On Fri, Feb 9, 2018 at 10:32 AM, Alvaro Herrera <alvhe...@alvh.no-ip.org> wrote:
> Claudio Freire wrote:
>> On Thu, Feb 8, 2018 at 8:39 PM, Alvaro Herrera <alvhe...@alvh.no-ip.org>
>> wrote:
>
>> During the process of developing the patch, I got seriously broken
>> code that passed the tests nonetheless. The test as it was was very
>> ineffective at actually detecting issues.
>>
>> This new test may be slow, but it's effective. That's a very good
>> reason to make it slower, if you ask me.
>
> OK, I don't disagree with improving the test, but if we can make it fast
> *and* effective, that's better than slow and effective.
I'd love to have a test that uses multiple segments of dead tuples,
but for that, it needs to use more than 64MB of mwm. That amounts to,
basically, ~12M rows.
Is there a "slow test suite" where such a test could be added that
won't bother regular "make check"?
That, or we turn the initial segment size into a GUC, but I don't
think it's a useful GUC outside of the test suite.
>> > 3. Figure out the minimum size for the table that triggers the behavior
>> > you want. Right now you use 400k tuples -- maybe 100k are sufficient?
>> > Don't know.
>>
>> For that test, I need enough *dead* tuples to cause several passes.
>> Even small mwm settings require tons of tuples for this. In fact, I'm
>> thinking that number might be too low for its purpose, even. I'll
>> re-check, but I doubt it's too high. If anything, it's too low.
>
> OK.
Turns out that it was a tad oversized. 300k tuples seems enough.
Attached is a new patch version that:
- Uses an unlogged table to make the large mwm test faster
- Uses a wait_barrier helper that waits for concurrent transactions
to finish before vacuuming tables, to make sure deleted tuples
actually are vacuumable
- Tweaks the size of the large mwm test to be as small as possible
- Optimizes the delete to avoid expensive operations yet attain
the same end result
From 20ca670215a7bf63f1bc2290efd5a89bcafabb20 Mon Sep 17 00:00:00 2001
From: Claudio Freire <klaussfre...@gmail.com>
Date: Mon, 12 Sep 2016 23:36:42 -0300
Subject: [PATCH 1/2] Vacuum: allow using more than 1GB work mem
Turn the dead_tuples array into a structure composed of several
exponentially bigger arrays, to enable usage of more than 1GB
of work mem during vacuum and thus reduce the number of full
index scans necessary to remove all dead tids when the memory is
available.
Improve test ability to spot vacuum errors
---
src/backend/commands/vacuumlazy.c | 402 ++++++++++++++++++++++++++++-------
src/test/regress/expected/vacuum.out | 72 ++++++-
src/test/regress/sql/vacuum.sql | 40 +++-
3 files changed, 438 insertions(+), 76 deletions(-)
diff --git a/src/backend/commands/vacuumlazy.c b/src/backend/commands/vacuumlazy.c
index cf7f5e1..1e82d26 100644
--- a/src/backend/commands/vacuumlazy.c
+++ b/src/backend/commands/vacuumlazy.c
@@ -11,11 +11,24 @@
*
* We are willing to use at most maintenance_work_mem (or perhaps
* autovacuum_work_mem) memory space to keep track of dead tuples. We
- * initially allocate an array of TIDs of that size, with an upper limit that
+ * initially allocate an array of TIDs of 128MB, or an upper limit that
* depends on table size (this limit ensures we don't allocate a huge area
- * uselessly for vacuuming small tables). If the array threatens to overflow,
+ * uselessly for vacuuming small tables). Additional arrays of increasingly
+ * large sizes are allocated as they become necessary.
+ *
+ * The TID array is thus represented as a list of multiple segments of
+ * varying size, beginning with the initial size of up to 128MB, and growing
+ * exponentially until the whole budget of (autovacuum_)maintenance_work_mem
+ * is used up.
+ *
+ * Lookup in that structure happens with a binary search of segments, and then
+ * with a binary search within each segment. Since segment's size grows
+ * exponentially, this retains O(log N) lookup complexity.
+ *
+ * If the multiarray's total size threatens to grow beyond maintenance_work_mem,
* we suspend the heap scan phase and perform a pass of index cleanup and page
- * compaction, then resume the heap scan with an empty TID array.
+ * compaction, then resume the heap scan with an array of logically empty but
+ * already preallocated TID segments to be refilled with more dead tuple TIDs.
*
* If we're processing a table with no indexes, we can just vacuum each page
* as we go; there's no need to save up multiple tuples to minimize the number
@@ -92,6 +105,14 @@
#define LAZY_ALLOC_TUPLES MaxHeapTuplesPerPage
/*
+ * Minimum (starting) size of the dead_tuples array segments. Will allocate
+ * space for 128MB worth of tid pointers in the first segment, further segments
+ * will grow in size exponentially. Don't make it too small or the segment list
+ * will grow bigger than the sweetspot for search efficiency on big vacuums.
