No need to checkpoint and restart if pg is not running

When updating replication settings, no need to attempt a checkpoint and
restart when postgres keeper is not running. Just start the service.

Fixes hapostgres#837

src/bin/pg_autoctl/keeper.c

@@ -973,17 +973,29 @@ keeper_ensure_configuration(Keeper *keeper, bool postgresNotRunningIsOk)
 			log_info("Replication settings at \"%s\" have changed, "
 					 "restarting Postgres", upstreamConfPath);
 
-			if (!pgsql_checkpoint(&(postgres->sqlClient)))
+			if (pg_setup_is_running(pgSetup))
 			{
-				log_warn("Failed to CHECKPOINT before restart, "
-						 "see above for details");
-			}
+				if (!pgsql_checkpoint(&(postgres->sqlClient)))
+				{
+					log_warn("Failed to CHECKPOINT before restart, "
+							 "see above for details");
+				}
 
-			if (!keeper_restart_postgres(keeper))
+				if (!keeper_restart_postgres(keeper))
+				{
+					log_error("Failed to restart Postgres to enable new "
+							  "replication settings, see above for details");
+					return false;
+				}
+			}
+			else
 			{
-				log_error("Failed to restart Postgres to enable new "
-						  "replication settings, see above for details");
-				return false;
+				if (!ensure_postgres_service_is_running(postgres))
+				{
+					log_error("Failed to start Postgres with new "
+							  "replication settings, see above for details");
+					return false;
+				}
 			}
 		}
 	}

tests/pgautofailover_utils.py

@@ -655,16 +655,20 @@ def show_uri(self, json=False):
             out, err, ret = command.execute("show uri", "show", "uri")
         return out
 
-    def logs(self):
+    def logs(self, log_type=""):
         log_string = ""
         if self.running():
             out, err, ret = self.stop_pg_autoctl()
-            log_string += f"STDOUT OF PG_AUTOCTL FOR {self.datadir}:\n"
-            log_string += f"{self.pg_autoctl.cmd}\n{out}\n"
-            log_string += f"STDERR OF PG_AUTOCTL FOR {self.datadir}:\n{err}\n"
-
-        pglogs = self.get_postgres_logs()
-        log_string += f"POSTGRES LOGS FOR {self.datadir}:\n{pglogs}\n"
+            if not log_type or (log_type == "STDOUT"):
+                log_string += f"STDOUT OF PG_AUTOCTL FOR {self.datadir}:\n"
+                log_string += f"{self.pg_autoctl.cmd}\n{out}\n"
+            if not log_type or (log_type == "STDERR"):
+                log_string += (
+                    f"STDERR OF PG_AUTOCTL FOR {self.datadir}:\n{err}\n"
+                )
+        if not log_type or (log_type == "POSTGRES"):
+            pglogs = self.get_postgres_logs()
+            log_string += f"POSTGRES LOGS FOR {self.datadir}:\n{pglogs}\n"
         return log_string
 
     def get_events_str(self):

tests/test_auth.py

@@ -68,6 +68,14 @@ def test_003_init_secondary():
         node1.get_synchronous_standby_names_local(),
         "ANY 1 (pgautofailover_standby_2)",
     )
+    logs = node2.logs("STDERR")
+    match = re.search(
+        "Failed to connect to .*, retrying until the server is ready", logs
+    )
+    if match:
+        print("Found connection failure in logs: ", match[0])
+    assert not match
+    node2.run()
 
 
 def test_004_failover():