Sonja K
10/07/2026, 8:10 AMkapa.ai
10/07/2026, 8:10 AMlost state, the Postgres connector cannot automatically recover from it within the sync [Github Issue (closed): [source-postgres] Invalid CDC Position Behavior not working; feat(source-postgres): remove non-functional invalid_cdc_cursor_position_behavior option].
To fix this issue and prevent it from recurring, follow these steps:
Step 1: Drop and Recreate the Replication Slot An invalidated replication slot cannot be resumed. You must manually drop and recreate the slot on your database [feat(source-postgres): remove non-functional invalid_cdc_cursor_position_behavior option]:
1. In PostgreSQL, drop the broken slot:SELECT pg_drop_replication_slot('airbyte_slot');
2. Recreate the logical replication slot:SELECT pg_create_logical_replication_slot('airbyte_slot', 'pgoutput');
Step 2: Reset the Airbyte Connection Because the replication slot has been recreated and previous LSN offsets are lost:
1. In the Airbyte UI, go to your connection.
2. Trigger a Reset on the connection/streams to clear the existing state [Github Issue (closed): [source-postgres] Invalid CDC Position Behavior not working].
3. Run a sync to trigger a fresh initial snapshot/sync.
Step 3: Prevent Conflicts with Recovery If Airbyte is reading from a Hot Standby (read replica), queries and replication slots conflict with changes being replayed from the primary. To prevent the slot from being invalidated again:
• Enable hot_standby_feedback (Recommended):
Set hot_standby_feedback = true on the standby server [Documentation Integrations > Sources > Postgres > Troubleshooting Postgres Sources>]. This informs the primary server about what the replica needs and prevents row versions/WAL needed by queries from being prematurely cleaned up.
• Sync from the primary:
If possible, point the Airbyte source to the primary PostgreSQL server rather than the hot standby [Documentation Integrations > Sources > Postgres > Troubleshooting Postgres Sources>].
• Tune standby delay parameters:
You can increase max_standby_archive_delay and max_standby_streaming_delay to give queries on the replica more time before being canceled [Documentation Integrations > Sources > Postgres > Troubleshooting Postgres Sources>].
• Tune WAL retention & sync frequency:
Ensure syncs run frequently enough, and configure sufficient WAL retention (such as tuning wal_keep_size on PG 13+) so the slot does not fall behind and get dropped [Github Issue (closed): [source-postgres] Invalid CDC Position Behavior not working; Documentation Integrations > Sources > Postgres > Troubleshooting Postgres Sources>]. 👍 0 button 👎 0 button ℹ️ Tip: Mention @kapa.ai in the thread for follow-ups or DM the app for 1-to-1 help.