Autovacuum loose end mop-up. Provide autovacuum-specific vacuum cost

delay and limit, both as global GUCs and as table-specific entries in
pg_autovacuum.  stats_reset_on_server_start is now OFF by default,
but a reset is forced if we did WAL replay.  XID-wrap vacuums do not
ANALYZE, but do FREEZE if it's a template database.  Alvaro Herrera

src/backend/utils/init/flatfiles.c

@@ -23,7 +23,7 @@
  * Portions Copyright (c) 1996-2005, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/backend/utils/init/flatfiles.c,v 1.13 2005/07/28 22:27:02 tgl Exp $
+ * $PostgreSQL: pgsql/src/backend/utils/init/flatfiles.c,v 1.14 2005/08/11 21:11:46 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -208,12 +208,14 @@ write_database_file(Relation drel)
 		char	   *datname;
 		Oid			datoid;
 		Oid			dattablespace;
-		TransactionId datfrozenxid;
+		TransactionId datfrozenxid,
+					  datvacuumxid;
 
 		datname = NameStr(dbform->datname);
 		datoid = HeapTupleGetOid(tuple);
 		dattablespace = dbform->dattablespace;
 		datfrozenxid = dbform->datfrozenxid;
+		datvacuumxid = dbform->datvacuumxid;
 
 		/*
 		 * Identify the oldest datfrozenxid, ignoring databases that are not
@@ -242,13 +244,14 @@ write_database_file(Relation drel)
 		}
 
 		/*
-		 * The file format is: "dbname" oid tablespace frozenxid
+		 * The file format is: "dbname" oid tablespace frozenxid vacuumxid
 		 *
-		 * The xid is not needed for backend startup, but may be of use
-		 * for forensic purposes.
+		 * The xids are not needed for backend startup, but are of use to
+		 * autovacuum, and might also be helpful for forensic purposes.
 		 */
 		fputs_quote(datname, fp);
-		fprintf(fp, " %u %u %u\n", datoid, dattablespace, datfrozenxid);
+		fprintf(fp, " %u %u %u %u\n",
+				datoid, dattablespace, datfrozenxid, datvacuumxid);
 	}
 	heap_endscan(scan);
 
@@ -654,8 +657,10 @@ write_auth_file(Relation rel_authid, Relation rel_authmem)
  * base backup which may be far out of sync with the current state.
  *
  * In theory we could skip rebuilding the flat files if no WAL replay
- * occurred, but it seems safest to just do it always.  We have to
- * scan pg_database to compute the XID wrap limit anyway.
+ * occurred, but it seems best to just do it always.  We have to
+ * scan pg_database to compute the XID wrap limit anyway.  Also, this
+ * policy means we need not force initdb to change the format of the
+ * flat files.
  *
  * In a standalone backend we pass database_only = true to skip processing
  * the auth file.  We won't need it, and building it could fail if there's

src/backend/utils/init/postinit.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/init/postinit.c,v 1.156 2005/08/08 03:12:14 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/init/postinit.c,v 1.157 2005/08/11 21:11:46 tgl Exp $
  *
  *
  *-------------------------------------------------------------------------
@@ -20,8 +20,10 @@
 #include <math.h>
 #include <unistd.h>
 
+#include "access/genam.h"
 #include "access/heapam.h"
 #include "catalog/catalog.h"
+#include "catalog/indexing.h"
 #include "catalog/namespace.h"
 #include "catalog/pg_authid.h"
 #include "catalog/pg_database.h"
@@ -79,7 +81,7 @@ FindMyDatabase(const char *name, Oid *db_id, Oid *db_tablespace)
 	char	   *filename;
 	FILE	   *db_file;
 	char		thisname[NAMEDATALEN];
-	TransactionId frozenxid;
+	TransactionId dummyxid;
 
 	filename = database_getflatfilename();
 	db_file = AllocateFile(filename, "r");
@@ -89,7 +91,8 @@ FindMyDatabase(const char *name, Oid *db_id, Oid *db_tablespace)
 				 errmsg("could not open file \"%s\": %m", filename)));
 
 	while (read_pg_database_line(db_file, thisname, db_id,
-								 db_tablespace, &frozenxid))
+								 db_tablespace, &dummyxid,
+								 &dummyxid))
 	{
 		if (strcmp(thisname, name) == 0)
 		{
@@ -131,7 +134,7 @@ static void
 ReverifyMyDatabase(const char *name)
 {
 	Relation	pgdbrel;
-	HeapScanDesc pgdbscan;
+	SysScanDesc	pgdbscan;
 	ScanKeyData key;
 	HeapTuple	tup;
 	Form_pg_database dbform;
@@ -147,9 +150,10 @@ ReverifyMyDatabase(const char *name)
 				BTEqualStrategyNumber, F_NAMEEQ,
 				NameGetDatum(name));
 
-	pgdbscan = heap_beginscan(pgdbrel, SnapshotNow, 1, &key);
+	pgdbscan = systable_beginscan(pgdbrel, DatabaseNameIndexId, true,
+								  SnapshotNow, 1, &key);
 
-	tup = heap_getnext(pgdbscan, ForwardScanDirection);
+	tup = systable_getnext(pgdbscan);
 	if (!HeapTupleIsValid(tup) ||
 		HeapTupleGetOid(tup) != MyDatabaseId)
 	{
@@ -238,7 +242,7 @@ ReverifyMyDatabase(const char *name)
 		}
 	}
 
-	heap_endscan(pgdbscan);
+	systable_endscan(pgdbscan);
 	heap_close(pgdbrel, RowShareLock);
 }
 
