vacuumlazy.c

PostgreSQL7.4.6 for Linux

			lazy_record_free_space(vacrelstats, blkno,
								   PageGetFreeSpace(page));
		}

		/* Remember the location of the last page with nonremovable tuples */
		if (hastup)
			vacrelstats->nonempty_pages = blkno + 1;

		LockBuffer(buf, BUFFER_LOCK_UNLOCK);

		if (pgchanged)
			SetBufferCommitInfoNeedsSave(buf);

		ReleaseBuffer(buf);
	}

	/* save stats for use later */
	vacrelstats->rel_tuples = num_tuples;

	/* If any tuples need to be deleted, perform final vacuum cycle */
	/* XXX put a threshold on min number of tuples here? */
	if (vacrelstats->num_dead_tuples > 0)
	{
		/* Remove index entries */
		for (i = 0; i < nindexes; i++)
			lazy_vacuum_index(Irel[i], vacrelstats);
		/* Remove tuples from heap */
		lazy_vacuum_heap(onerel, vacrelstats);
	}
	else
	{
		/* Must do post-vacuum cleanup and statistics update anyway */
		for (i = 0; i < nindexes; i++)
			lazy_scan_index(Irel[i], vacrelstats);
	}

	ereport(elevel,
			(errmsg("\"%s\": found %.0f removable, %.0f nonremovable row versions in %u pages",
					RelationGetRelationName(onerel),
					tups_vacuumed, num_tuples, nblocks),
			 errdetail("%.0f dead row versions cannot be removed yet.\n"
					   "There were %.0f unused item pointers.\n"
					   "%u pages are entirely empty.\n"
					   "%s",
					   nkeep,
					   nunused,
					   empty_pages,
					   vac_show_rusage(&ru0))));
}


/*
 *	lazy_vacuum_heap() -- second pass over the heap
 *
 *		This routine marks dead tuples as unused and compacts out free
 *		space on their pages.  Pages not having dead tuples recorded from
 *		lazy_scan_heap are not visited at all.
 *
 * Note: the reason for doing this as a second pass is we cannot remove
 * the tuples until we've removed their index entries, and we want to
 * process index entry removal in batches as large as possible.
 */
static void
lazy_vacuum_heap(Relation onerel, LVRelStats *vacrelstats)
{
	int			tupindex;
	int			npages;
	VacRUsage	ru0;

	vac_init_rusage(&ru0);
	npages = 0;

	tupindex = 0;
	while (tupindex < vacrelstats->num_dead_tuples)
	{
		BlockNumber tblk;
		Buffer		buf;
		Page		page;

		CHECK_FOR_INTERRUPTS();

		tblk = ItemPointerGetBlockNumber(&vacrelstats->dead_tuples[tupindex]);
		buf = ReadBuffer(onerel, tblk);
		LockBufferForCleanup(buf);
		tupindex = lazy_vacuum_page(onerel, tblk, buf, tupindex, vacrelstats);
		/* Now that we've compacted the page, record its available space */
		page = BufferGetPage(buf);
		lazy_record_free_space(vacrelstats, tblk,
							   PageGetFreeSpace(page));
		LockBuffer(buf, BUFFER_LOCK_UNLOCK);
		WriteBuffer(buf);
		npages++;
	}

	ereport(elevel,
			(errmsg("\"%s\": removed %d row versions in %d pages",
					RelationGetRelationName(onerel),
					tupindex, npages),
			 errdetail("%s",
					   vac_show_rusage(&ru0))));
}

/*
 *	lazy_vacuum_page() -- free dead tuples on a page
 *					 and repair its fragmentation.
 *
 * Caller is expected to handle reading, locking, and writing the buffer.
 *
 * tupindex is the index in vacrelstats->dead_tuples of the first dead
 * tuple for this page.  We assume the rest follow sequentially.
 * The return value is the first tupindex after the tuples of this page.
 */
static int
lazy_vacuum_page(Relation onerel, BlockNumber blkno, Buffer buffer,
				 int tupindex, LVRelStats *vacrelstats)
{
	OffsetNumber unused[BLCKSZ / sizeof(OffsetNumber)];
	int			uncnt;
	Page		page = BufferGetPage(buffer);
	ItemId		itemid;