+ */
+#define LAZY_INIT_TUPLES Max(MaxHeapTuplesPerPage, (128<<20) / sizeof(ItemPointerData))
+
+/*
* Before we consider skipping a page that's marked as clean in
* visibility map, we must've seen at least this many clean pages.
*/
@@ -103,6 +124,27 @@
*/
#define PREFETCH_SIZE ((BlockNumber) 32)
+typedef struct DeadTuplesSegment
+{
+ ItemPointerData last_dead_tuple; /* Copy of the last dead tuple (unset
+ * until the segment is fully
+ * populated). Keep it first to simplify
+ * binary searches */
+ int num_dead_tuples; /* # of entries in the segment */
+ int max_dead_tuples; /* # of entries allocated in the segment */
+ ItemPointer dt_tids; /* Array of dead tuples */
+} DeadTuplesSegment;
+
+typedef struct DeadTuplesMultiArray
+{
+ int num_entries; /* current # of entries */
+ int max_entries; /* total # of slots that can be allocated in
+ * array */
+ int num_segs; /* number of dead tuple segments allocated */
+ int last_seg; /* last dead tuple segment with data (or 0) */
+ DeadTuplesSegment *dt_segments; /* array of num_segs segments */
+} DeadTuplesMultiArray;
+
typedef struct LVRelStats
{
/* hasindex = true means two-pass strategy; false means one-pass */
@@ -123,14 +165,20 @@ typedef struct LVRelStats
BlockNumber nonempty_pages; /* actually, last nonempty page + 1 */
/* List of TIDs of tuples we intend to delete */
/* NB: this list is ordered by TID address */
- int num_dead_tuples; /* current # of entries */
- int max_dead_tuples; /* # slots allocated in array */
- ItemPointer dead_tuples; /* array of ItemPointerData */
+ DeadTuplesMultiArray dead_tuples;
int num_index_scans;
TransactionId latestRemovedXid;
bool lock_waiter_detected;
} LVRelStats;
+#define GetNumDeadTuplesSegments(lvrelstats) \
+ ((lvrelstats)->dead_tuples.num_segs)
+
+#define GetDeadTuplesSegment(lvrelstats, seg) \
+ (&((lvrelstats)->dead_tuples.dt_segments[seg]))
+
+#define DeadTuplesCurrentSegment(lvrelstats) \
+ GetDeadTuplesSegment(lvrelstats, (lvrelstats)->dead_tuples.last_seg)
/* A few variables that don't seem worth passing around as parameters */
static int elevel = -1;
@@ -155,7 +203,7 @@ static void lazy_cleanup_index(Relation indrel,
IndexBulkDeleteResult *stats,
LVRelStats *vacrelstats);
static int lazy_vacuum_page(Relation onerel, BlockNumber blkno, Buffer buffer,
- int tupindex, LVRelStats *vacrelstats, Buffer *vmbuffer);
+ int tupindex, LVRelStats *vacrelstats, DeadTuplesSegment * seg, Buffer *vmbuffer);
static bool should_attempt_truncation(LVRelStats *vacrelstats);
static void lazy_truncate_heap(Relation onerel, LVRelStats *vacrelstats);
static BlockNumber count_nondeletable_pages(Relation onerel,
@@ -163,8 +211,8 @@ static BlockNumber count_nondeletable_pages(Relation onerel,
static void lazy_space_alloc(LVRelStats *vacrelstats, BlockNumber relblocks);
static void lazy_record_dead_tuple(LVRelStats *vacrelstats,
ItemPointer itemptr);
+static void lazy_clear_dead_tuples(LVRelStats *vacrelstats);
static bool lazy_tid_reaped(ItemPointer itemptr, void *state);
-static int vac_cmp_itemptr(const void *left, const void *right);
static bool heap_page_is_all_visible(Relation rel, Buffer buf,
TransactionId *visibility_cutoff_xid, bool *all_frozen);
@@ -523,7 +571,7 @@ lazy_scan_heap(Relation onerel, int options, LVRelStats *vacrelstats,
/* Report that we're scanning the heap, advertising total # of blocks */
initprog_val[0] = PROGRESS_VACUUM_PHASE_SCAN_HEAP;
initprog_val[1] = nblocks;
- initprog_val[2] = vacrelstats->max_dead_tuples;
+ initprog_val[2] = vacrelstats->dead_tuples.max_entries;
pgstat_progress_update_multi_param(3, initprog_index, initprog_val);
/*
@@ -701,8 +749,8 @@ lazy_scan_heap(Relation onerel, int options, LVRelStats *vacrelstats,
* If we are close to overrunning the available space for dead-tuple
* TIDs, pause and do a cycle of vacuuming before we tackle this page.
