pgstat: scaffolding for transactional stats creation / drop.

One problematic part of the current statistics collector design is that there
is no reliable way of getting rid of statistics entries. Because of that
pgstat_vacuum_stat() (called by [auto-]vacuum) matches all stats for the
current database with the catalog contents and tries to drop now-superfluous
entries. That's quite expensive. What's worse, it doesn't work on physical
replicas, despite physical replicas collection statistics entries.

This commit introduces infrastructure to create / drop statistics entries
transactionally, together with the underlying catalog objects (functions,
relations, subscriptions). pgstat_xact.c maintains a list of stats entries
created / dropped transactionally in the current transaction. To ensure the
removal of statistics entries is durable dropped statistics entries are
included in commit / abort (and prepare) records, which also ensures that
stats entries are dropped on standbys.

Statistics entries created separately from creating the underlying catalog
object (e.g. when stats were previously lost due to an immediate restart)
are *not* WAL logged. However that can only happen outside of the transaction
creating the catalog object, so it does not lead to "leaked" statistics
entries.

For this to work, functions creating / dropping functions / relations /
subscriptions need to call into pgstat. For subscriptions this was already
done when dropping subscriptions, via pgstat_report_subscription_drop() (now
renamed to pgstat_drop_subscription()).

This commit does not actually drop stats yet, it just provides the
infrastructure. It is however a largely independent piece of infrastructure,
so committing it separately makes sense.

Bumps XLOG_PAGE_MAGIC.

Author: Andres Freund <andres@anarazel.de>
Reviewed-By: Thomas Munro <thomas.munro@gmail.com>
Reviewed-By: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
Discussion: https://postgr.es/m/20220303021600.hs34ghqcw6zcokdh@alap3.anarazel.de

4 files changed, 63 insertions, 3 deletions

diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 062cc7e17d8..4e1e8735010 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -180,6 +180,7 @@ typedef struct SavedTransactionCharacteristics
 #define XACT_XINFO_HAS_ORIGIN			(1U << 5)
 #define XACT_XINFO_HAS_AE_LOCKS			(1U << 6)
 #define XACT_XINFO_HAS_GID				(1U << 7)
+#define XACT_XINFO_HAS_DROPPED_STATS	(1U << 8)
 
 /*
  * Also stored in xinfo, these indicating a variety of additional actions that
@@ -230,7 +231,7 @@ typedef struct xl_xact_assignment
 typedef struct xl_xact_xinfo
 {
 	/*
-	 * Even though we right now only require 1 byte of space in xinfo we use
+	 * Even though we right now only require two bytes of space in xinfo we use
 	 * four so following records don't have to care about alignment. Commit
 	 * records can be large, so copying large portions isn't attractive.
 	 */
@@ -257,6 +258,27 @@ typedef struct xl_xact_relfilenodes
 } xl_xact_relfilenodes;
 #define MinSizeOfXactRelfilenodes offsetof(xl_xact_relfilenodes, xnodes)
 
+/*
+ * A transactionally dropped statistics entry.
+ *
+ * Declared here rather than pgstat.h because pgstat.h can't be included from
+ * frontend code, but the WAL format needs to be readable by frontend
+ * programs.
+ */
+typedef struct xl_xact_stats_item
+{
+	int			kind;
+	Oid			dboid;
+	Oid			objoid;
+} xl_xact_stats_item;
+
+typedef struct xl_xact_stats_items
+{
+	int		nitems;
+	xl_xact_stats_item items[FLEXIBLE_ARRAY_MEMBER];
+} xl_xact_stats_items;
+#define MinSizeOfXactStatsItems offsetof(xl_xact_stats_items, items)
+
 typedef struct xl_xact_invals
 {
 	int			nmsgs;			/* number of shared inval msgs */
@@ -283,6 +305,7 @@ typedef struct xl_xact_commit
 	/* xl_xact_dbinfo follows if XINFO_HAS_DBINFO */
 	/* xl_xact_subxacts follows if XINFO_HAS_SUBXACT */
 	/* xl_xact_relfilenodes follows if XINFO_HAS_RELFILENODES */
+	/* xl_xact_stats_items follows if XINFO_HAS_DROPPED_STATS */
 	/* xl_xact_invals follows if XINFO_HAS_INVALS */
 	/* xl_xact_twophase follows if XINFO_HAS_TWOPHASE */
 	/* twophase_gid follows if XINFO_HAS_GID. As a null-terminated string. */
@@ -298,6 +321,7 @@ typedef struct xl_xact_abort
 	/* xl_xact_dbinfo follows if XINFO_HAS_DBINFO */
 	/* xl_xact_subxacts follows if XINFO_HAS_SUBXACT */
 	/* xl_xact_relfilenodes follows if XINFO_HAS_RELFILENODES */
+	/* xl_xact_stats_items follows if XINFO_HAS_DROPPED_STATS */
 	/* No invalidation messages needed. */
 	/* xl_xact_twophase follows if XINFO_HAS_TWOPHASE */
 	/* twophase_gid follows if XINFO_HAS_GID. As a null-terminated string. */
@@ -316,6 +340,8 @@ typedef struct xl_xact_prepare
 	int32		nsubxacts;		/* number of following subxact XIDs */
 	int32		ncommitrels;	/* number of delete-on-commit rels */
 	int32		nabortrels;		/* number of delete-on-abort rels */
+	int32		ncommitstats;	/* number of stats to drop on commit */
+	int32		nabortstats;	/* number of stats to drop on abort */
 	int32		ninvalmsgs;		/* number of cache invalidation messages */
 	bool		initfileinval;	/* does relcache init file need invalidation? */
 	uint16		gidlen;			/* length of the GID - GID follows the header */
@@ -342,6 +368,9 @@ typedef struct xl_xact_parsed_commit
 	int			nrels;
 	RelFileNode *xnodes;
 
+	int			nstats;
+	xl_xact_stats_item *stats;
+
 	int			nmsgs;
 	SharedInvalidationMessage *msgs;
 
@@ -349,6 +378,8 @@ typedef struct xl_xact_parsed_commit
 	char		twophase_gid[GIDSIZE];	/* only for 2PC */
 	int			nabortrels;		/* only for 2PC */
 	RelFileNode *abortnodes;	/* only for 2PC */
+	int			nabortstats;		/* only for 2PC */
+	xl_xact_stats_item *abortstats; /* only for 2PC */
 
 	XLogRecPtr	origin_lsn;
 	TimestampTz origin_timestamp;
@@ -370,6 +401,9 @@ typedef struct xl_xact_parsed_abort
 	int			nrels;
 	RelFileNode *xnodes;
 
+	int			nstats;
+	xl_xact_stats_item *stats;
+
 	TransactionId twophase_xid; /* only for 2PC */
 	char		twophase_gid[GIDSIZE];	/* only for 2PC */
 
@@ -449,6 +483,8 @@ extern int	xactGetCommittedChildren(TransactionId **ptr);
 extern XLogRecPtr XactLogCommitRecord(TimestampTz commit_time,
 									  int nsubxacts, TransactionId *subxacts,
 									  int nrels, RelFileNode *rels,
+									  int nstats,
+									  xl_xact_stats_item *stats,
 									  int nmsgs, SharedInvalidationMessage *msgs,
 									  bool relcacheInval,
 									  int xactflags,
@@ -458,6 +494,8 @@ extern XLogRecPtr XactLogCommitRecord(TimestampTz commit_time,
 extern XLogRecPtr XactLogAbortRecord(TimestampTz abort_time,
 									 int nsubxacts, TransactionId *subxacts,
 									 int nrels, RelFileNode *rels,
+									 int nstats,
+									 xl_xact_stats_item *stats,
 									 int xactflags, TransactionId twophase_xid,
 									 const char *twophase_gid);
 extern void xact_redo(XLogReaderState *record);
diff --git a/src/include/access/xlog_internal.h b/src/include/access/xlog_internal.h
index 0e94833129a..b7c375fed1c 100644
--- a/src/include/access/xlog_internal.h
+++ b/src/include/access/xlog_internal.h
@@ -31,7 +31,7 @@
 /*
  * Each page of XLOG file has a header like this:
  */
-#define XLOG_PAGE_MAGIC 0xD10F	/* can be used as WAL version indicator */
+#define XLOG_PAGE_MAGIC 0xD110	/* can be used as WAL version indicator */
 
