Skip WAL for new relfilenodes, under wal_level=minimal.

Until now, only selected bulk operations (e.g. COPY) did this.  If a
given relfilenode received both a WAL-skipping COPY and a WAL-logged
operation (e.g. INSERT), recovery could lose tuples from the COPY.  See
src/backend/access/transam/README section "Skipping WAL for New
RelFileNode" for the new coding rules.  Maintainers of table access
methods should examine that section.

To maintain data durability, just before commit, we choose between an
fsync of the relfilenode and copying its contents to WAL.  A new GUC,
wal_skip_threshold, guides that choice.  If this change slows a workload
that creates small, permanent relfilenodes under wal_level=minimal, try
adjusting wal_skip_threshold.  Users setting a timeout on COMMIT may
need to adjust that timeout, and log_min_duration_statement analysis
will reflect time consumption moving to COMMIT from commands like COPY.

Internally, this requires a reliable determination of whether
RollbackAndReleaseCurrentSubTransaction() would unlink a relation's
current relfilenode.  Introduce rd_firstRelfilenodeSubid.  Amend the
specification of rd_createSubid such that the field is zero when a new
rel has an old rd_node.  Make relcache.c retain entries for certain
dropped relations until end of transaction.

Back-patch to 9.5 (all supported versions).  This introduces a new WAL
record type, XLOG_GIST_ASSIGN_LSN, without bumping XLOG_PAGE_MAGIC.  As
always, update standby systems before master systems.  This changes
sizeof(RelationData) and sizeof(IndexStmt), breaking binary
compatibility for affected extensions.  (The most recent commit to
affect the same class of extensions was
089e4d405d0f3b94c74a2c6a54357a84a681754b.)

Kyotaro Horiguchi, reviewed (in earlier, similar versions) by Robert
Haas.  Heikki Linnakangas and Michael Paquier implemented earlier
designs that materially clarified the problem.  Reviewed, in earlier
designs, by Andrew Dunstan, Andres Freund, Alvaro Herrera, Tom Lane,
Fujii Masao, and Simon Riggs.  Reported by Martijn van Oosterhout.

Discussion: https://postgr.es/m/20150702220524.GA9392@svana.org

1 files changed, 26 insertions, 30 deletions

diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 82004117522..54cbbd4a4be 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -4013,19 +4013,14 @@ ATRewriteTable(AlteredTableInfo *tab, Oid OIDNewHeap, LOCKMODE lockmode)
 		newrel = NULL;
 
 	/*
-	 * Prepare a BulkInsertState and options for heap_insert. Because we're
-	 * building a new heap, we can skip WAL-logging and fsync it to disk at
-	 * the end instead (unless WAL-logging is required for archiving or
-	 * streaming replication). The FSM is empty too, so don't bother using it.
+	 * Prepare a BulkInsertState and options for heap_insert.  The FSM is
+	 * empty, so don't bother using it.
 	 */
 	if (newrel)
 	{
 		mycid = GetCurrentCommandId(true);
 		bistate = GetBulkInsertState();
-
 		hi_options = HEAP_INSERT_SKIP_FSM;
-		if (!XLogIsNeeded())
-			hi_options |= HEAP_INSERT_SKIP_WAL;
 	}
 	else
 	{
@@ -4275,10 +4270,6 @@ ATRewriteTable(AlteredTableInfo *tab, Oid OIDNewHeap, LOCKMODE lockmode)
 	{
 		FreeBulkInsertState(bistate);
 
-		/* If we skipped writing WAL, then we need to sync the heap. */
-		if (hi_options & HEAP_INSERT_SKIP_WAL)
-			heap_sync(newrel);
-
 		heap_close(newrel, NoLock);
 	}
 }
@@ -5958,14 +5949,19 @@ ATExecAddIndex(AlteredTableInfo *tab, Relation rel,
 
