VACUUM: ignore indexing operations with CONCURRENTLY

alvherre · alvherre · commit d9d076222f5b · 2021-02-23T12:15:09.000-03:00
As envisioned in commit c98763b, it is possible for VACUUM to ignore certain transactions that are executing CREATE INDEX CONCURRENTLY and REINDEX CONCURRENTLY for the purposes of computing Xmin; that's because we know those transactions are not going to examine any other tables, and are not going to execute anything else in the same transaction. (Only operations on "safe" indexes can be ignored: those on indexes that are neither partial nor expressional). This is extremely useful in cases where CIC/RC can run for a very long time, because that used to be a significant headache for concurrent vacuuming of other tables. Reviewed-by: Matthias van de Meent <boekewurm+postgres@gmail.com> Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com> Discussion: https://postgr.es/m/20210115133858.GA18931@alvherre.pgsql
diff --git a/doc/src/sgml/ref/create_index.sgml b/doc/src/sgml/ref/create_index.sgml
@@ -855,6 +855,8 @@ Indexes:
    Like any long-running transaction, <command>CREATE INDEX</command> on a
    table can affect which tuples can be removed by concurrent
    <command>VACUUM</command> on any other table.
+   Excepted from this are operations with the <literal>CONCURRENTLY</literal>
+   option for indexes that are not partial and do not index any expressions.
   </para>
 
   <para>
diff --git a/doc/src/sgml/ref/reindex.sgml b/doc/src/sgml/ref/reindex.sgml
@@ -478,6 +478,8 @@ Indexes:
     Like any long-running transaction, <command>REINDEX</command> on a table
     can affect which tuples can be removed by concurrent
     <command>VACUUM</command> on any other table.
+    Excepted from this are operations with the <literal>CONCURRENTLY</literal>
+    option for indexes that are not partial and do not index any expressions.
    </para>
 
    <para>
diff --git a/src/backend/storage/ipc/procarray.c b/src/backend/storage/ipc/procarray.c
@@ -1610,7 +1610,13 @@ TransactionIdIsActive(TransactionId xid)
  * relations that's not required, since only backends in my own database could
  * ever see the tuples in them. Also, we can ignore concurrently running lazy
  * VACUUMs because (a) they must be working on other tables, and (b) they
- * don't need to do snapshot-based lookups.
+ * don't need to do snapshot-based lookups.  Similarly, for the non-catalog
+ * horizon, we can ignore CREATE INDEX CONCURRENTLY and REINDEX CONCURRENTLY
+ * when they are working on non-partial, non-expressional indexes, for the
+ * same reasons and because they can't run in transaction blocks.  (They are
+ * not possible to ignore for catalogs, because CIC and RC do some catalog
+ * operations.)  Do note that this means that CIC and RC must use a lock level
+ * that conflicts with VACUUM.
  *
  * This also computes a horizon used to truncate pg_subtrans. For that
  * backends in all databases have to be considered, and concurrently running
@@ -1660,9 +1666,6 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h)
 	bool		in_recovery = RecoveryInProgress();
 	TransactionId *other_xids = ProcGlobal->xids;
 
-	/* inferred after ProcArrayLock is released */
-	h->catalog_oldest_nonremovable = InvalidTransactionId;
-
 	LWLockAcquire(ProcArrayLock, LW_SHARED);
 
 	h->latest_completed = ShmemVariableCache->latestCompletedXid;
@@ -1682,6 +1685,7 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h)
 
 		h->oldest_considered_running = initial;
 		h->shared_oldest_nonremovable = initial;
+		h->catalog_oldest_nonremovable = initial;
 		h->data_oldest_nonremovable = initial;
 
 		/*
@@ -1752,7 +1756,7 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h)
 		if (statusFlags & (PROC_IN_VACUUM | PROC_IN_LOGICAL_DECODING))
 			continue;
 
-		/* shared tables need to take backends in all database into account */
+		/* shared tables need to take backends in all databases into account */
 		h->shared_oldest_nonremovable =
 			TransactionIdOlder(h->shared_oldest_nonremovable, xmin);
 
@@ -1773,11 +1777,26 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h)
 			MyDatabaseId == InvalidOid || proc->databaseId == MyDatabaseId ||
 			proc->databaseId == 0)	/* always include WalSender */
 		{
-			h->data_oldest_nonremovable =
-				TransactionIdOlder(h->data_oldest_nonremovable, xmin);
+			/*
+			 * We can ignore this backend if it's running CREATE INDEX
+			 * CONCURRENTLY or REINDEX CONCURRENTLY on a "safe" index -- but
+			 * only on vacuums of user-defined tables.
+			 */
+			if (!(statusFlags & PROC_IN_SAFE_IC))
+				h->data_oldest_nonremovable =
+					TransactionIdOlder(h->data_oldest_nonremovable, xmin);
+
+			/* Catalog tables need to consider all backends in this db */
+			h->catalog_oldest_nonremovable =
+				TransactionIdOlder(h->catalog_oldest_nonremovable, xmin);
+
 		}
 	}
 
+	/* catalog horizon should never be later than data */
+	Assert(TransactionIdPrecedesOrEquals(h->catalog_oldest_nonremovable,
+										 h->data_oldest_nonremovable));
+
 	/*
 	 * If in recovery fetch oldest xid in KnownAssignedXids, will be applied
 	 * after lock is released.
@@ -1799,6 +1818,8 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h)
 			TransactionIdOlder(h->shared_oldest_nonremovable, kaxmin);
 		h->data_oldest_nonremovable =
 			TransactionIdOlder(h->data_oldest_nonremovable, kaxmin);
+		h->catalog_oldest_nonremovable =
+			TransactionIdOlder(h->catalog_oldest_nonremovable, kaxmin);
 		/* temp relations cannot be accessed in recovery */
 	}
 	else
@@ -1825,6 +1846,9 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h)
 		h->data_oldest_nonremovable =
 			TransactionIdRetreatedBy(h->data_oldest_nonremovable,
 									 vacuum_defer_cleanup_age);
+		h->catalog_oldest_nonremovable =
+			TransactionIdRetreatedBy(h->catalog_oldest_nonremovable,
+									 vacuum_defer_cleanup_age);
 		/* defer doesn't apply to temp relations */
 	}
 
@@ -1847,7 +1871,9 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h)
 	h->shared_oldest_nonremovable =
 		TransactionIdOlder(h->shared_oldest_nonremovable,
 						   h->slot_catalog_xmin);
-	h->catalog_oldest_nonremovable = h->data_oldest_nonremovable;
+	h->catalog_oldest_nonremovable =
+		TransactionIdOlder(h->catalog_oldest_nonremovable,
+						   h->slot_xmin);
 	h->catalog_oldest_nonremovable =
 		TransactionIdOlder(h->catalog_oldest_nonremovable,
 						   h->slot_catalog_xmin);
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
@@ -2583,7 +2583,7 @@ static struct config_int ConfigureNamesInt[] =
 			NULL
 		},
 		&vacuum_defer_cleanup_age,
-		0, 0, 1000000,
+		0, 0, 1000000,		/* see ComputeXidHorizons */
 		NULL, NULL, NULL
 	},
 