	START_CRIT_SECTION();
	for (; tupindex < vacrelstats->num_dead_tuples; tupindex++)
	{
		BlockNumber tblk;
		OffsetNumber toff;

		tblk = ItemPointerGetBlockNumber(&vacrelstats->dead_tuples[tupindex]);
		if (tblk != blkno)
			break;				/* past end of tuples for this block */
		toff = ItemPointerGetOffsetNumber(&vacrelstats->dead_tuples[tupindex]);
		itemid = PageGetItemId(page, toff);
		itemid->lp_flags &= ~LP_USED;
	}

	uncnt = PageRepairFragmentation(page, unused);

	/* XLOG stuff */
	if (!onerel->rd_istemp)
	{
		XLogRecPtr	recptr;

		recptr = log_heap_clean(onerel, buffer, unused, uncnt);
		PageSetLSN(page, recptr);
		PageSetSUI(page, ThisStartUpID);
	}
	else
	{
		/* No XLOG record, but still need to flag that XID exists on disk */
		MyXactMadeTempRelUpdate = true;
	}

	END_CRIT_SECTION();

	return tupindex;
}

/*
 *	lazy_scan_index() -- scan one index relation to update pg_class statistic.
 *
 * We use this when we have no deletions to do.
 */
static void
lazy_scan_index(Relation indrel, LVRelStats *vacrelstats)
{
	IndexBulkDeleteResult *stats;
	IndexVacuumCleanupInfo vcinfo;
	VacRUsage	ru0;

	vac_init_rusage(&ru0);

	/*
	 * If index is unsafe for concurrent access, must lock it.
	 */
	if (!indrel->rd_am->amconcurrent)
		LockRelation(indrel, AccessExclusiveLock);

	/*
	 * Even though we're not planning to delete anything, we use the
	 * ambulkdelete call, because (a) the scan happens within the index AM
	 * for more speed, and (b) it may want to pass private statistics to
	 * the amvacuumcleanup call.
	 */
	stats = index_bulk_delete(indrel, dummy_tid_reaped, NULL);

	/* Do post-VACUUM cleanup, even though we deleted nothing */
	vcinfo.vacuum_full = false;
	vcinfo.message_level = elevel;

	stats = index_vacuum_cleanup(indrel, &vcinfo, stats);

	/*
	 * Release lock acquired above.
	 */
	if (!indrel->rd_am->amconcurrent)
		UnlockRelation(indrel, AccessExclusiveLock);

	if (!stats)
		return;

	/* now update statistics in pg_class */
	vac_update_relstats(RelationGetRelid(indrel),
						stats->num_pages, stats->num_index_tuples,
						false);

	ereport(elevel,
			(errmsg("index \"%s\" now contains %.0f row versions in %u pages",
					RelationGetRelationName(indrel),
					stats->num_index_tuples,
					stats->num_pages),
			 errdetail("%u index pages have been deleted, %u are currently reusable.\n"
					   "%s",
					   stats->pages_deleted, stats->pages_free,
					   vac_show_rusage(&ru0))));

	pfree(stats);
}

/*
 *	lazy_vacuum_index() -- vacuum one index relation.
 *
 *		Delete all the index entries pointing to tuples listed in
 *		vacrelstats->dead_tuples.
 *
 *		Finally, we arrange to update the index relation's statistics in
 *		pg_class.
 */
static void
lazy_vacuum_index(Relation indrel, LVRelStats *vacrelstats)
{
	IndexBulkDeleteResult *stats;
	IndexVacuumCleanupInfo vcinfo;
	VacRUsage	ru0;

	vac_init_rusage(&ru0);

	/*
	 * If index is unsafe for concurrent access, must lock it.
	 */
	if (!indrel->rd_am->amconcurrent)
		LockRelation(indrel, AccessExclusiveLock);

	/* Do bulk deletion */
	stats = index_bulk_delete(indrel, lazy_tid_reaped, (void *) vacrelstats);

	/* Do post-VACUUM cleanup */
	vcinfo.vacuum_full = false;
	vcinfo.message_level = elevel;

	stats = index_vacuum_cleanup(indrel, &vcinfo, stats);

	/*
	 * Release lock acquired above.
	 */
	if (!indrel->rd_am->amconcurrent)
		UnlockRelation(indrel, AccessExclusiveLock);

	if (!stats)
		return;