 	/*
 	 * If TryReuseIndex() stashed a relfilenode for us, we used it for the new
-	 * index instead of building from scratch.  The DROP of the old edition of
-	 * this index will have scheduled the storage for deletion at commit, so
-	 * cancel that pending deletion.
+	 * index instead of building from scratch.  Restore associated fields.
+	 * This may store InvalidSubTransactionId in both fields, in which case
+	 * relcache.c will assume it can rebuild the relcache entry.  Hence, do
+	 * this after the CCI that made catalog rows visible to any rebuild.  The
+	 * DROP of the old edition of this index will have scheduled the storage
+	 * for deletion at commit, so cancel that pending deletion.
 	 */
 	if (OidIsValid(stmt->oldNode))
 	{
 		Relation	irel = index_open(address.objectId, NoLock);
 
+		irel->rd_createSubid = stmt->oldCreateSubid;
+		irel->rd_firstRelfilenodeSubid = stmt->oldFirstRelfilenodeSubid;
 		RelationPreserveStorage(irel->rd_node, true);
 		index_close(irel, NoLock);
 	}
@@ -9104,6 +9100,8 @@ TryReuseIndex(Oid oldId, IndexStmt *stmt)
 		Relation	irel = index_open(oldId, NoLock);
 
 		stmt->oldNode = irel->rd_node.relNode;
+		stmt->oldCreateSubid = irel->rd_createSubid;
+		stmt->oldFirstRelfilenodeSubid = irel->rd_firstRelfilenodeSubid;
 		index_close(irel, NoLock);
 	}
 }
@@ -9949,6 +9947,8 @@ ATExecSetTableSpace(Oid tableOid, Oid newTableSpace, LOCKMODE lockmode)
 
 	heap_close(pg_class, RowExclusiveLock);
 
+	RelationAssumeNewRelfilenode(rel);
+
 	relation_close(rel, NoLock);
 
 	/* Make sure the reltablespace change is visible */
@@ -10163,7 +10163,9 @@ copy_relation_data(SMgrRelation src, SMgrRelation dst,
 
 	/*
 	 * We need to log the copied data in WAL iff WAL archiving/streaming is
-	 * enabled AND it's a permanent relation.
+	 * enabled AND it's a permanent relation.  This gives the same answer as
+	 * "RelationNeedsWAL(rel) || copying_initfork", because we know the
+	 * current operation created a new relfilenode.
 	 */
 	use_wal = XLogIsNeeded() &&
 		(relpersistence == RELPERSISTENCE_PERMANENT || copying_initfork);
@@ -10205,21 +10207,15 @@ copy_relation_data(SMgrRelation src, SMgrRelation dst,
 	}
 
 	/*
-	 * If the rel is WAL-logged, must fsync before commit.  We use heap_sync
-	 * to ensure that the toast table gets fsync'd too.  (For a temp or
-	 * unlogged rel we don't care since the data will be gone after a crash
-	 * anyway.)
-	 *
-	 * It's obvious that we must do this when not WAL-logging the copy. It's
-	 * less obvious that we have to do it even if we did WAL-log the copied
-	 * pages. The reason is that since we're copying outside shared buffers, a
-	 * CHECKPOINT occurring during the copy has no way to flush the previously
-	 * written data to disk (indeed it won't know the new rel even exists).  A
-	 * crash later on would replay WAL from the checkpoint, therefore it
-	 * wouldn't replay our earlier WAL entries. If we do not fsync those pages
-	 * here, they might still not be on disk when the crash occurs.
-	 */
-	if (relpersistence == RELPERSISTENCE_PERMANENT || copying_initfork)
+	 * When we WAL-logged rel pages, we must nonetheless fsync them.  The
+	 * reason is that since we're copying outside shared buffers, a CHECKPOINT
+	 * occurring during the copy has no way to flush the previously written
+	 * data to disk (indeed it won't know the new rel even exists).  A crash
+	 * later on would replay WAL from the checkpoint, therefore it wouldn't
+	 * replay our earlier WAL entries. If we do not fsync those pages here,
+	 * they might still not be on disk when the crash occurs.
+	 */
+	if (use_wal || copying_initfork)
 		smgrimmedsync(dst, forkNum);
 }