Decide to failover based on uptodate info

When we notice the primary node has gone, we need to get up to date
information from the other standby nodes if needed.

pglookout/pglookout.py

@@ -494,9 +494,11 @@ def consider_failover(self, own_state, master_node, standby_nodes):
                 self.log.warning("Not considering failover, because it's not enabled by configuration")
             elif self.current_master:
                 self.cluster_monitor_check_queue.put("Master is missing, ask for immediate state check")
+                self.failover_decision_queue.get(timeout=self.missing_master_from_config_timeout)
                 master_known_to_be_gone = self.current_master in self.known_gone_nodes
                 now = time.monotonic()
                 config_timeout_exceeded = (now - self.cluster_nodes_change_time) >= self.missing_master_from_config_timeout
+
                 if master_known_to_be_gone or config_timeout_exceeded:
                     # we've seen a master at some point in time, but now it's
                     # not reachable or removed from configuration, perform an

test/conftest.py

@@ -30,6 +30,7 @@ def pgl():
     pgl_.cluster_monitor._connect_to_db = Mock()  # pylint: disable=protected-access
     pgl_.create_alert_file = Mock()
     pgl_.execute_external_command = Mock()
+    pgl_.failover_decision_queue = Mock()
     try:
         yield pgl_
     finally: