Avoid having vacuum set reltuples to 0 on non-empty relations in the

RhodiumToad · RhodiumToad · commit 8f8a5aefc735 · 2017-03-16T22:33:18.000Z
presence of page pins, which leads to serious estimation errors in the planner. This particularly affects small heavily-accessed tables, especially where locking (e.g. from FK constraints) forces frequent vacuums for mxid cleanup. Fix by keeping separate track of pages whose live tuples were actually counted vs. pages that were only scanned for freezing purposes. Thus, reltuples can only be set to 0 if all pages of the relation were actually counted. Backpatch to all supported versions. Per bug #14057 from Nicolas Baccelli, analyzed by me. Discussion: https://postgr.es/m/20160331103739.8956.94469@wrigleys.postgresql.org
diff --git a/src/backend/commands/vacuumlazy.c b/src/backend/commands/vacuumlazy.c
@@ -103,7 +103,8 @@ typedef struct LVRelStats
 	BlockNumber old_rel_pages;	/* previous value of pg_class.relpages */
 	BlockNumber rel_pages;		/* total number of pages */
 	BlockNumber scanned_pages;	/* number of pages we examined */
-	double		scanned_tuples; /* counts only tuples on scanned pages */
+	BlockNumber	tupcount_pages;	/* pages whose tuples we counted */
+	double		scanned_tuples; /* counts only tuples on tupcount_pages */
 	double		old_rel_tuples; /* previous value of pg_class.reltuples */
 	double		new_rel_tuples; /* new estimated total # of tuples */
 	BlockNumber pages_removed;
@@ -278,6 +279,10 @@ lazy_vacuum_rel(Relation onerel, VacuumStmt *vacstmt,
 	 * density") with nonzero relpages and reltuples=0 (which means "zero
 	 * tuple density") unless there's some actual evidence for the latter.
 	 *
+	 * It's important that we use tupcount_pages and not scanned_pages for the
+	 * check described above; scanned_pages counts pages where we could not get
+	 * cleanup lock, and which were processed only for frozenxid purposes.
+	 *
 	 * We do update relallvisible even in the corner case, since if the table
 	 * is all-visible we'd definitely like to know that.  But clamp the value
 	 * to be not more than what we're setting relpages to.
@@ -287,7 +292,7 @@ lazy_vacuum_rel(Relation onerel, VacuumStmt *vacstmt,
 	 */
 	new_rel_pages = vacrelstats->rel_pages;
 	new_rel_tuples = vacrelstats->new_rel_tuples;
-	if (vacrelstats->scanned_pages == 0 && new_rel_pages > 0)
+	if (vacrelstats->tupcount_pages == 0 && new_rel_pages > 0)
 	{
 		new_rel_pages = vacrelstats->old_rel_pages;
 		new_rel_tuples = vacrelstats->old_rel_tuples;
@@ -446,6 +451,7 @@ lazy_scan_heap(Relation onerel, LVRelStats *vacrelstats,
 	nblocks = RelationGetNumberOfBlocks(onerel);
 	vacrelstats->rel_pages = nblocks;
 	vacrelstats->scanned_pages = 0;
+	vacrelstats->tupcount_pages = 0;
 	vacrelstats->nonempty_pages = 0;
 	vacrelstats->latestRemovedXid = InvalidTransactionId;
 
@@ -636,6 +642,7 @@ lazy_scan_heap(Relation onerel, LVRelStats *vacrelstats,
 		}
 
 		vacrelstats->scanned_pages++;
+		vacrelstats->tupcount_pages++;
 
 		page = BufferGetPage(buf);
 
@@ -1046,7 +1053,7 @@ lazy_scan_heap(Relation onerel, LVRelStats *vacrelstats,
 	/* now we can compute the new value for pg_class.reltuples */
 	vacrelstats->new_rel_tuples = vac_estimate_reltuples(onerel, false,
 														 nblocks,
-												  vacrelstats->scanned_pages,
+												  vacrelstats->tupcount_pages,
 														 num_tuples);
 
 	/*
@@ -1350,7 +1357,7 @@ lazy_cleanup_index(Relation indrel,
 
 	ivinfo.index = indrel;
 	ivinfo.analyze_only = false;
-	ivinfo.estimated_count = (vacrelstats->scanned_pages < vacrelstats->rel_pages);
+	ivinfo.estimated_count = (vacrelstats->tupcount_pages < vacrelstats->rel_pages);
 	ivinfo.message_level = elevel;
 	ivinfo.num_heap_tuples = vacrelstats->new_rel_tuples;
 	ivinfo.strategy = vac_strategy;
