MDB-30823: use replication lag in seconds instead of bytes, allow lag…

… in scheduled switchover (#35)

in scheduler switchover, replication lag is allowed, because primary will be stopped before replica promotion

Co-authored-by: reshke <reshkekirill@gmail.com>

src/__init__.py

@@ -79,6 +79,7 @@ def read_config(filename=None, options=None):
             'verify_certs': 'no',
             'drop_slot_countdown': 10,
             'replication_slots_polling': None,
+            'max_allowed_switchover_lag_ms': 60000,
         },
         'primary': {
             'change_replication_type': 'yes',

src/main.py

@@ -1845,16 +1845,18 @@ def _candidate_is_sync_with_primary(self, db_state, switchover_candidate):
             return True
 
         replics_info = db_state.get('replics_info', list())
+        max_allowed_lag_ms = self.config.getint('global', 'max_allowed_switchover_lag_ms')
         for replica in replics_info:
             if replica.get('sync_state', '') != 'quorum':
                 continue
             if replica.get('application_name', '') != helpers.app_name_from_fqdn(switchover_candidate):
                 continue
-            replay_lag = replica.get('replay_location_diff', -1)
-            if replay_lag != 0:
+            replay_lag = replica.get('replay_lag_msec', -1)
+            logging.info(f"Replica {switchover_candidate} has replay lag {replay_lag}ms")
+            if replay_lag > max_allowed_lag_ms:
                 if not self.config.getboolean('replica', 'allow_potential_data_loss'):
                     logging.warning(
-                        f"Replica {switchover_candidate} has replay lag {replay_lag} so cannot be primary for switchover"
+                        f"Replica {switchover_candidate} has replay lag {replay_lag}ms so cannot be primary for switchover, max allowed lag {max_allowed_lag_ms}ms"
                     )
                     return None
                 else: