Add C comment about why synchronous_commit=off behavior can lose

bmomjian · bmomjian · commit b57ddccf0503 · 2010-06-29T18:44:58.000Z
committed transactions in a postmaster crash.
diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
@@ -10,7 +10,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/transam/xact.c,v 1.291 2010/05/13 11:39:30 sriggs Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/transam/xact.c,v 1.292 2010/06/29 18:44:58 momjian Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1028,7 +1028,7 @@ RecordTransactionCommit(void)
 	if (XactSyncCommit || forceSyncCommit || haveNonTemp)
 	{
 		/*
-		 * Synchronous commit case.
+		 * Synchronous commit case:
 		 *
 		 * Sleep before flush! So we can flush more than one commit records
 		 * per single fsync.  (The idea is some other backend may do the
@@ -1054,7 +1054,12 @@ RecordTransactionCommit(void)
 	else
 	{
 		/*
-		 * Asynchronous commit case.
+		 * Asynchronous commit case:
+		 *
+		 * This enables possible committed transaction loss in the case of a
+		 * postmaster crash because WAL buffers are left unwritten.
+		 * Ideally we could issue the WAL write without the fsync, but
+		 * some wal_sync_methods do not allow separate write/fsync.
 		 *
 		 * Report the latest async commit LSN, so that the WAL writer knows to
 		 * flush this commit.

Original file line number	Diff line number	Diff line change
`@@ -10,7 +10,7 @@`
`10`	`10`	`*`
`11`	`11`	`*`
`12`	`12`	`* IDENTIFICATION`
`13`		`- * $PostgreSQL: pgsql/src/backend/access/transam/xact.c,v 1.291 2010/05/13 11:39:30 sriggs Exp $`
	`13`	`+ * $PostgreSQL: pgsql/src/backend/access/transam/xact.c,v 1.292 2010/06/29 18:44:58 momjian Exp $`
`14`	`14`	`*`
`15`	`15`	`*-------------------------------------------------------------------------`
`16`	`16`	`*/`
`@@ -1028,7 +1028,7 @@ RecordTransactionCommit(void)`
`1028`	`1028`	`if (XactSyncCommit \|\| forceSyncCommit \|\| haveNonTemp)`
`1029`	`1029`	`{`
`1030`	`1030`	`/*`
`1031`		`- * Synchronous commit case.`
	`1031`	`+ * Synchronous commit case:`
`1032`	`1032`	`*`
`1033`	`1033`	`* Sleep before flush! So we can flush more than one commit records`
`1034`	`1034`	`* per single fsync. (The idea is some other backend may do the`
`@@ -1054,7 +1054,12 @@ RecordTransactionCommit(void)`
`1054`	`1054`	`else`
`1055`	`1055`	`{`
`1056`	`1056`	`/*`
`1057`		`- * Asynchronous commit case.`
	`1057`	`+ * Asynchronous commit case:`
	`1058`	`+ *`
	`1059`	`+ * This enables possible committed transaction loss in the case of a`
	`1060`	`+ * postmaster crash because WAL buffers are left unwritten.`
	`1061`	`+ * Ideally we could issue the WAL write without the fsync, but`
	`1062`	`+ * some wal_sync_methods do not allow separate write/fsync.`
`1058`	`1063`	`*`
`1059`	`1064`	`* Report the latest async commit LSN, so that the WAL writer knows to`
`1060`	`1065`	`* flush this commit.`