From afe1a2c92cf5a75607fe12d87ad80e56029fc139 Mon Sep 17 00:00:00 2001
From: Alexander Kukushkin <alexander.kukushkin@zalando.de>
Date: Fri, 18 Sep 2015 12:41:36 +0200
Subject: [PATCH 1/4] During shutdown sigterm is send to all processes every
 second

First time is triggers some shutdown mechanisms in a patroni, but the
second one just terminates patroni. Solution is pretty simple, we need
to ignore all other sigterms which are coming after the first one.
---
 patroni/__init__.py   | 1 +
 patroni/utils.py      | 1 +
 tests/test_patroni.py | 3 +++
 3 files changed, 5 insertions(+)

diff --git a/patroni/__init__.py b/patroni/__init__.py
index c190e2de..5a8588ce 100644
--- a/patroni/__init__.py
+++ b/patroni/__init__.py
@@ -126,6 +126,7 @@ def main():
     except KeyboardInterrupt:
         pass
     finally:
+        patroni.api.shutdown()
         patroni.touch_member(patroni.shutdown_member_ttl)  # schedule member removal
         patroni.postgresql.stop()
         patroni.ha.dcs.delete_leader()
diff --git a/patroni/utils.py b/patroni/utils.py
index 45ada864..898b058e 100644
--- a/patroni/utils.py
+++ b/patroni/utils.py
@@ -46,6 +46,7 @@ def calculate_ttl(expiration):
 
 
 def sigterm_handler(signo, stack_frame):
+    signal.signal(signal.SIGTERM, signal.SIG_IGN)
     sys.exit()
 
 
diff --git a/tests/test_patroni.py b/tests/test_patroni.py
index 65293d03..7481a6fd 100644
--- a/tests/test_patroni.py
+++ b/tests/test_patroni.py
@@ -40,6 +40,9 @@ def keyboard_interrupt(*args):
 
 class Mock_BaseServer__is_shut_down:
 
+    def wait(self):
+        pass
+
     def set(self):
         pass
 

From 3145e94797af2e175fc7ccf8c017110c3b87cddb Mon Sep 17 00:00:00 2001
From: Alexander Kukushkin <alexander.kukushkin@zalando.de>
Date: Fri, 18 Sep 2015 15:27:09 +0200
Subject: [PATCH 2/4] check ignore_sigterm flag instead of setting new handler
 for SIGTERM

---
 patroni/utils.py | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/patroni/utils.py b/patroni/utils.py
index 898b058e..19d5837a 100644
--- a/patroni/utils.py
+++ b/patroni/utils.py
@@ -8,6 +8,7 @@ import time
 
 from patroni.exceptions import DCSError
 
+ignore_sigterm = False
 interrupted_sleep = False
 reap_children = False
 
@@ -46,8 +47,10 @@ def calculate_ttl(expiration):
 
 
 def sigterm_handler(signo, stack_frame):
-    signal.signal(signal.SIGTERM, signal.SIG_IGN)
-    sys.exit()
+    global ignore_sigterm
+    if not ignore_sigterm:
+        ignore_sigterm = True
+        sys.exit()
 
 
 def sigchld_handler(signo, stack_frame):

From 9dfb562a4f560c5c9d0349a9d25dad24e547688b Mon Sep 17 00:00:00 2001
From: Alexander Kukushkin <alexander.kukushkin@zalando.de>
Date: Fri, 18 Sep 2015 15:29:50 +0200
Subject: [PATCH 3/4] wait for event on leader key a little bit longer than
 timeout

---
 patroni/etcd.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/patroni/etcd.py b/patroni/etcd.py
index c0950c85..fc6d5a97 100644
--- a/patroni/etcd.py
+++ b/patroni/etcd.py
@@ -251,7 +251,7 @@ class Etcd(AbstractDCS):
 
             while index and timeout >= 1:  # when timeout is too small urllib3 doesn't have enough time to connect
                 try:
-                    self.client.watch(self.leader_path, index=index + 1, timeout=timeout)
+                    self.client.watch(self.leader_path, index=index + 1, timeout=timeout + 0.5)
                     # Synchronous work of all cluster members with etcd is less expensive
                     # than reestablishing http connection every time from every replica.
                     return True

From 0a3a6d72cf98e8a81e212d45cba3a7b9f2b3ccc9 Mon Sep 17 00:00:00 2001
From: Alexander Kukushkin <alexander.kukushkin@zalando.de>
Date: Fri, 18 Sep 2015 15:33:58 +0200
Subject: [PATCH 4/4] Fix etcd unit tests

---
 tests/test_etcd.py | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/tests/test_etcd.py b/tests/test_etcd.py
index 10a602db..9ded883b 100644
--- a/tests/test_etcd.py
+++ b/tests/test_etcd.py
@@ -67,11 +67,11 @@ def requests_get(url, **kwargs):
 
 
 def etcd_watch(key, index=None, timeout=None, recursive=None):
-    if timeout == 1:
+    if timeout == 2.0:
         raise urllib3.exceptions.TimeoutError
-    elif timeout == 5:
+    elif timeout == 5.0:
         return etcd.EtcdResult('delete', {})
-    elif timeout == 10:
+    elif timeout == 10.0:
         raise etcd.EtcdException
     elif index == 20729:
         return etcd.EtcdResult('set', {'value': 'postgresql1', 'modifiedIndex': index + 1})
@@ -281,7 +281,7 @@ class TestEtcd(unittest.TestCase):
         self.etcd.client.watch = etcd_watch
         self.etcd.watch(100)
         self.etcd.get_cluster()
-        self.etcd.watch(1)
-        self.etcd.watch(5)
-        self.etcd.watch(10)
+        self.etcd.watch(1.5)
+        self.etcd.watch(4.5)
+        self.etcd.watch(9.5)
         self.etcd.watch(100)