@@ -428,6 +432,18 @@ InitPostgres(const char *dbname, const char *username)
 	/* Initialize portal manager */
 	EnablePortalManager();
 
+	/*
+	 * Set up process-exit callback to do pre-shutdown cleanup.  This
+	 * has to be after we've initialized all the low-level modules
+	 * like the buffer manager, because during shutdown this has to
+	 * run before the low-level modules start to close down.  On the
+	 * other hand, we want it in place before we begin our first
+	 * transaction --- if we fail during the initialization transaction,
+	 * as is entirely possible, we need the AbortTransaction call to
+	 * clean up.
+	 */
+	on_shmem_exit(ShutdownPostgres, 0);
+
 	/* start a new transaction here before access to db */
 	if (!bootstrap)
 		StartTransactionCommand();
@@ -465,7 +481,8 @@ InitPostgres(const char *dbname, const char *username)
 	/*
 	 * Unless we are bootstrapping, double-check that InitMyDatabaseInfo()
 	 * got a correct result.  We can't do this until all the
-	 * database-access infrastructure is up.
+	 * database-access infrastructure is up.  (Also, it wants to know if
+	 * the user is a superuser, so the above stuff has to happen first.)
 	 */
 	if (!bootstrap)
 		ReverifyMyDatabase(dbname);
@@ -509,24 +526,10 @@ InitPostgres(const char *dbname, const char *username)
 	/* initialize client encoding */
 	InitializeClientEncoding();
 
-	/*
-	 * Initialize statistics collection for this backend.  We do this
-	 * here because the shutdown hook it sets up needs to be invoked
-	 * at the corresponding phase of backend shutdown: after
-	 * ShutdownPostgres and before we drop access to shared memory.
-	 */
+	/* initialize statistics collection for this backend */
 	if (IsUnderPostmaster)
 		pgstat_bestart();
 
-	/*
-	 * Set up process-exit callback to do pre-shutdown cleanup.  This
-	 * should be last because we want shmem_exit to call this routine
-	 * before the exit callbacks that are registered by buffer manager,
-	 * lock manager, etc. We need to run this code before we close down
-	 * database access!
-	 */
-	on_shmem_exit(ShutdownPostgres, 0);
-
 	/* close the transaction we started above */
 	if (!bootstrap)
 		CommitTransactionCommand();
@@ -538,9 +541,7 @@ InitPostgres(const char *dbname, const char *username)
 /*
  * Backend-shutdown callback.  Do cleanup that we want to be sure happens
  * before all the supporting modules begin to nail their doors shut via
- * their own callbacks.  Note that because this has to be registered very
- * late in startup, it will not get called if we suffer a failure *during*
- * startup.
+ * their own callbacks.
  *
  * User-level cleanup, such as temp-relation removal and UNLISTEN, happens
  * via separate callbacks that execute before this one.  We don't combine the

src/backend/utils/misc/guc.c

@@ -10,7 +10,7 @@
  * Written by Peter Eisentraut <peter_e@gmx.net>.
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/misc/guc.c,v 1.280 2005/07/30 15:17:20 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/misc/guc.c,v 1.281 2005/08/11 21:11:47 tgl Exp $
  *
  *--------------------------------------------------------------------
  */
@@ -672,7 +672,7 @@ static struct config_bool ConfigureNamesBool[] =
 			NULL
 		},
 		&pgstat_collect_resetonpmstart,
-		true, NULL, NULL
+		false, NULL, NULL
 	},
 	{
 		{"stats_command_string", PGC_SUSET, STATS_COLLECTOR,
@@ -1160,6 +1160,24 @@ static struct config_int ConfigureNamesInt[] =
 		0, 0, 1000, NULL, NULL
 	},
 
+	{
+		{"autovacuum_vacuum_cost_delay", PGC_SIGHUP, AUTOVACUUM,
+			gettext_noop("Vacuum cost delay in milliseconds, for autovacuum."),
+			NULL
+		},
+		&autovacuum_vac_cost_delay,
+		-1, -1, 1000, NULL, NULL
+	},
+
+	{
+		{"autovacuum_vacuum_cost_limit", PGC_SIGHUP, AUTOVACUUM,
+			gettext_noop("Vacuum cost amount available before napping, for autovacuum."),
+			NULL
+		},
+		&autovacuum_vac_cost_limit,
+		-1, -1, 10000, NULL, NULL
+	},
+
 	{
 		{"max_files_per_process", PGC_POSTMASTER, RESOURCES_KERNEL,
 			gettext_noop("Sets the maximum number of simultaneously open files for each server process."),

src/backend/utils/misc/postgresql.conf.sample

@@ -288,7 +288,7 @@
 #stats_command_string = off
 #stats_block_level = off
 #stats_row_level = off
-#stats_reset_on_server_start = on
+#stats_reset_on_server_start = off
 
 
 #---------------------------------------------------------------------------
@@ -301,6 +301,10 @@
 #autovacuum_analyze_threshold = 500	# min # of tuple updates before analyze
 #autovacuum_vacuum_scale_factor = 0.4	# fraction of rel size before vacuum
 #autovacuum_analyze_scale_factor = 0.2	# fraction of rel size before analyze
+#autovacuum_vacuum_cost_delay = -1  # default vacuum cost delay for autovac
+                                    # -1 means use vacuum_cost_delay
+#autovacuum_vacuum_cost_limit = -1  # default vacuum cost limit for autovac
+                                    # -1 means use vacuum_cost_limit
 
 
 #---------------------------------------------------------------------------