 typedef struct XLogPageHeaderData
 {
diff --git a/src/include/pgstat.h b/src/include/pgstat.h
index 9235f4dc4ce..7981a816565 100644
--- a/src/include/pgstat.h
+++ b/src/include/pgstat.h
@@ -1049,6 +1049,9 @@ extern void pgstat_report_connect(Oid dboid);
  * Functions in pgstat_function.c
  */
 
+extern void pgstat_create_function(Oid proid);
+extern void pgstat_drop_function(Oid proid);
+
 struct FunctionCallInfoBaseData;
 extern void pgstat_init_function_usage(struct FunctionCallInfoBaseData *fcinfo,
 									   PgStat_FunctionCallUsage *fcu);
@@ -1062,6 +1065,8 @@ extern PgStat_BackendFunctionEntry *find_funcstat_entry(Oid func_id);
  * Functions in pgstat_relation.c
  */
 
+extern void pgstat_create_relation(Relation rel);
+extern void pgstat_drop_relation(Relation rel);
 extern void pgstat_copy_relation_stats(Relation dstrel, Relation srcrel);
 
 extern void pgstat_relation_init(Relation rel);
@@ -1158,7 +1163,8 @@ extern int	pgstat_slru_index(const char *name);
  */
 
 extern void pgstat_report_subscription_error(Oid subid, bool is_apply_error);
-extern void pgstat_report_subscription_drop(Oid subid);
+extern void pgstat_create_subscription(Oid subid);
+extern void pgstat_drop_subscription(Oid subid);
 
 
 /*
@@ -1169,6 +1175,9 @@ extern void AtEOXact_PgStat(bool isCommit, bool parallel);
 extern void AtEOSubXact_PgStat(bool isCommit, int nestDepth);
 extern void AtPrepare_PgStat(void);
 extern void PostPrepare_PgStat(void);
+struct xl_xact_stats_item;
+extern int	pgstat_get_transactional_drops(bool isCommit, struct xl_xact_stats_item **items);
+extern void pgstat_execute_transactional_drops(int ndrops, struct xl_xact_stats_item *items, bool is_redo);
 
 
 /*
diff --git a/src/include/utils/pgstat_internal.h b/src/include/utils/pgstat_internal.h
index 1923f56a3ad..da66e924dc0 100644
--- a/src/include/utils/pgstat_internal.h
+++ b/src/include/utils/pgstat_internal.h
@@ -43,6 +43,16 @@ typedef struct PgStat_SubXactStatus
 	struct PgStat_SubXactStatus *prev;	/* higher-level subxact if any */
 
 	/*
+	 * Dropping the statistics for objects that dropped transactionally itself
+	 * needs to be transactional. Therefore we collect the stats dropped in
+	 * the current (sub-)transaction and only execute the stats drop when we
+	 * know if the transaction commits/aborts. To handle replicas and crashes,
+	 * stats drops are included in commit records.
+	 */
+	dlist_head	pending_drops;
+	int			pending_drops_count;
+
+	/*
 	 * Tuple insertion/deletion counts for an open transaction can't be
 	 * propagated into PgStat_TableStatus counters until we know if it is
 	 * going to commit or abort.  Hence, we keep these counts in per-subxact
@@ -133,6 +143,9 @@ extern bool pgstat_wal_pending(void);
  */
 
 extern PgStat_SubXactStatus *pgstat_xact_stack_level_get(int nest_level);
+extern void pgstat_drop_transactional(PgStat_Kind kind, Oid dboid, Oid objoid);
+extern void pgstat_create_transactional(PgStat_Kind kind, Oid dboid, Oid objoid);
+
 
 
 /*