	/* now update statistics in pg_class */
	vac_update_relstats(RelationGetRelid(indrel),
						stats->num_pages, stats->num_index_tuples,
						false);

	ereport(elevel,
			(errmsg("index \"%s\" now contains %.0f row versions in %u pages",
					RelationGetRelationName(indrel),
					stats->num_index_tuples,
					stats->num_pages),
			 errdetail("%.0f index row versions were removed.\n"
		 "%u index pages have been deleted, %u are currently reusable.\n"
					   "%s",
					   stats->tuples_removed,
					   stats->pages_deleted, stats->pages_free,
					   vac_show_rusage(&ru0))));

	pfree(stats);
}

/*
 * lazy_truncate_heap - try to truncate off any empty pages at the end
 */
static void
lazy_truncate_heap(Relation onerel, LVRelStats *vacrelstats)
{
	BlockNumber old_rel_pages = vacrelstats->rel_pages;
	BlockNumber new_rel_pages;
	PageFreeSpaceInfo *pageSpaces;
	int			n;
	int			i,
				j;
	VacRUsage	ru0;

	vac_init_rusage(&ru0);

	/*
	 * We need full exclusive lock on the relation in order to do
	 * truncation. If we can't get it, give up rather than waiting --- we
	 * don't want to block other backends, and we don't want to deadlock
	 * (which is quite possible considering we already hold a lower-grade
	 * lock).
	 */
	if (!ConditionalLockRelation(onerel, AccessExclusiveLock))
		return;

	/*
	 * Now that we have exclusive lock, look to see if the rel has grown
	 * whilst we were vacuuming with non-exclusive lock.  If so, give up;
	 * the newly added pages presumably contain non-deletable tuples.
	 */
	new_rel_pages = RelationGetNumberOfBlocks(onerel);
	if (new_rel_pages != old_rel_pages)
	{
		/* might as well use the latest news when we update pg_class stats */
		vacrelstats->rel_pages = new_rel_pages;
		UnlockRelation(onerel, AccessExclusiveLock);
		return;
	}

	/*
	 * Scan backwards from the end to verify that the end pages actually
	 * contain nothing we need to keep.  This is *necessary*, not
	 * optional, because other backends could have added tuples to these
	 * pages whilst we were vacuuming.
	 */
	new_rel_pages = count_nondeletable_pages(onerel, vacrelstats);

	if (new_rel_pages >= old_rel_pages)
	{
		/* can't do anything after all */
		UnlockRelation(onerel, AccessExclusiveLock);
		return;
	}

	/*
	 * Okay to truncate.
	 *
	 * First, flush any shared buffers for the blocks we intend to delete.
	 * FlushRelationBuffers is a bit more than we need for this, since it
	 * will also write out dirty buffers for blocks we aren't deleting,
	 * but it's the closest thing in bufmgr's API.
	 */
	i = FlushRelationBuffers(onerel, new_rel_pages);
	if (i < 0)
		elog(ERROR, "FlushRelationBuffers returned %d", i);

	/*
	 * Do the physical truncation.
	 */
	new_rel_pages = smgrtruncate(DEFAULT_SMGR, onerel, new_rel_pages);
	onerel->rd_nblocks = new_rel_pages; /* update relcache immediately */
	onerel->rd_targblock = InvalidBlockNumber;
	vacrelstats->rel_pages = new_rel_pages;		/* save new number of
												 * blocks */

	/*
	 * Drop free-space info for removed blocks; these must not get entered
	 * into the FSM!
	 */
	pageSpaces = vacrelstats->free_pages;
	n = vacrelstats->num_free_pages;
	j = 0;
	for (i = 0; i < n; i++)
	{
		if (pageSpaces[i].blkno < new_rel_pages)
		{
			pageSpaces[j] = pageSpaces[i];
			j++;
		}
	}
	vacrelstats->num_free_pages = j;
	/* We destroyed the heap ordering, so mark array unordered */
	vacrelstats->fs_is_heap = false;

	/*
	 * We keep the exclusive lock until commit (perhaps not necessary)?
	 */

	ereport(elevel,
			(errmsg("\"%s\": truncated %u to %u pages",
					RelationGetRelationName(onerel),
					old_rel_pages, new_rel_pages),
			 errdetail("%s",
					   vac_show_rusage(&ru0))));
}

/*
 * Rescan end pages to verify that they are (still) empty of needed tuples.
 *