty?
         if self.postgresql.data_directory_empty():
             while True:
-                # racing to initialize
-                if self.ha.dcs.initialize():
-                    try:
-                        self.postgresql.bootstrap()
-                    except:
-                        # bail out and clean the initialize flag.
-                        self.cleanup_on_failed_initialization()
-                        raise
-                    self.ha.dcs.take_leader()
-                    break
-                else:
-                    leader = self.ha.dcs.current_leader()
-                    if leader and self.postgresql.bootstrap(leader):
+                try:
+                    cluster = self.ha.dcs.get_cluster()
+                    if not cluster.is_unlocked():  # the leader already exists
+                        if not cluster.initialize:
+                            self.ha.dcs.initialize()
+                        self.postgresql.bootstrap(cluster.leader)
                         break
-                    sleep(5)
+                    # racing to initialize
+                    elif not cluster.initialize and self.ha.dcs.initialize():
+                        try:
+                            self.postgresql.bootstrap()
+                        except:
+                            # bail out and clean the initialize flag.
+                            self.cleanup_on_failed_initialization()
+                            raise
+                        self.ha.dcs.take_leader()
+                        break
+                except DCSError:
+                    logger.info('waiting on DCS')
+                sleep(5)
         elif self.postgresql.is_running():
             self.postgresql.load_replication_slots()
 

tests/test_patroni.py

@@ -9,7 +9,7 @@ import yaml
 
 from mock import Mock, patch
 from patroni.api import RestApiServer
-from patroni.dcs import Cluster, Member
+from patroni.dcs import Cluster, Member, Leader
 from patroni.etcd import Etcd
 from patroni.exceptions import PostgresException
 from patroni import Patroni, main
@@ -42,6 +42,30 @@ class Mock_BaseServer__is_shut_down:
         pass
 
 
+def get_cluster(initialize, leader):
+    return Cluster(initialize, leader, None, None)
+
+
+def get_cluster_not_initialized_without_leader():
+    return get_cluster(None, None)
+
+
+def get_cluster_initialized_without_leader():
+    return get_cluster(True, None)
+
+
+def get_cluster_not_initialized_with_leader():
+    return get_cluster(False, Leader(0, 0, 0,
+                       Member(0, 'leader', 'postgres://replicator:[email protected]:5435/postgres',
+                              None, None, 28)))
+
+
+def get_cluster_initialized_with_leader():
+    return get_cluster(True, Leader(0, 0, 0, 
+                       Member(0, 'leader', 'postgres://replicator:[email protected]:5435/postgres',
+                              None, None, 28)))
+
+
 class TestPatroni(unittest.TestCase):
 
     def __init__(self, method_name='runTest'):
@@ -129,24 +153,31 @@ class TestPatroni(unittest.TestCase):
 
     def test_patroni_initialize(self):
         self.p.ha.dcs.client.write = etcd_write
+        self.p.ha.dcs.client.read = etcd_read
         self.p.touch_member = self.touch_member
         self.p.postgresql.data_directory_empty = true
         self.p.ha.dcs.initialize = true
         self.p.postgresql.initialize = true
         self.p.postgresql.start = true
+        self.p.ha.dcs.get_cluster = get_cluster_not_initialized_without_leader
         self.p.initialize()
 
         self.p.ha.dcs.initialize = false
+        self.p.ha.dcs.get_cluster = get_cluster_initialized_with_leader
         time.sleep = time_sleep
         self.p.ha.dcs.client.read = etcd_read
         self.p.initialize()
 
-        self.p.ha.dcs.current_leader = nop
+        self.p.ha.dcs.get_cluster = get_cluster_initialized_without_leader
         self.assertRaises(SleepException, self.p.initialize)
 
         self.p.postgresql.data_directory_empty = false
         self.p.initialize()
 
+        self.p.ha.dcs.get_cluster = get_cluster_not_initialized_with_leader
+        self.p.postgresql.data_directory_empty = true
+        self.p.initialize()
+
     def test_schedule_next_run(self):
         self.p.next_run = time.time() - self.p.nap_time - 1
         self.p.schedule_next_run()
@@ -156,6 +187,8 @@ class TestPatroni(unittest.TestCase):
 
     def test_cleanup_on_initialization(self):
         self.p.ha.dcs.client.write = etcd_write
+        self.p.ha.dcs.client.read = etcd_read
+        self.p.ha.dcs.get_cluster = get_cluster_not_initialized_without_leader
         self.p.touch_member = self.touch_member
         self.p.postgresql.data_directory_empty = true
         self.p.ha.dcs.initialize = true

tests/test_postgresql.py

@@ -6,6 +6,7 @@ import unittest
 
 from patroni.dcs import Cluster, Leader, Member
 from patroni.postgresql import Postgresql
+from test_ha import true, false
 
 
 def nop(*args, **kwargs):
@@ -217,3 +218,9 @@ class TestPostgresql(unittest.TestCase):
         self.p.start()
         self.p.query = self.mock_query
         self.assertTrue(self.p.stop())
+
+    def test_move_data_directory(self):
+        self.p.is_running = is_running
+        os.rename = nop
+        os.path.isdir = true
+        self.p.move_data_directory()