Fix some comments related to logical repslot advancing
authorMichael Paquier <[email protected]>
Thu, 26 Dec 2019 13:26:09 +0000 (22:26 +0900)
committerMichael Paquier <[email protected]>
Thu, 26 Dec 2019 13:26:09 +0000 (22:26 +0900)
confirmed_flush is part of a replication slot's information, but not
confirmed_lsn.

Author: Kyotaro Horiguchi
Discussion: https://postgr.es/m/20191226.175919.17237335658671970[email protected]
Backpatch-through: 11

src/backend/replication/slotfuncs.c

index 6683fc3f9b749bc226866cf26c280847414a99f0..ba08ad405f463e116e794518782cd0bb6afd6bc7 100644 (file)
@@ -378,8 +378,8 @@ pg_physical_replication_slot_advance(XLogRecPtr moveto)
 /*
  * Helper function for advancing our logical replication slot forward.
  *
- * The slot's restart_lsn is used as start point for reading records,
- * while confirmed_lsn is used as base point for the decoding context.
+ * The slot's restart_lsn is used as start point for reading records, while
+ * confirmed_flush is used as base point for the decoding context.
  *
  * We cannot just do LogicalConfirmReceivedLocation to update confirmed_flush,
  * because we need to digest WAL to advance restart_lsn allowing to recycle
@@ -549,8 +549,8 @@ pg_replication_slot_advance(PG_FUNCTION_ARGS)
        /*
         * Check if the slot is not moving backwards.  Physical slots rely simply
         * on restart_lsn as a minimum point, while logical slots have confirmed
-        * consumption up to confirmed_lsn, meaning that in both cases data older
-        * than that is not available anymore.
+        * consumption up to confirmed_flush, meaning that in both cases data
+        * older than that is not available anymore.
         */
        if (OidIsValid(MyReplicationSlot->data.database))
                minlsn = MyReplicationSlot->data.confirmed_flush;