*/
- if ((vacrelstats->max_dead_tuples - vacrelstats->num_dead_tuples) < MaxHeapTuplesPerPage &&
- vacrelstats->num_dead_tuples > 0)
+ if ((vacrelstats->dead_tuples.max_entries - vacrelstats->dead_tuples.num_entries) < MaxHeapTuplesPerPage &&
+ vacrelstats->dead_tuples.num_entries > 0)
{
const int hvp_index[] = {
PROGRESS_VACUUM_PHASE,
@@ -753,7 +801,7 @@ lazy_scan_heap(Relation onerel, int options, LVRelStats *vacrelstats,
* not to reset latestRemovedXid since we want that value to be
* valid.
*/
- vacrelstats->num_dead_tuples = 0;
+ lazy_clear_dead_tuples(vacrelstats);
vacrelstats->num_index_scans++;
/* Report that we are once again scanning the heap */
@@ -936,7 +984,7 @@ lazy_scan_heap(Relation onerel, int options, LVRelStats *vacrelstats,
has_dead_tuples = false;
nfrozen = 0;
hastup = false;
- prev_dead_count = vacrelstats->num_dead_tuples;
+ prev_dead_count = vacrelstats->dead_tuples.num_entries;
maxoff = PageGetMaxOffsetNumber(page);
/*
@@ -1157,10 +1205,16 @@ lazy_scan_heap(Relation onerel, int options, LVRelStats *vacrelstats,
* instead of doing a second scan.
*/
if (nindexes == 0 &&
- vacrelstats->num_dead_tuples > 0)
+ vacrelstats->dead_tuples.num_entries > 0)
{
+ /* Should never need more than one segment per page */
+ Assert(vacrelstats->dead_tuples.last_seg == 0);
+
+ /* Should have been initialized */
+ Assert(GetNumDeadTuplesSegments(vacrelstats) > 0);
+
/* Remove tuples from heap */
- lazy_vacuum_page(onerel, blkno, buf, 0, vacrelstats, &vmbuffer);
+ lazy_vacuum_page(onerel, blkno, buf, 0, vacrelstats, DeadTuplesCurrentSegment(vacrelstats), &vmbuffer);
has_dead_tuples = false;
/*
@@ -1168,7 +1222,7 @@ lazy_scan_heap(Relation onerel, int options, LVRelStats *vacrelstats,
* not to reset latestRemovedXid since we want that value to be
* valid.
*/
- vacrelstats->num_dead_tuples = 0;
+ lazy_clear_dead_tuples(vacrelstats);
vacuumed_pages++;
}
@@ -1271,7 +1325,7 @@ lazy_scan_heap(Relation onerel, int options, LVRelStats *vacrelstats,
* page, so remember its free space as-is. (This path will always be
* taken if there are no indexes.)
*/
- if (vacrelstats->num_dead_tuples == prev_dead_count)
+ if (vacrelstats->dead_tuples.num_entries == prev_dead_count)
RecordPageWithFreeSpace(onerel, blkno, freespace);
}
@@ -1302,7 +1356,7 @@ lazy_scan_heap(Relation onerel, int options, LVRelStats *vacrelstats,
/* 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)
+ if (vacrelstats->dead_tuples.num_entries > 0)
{
const int hvp_index[] = {
PROGRESS_VACUUM_PHASE,
@@ -1399,43 +1453,56 @@ lazy_scan_heap(Relation onerel, int options, LVRelStats *vacrelstats,
static void
lazy_vacuum_heap(Relation onerel, LVRelStats *vacrelstats)
{
- int tupindex;
+ int tottuples;
+ int segindex;
int npages;
PGRUsage ru0;
Buffer vmbuffer = InvalidBuffer;
+ if (GetNumDeadTuplesSegments(vacrelstats) == 0)
+ return;
+
pg_rusage_init(&ru0);
npages = 0;
- tupindex = 0;
- while (tupindex < vacrelstats->num_dead_tuples)
+ segindex = 0;
+ tottuples = 0;
+ for (segindex = 0; segindex <= vacrelstats->dead_tuples.last_seg; segindex++)
{
- BlockNumber tblk;
- Buffer buf;
- Page page;
- Size freespace;
+ DeadTuplesSegment *seg = GetDeadTuplesSegment(vacrelstats, segindex);
+ int num_dead_tuples = seg->num_dead_tuples;
+ int tupindex = 0;
- vacuum_delay_point();
-
- tblk = ItemPointerGetBlockNumber(&vacrelstats->dead_tuples[tupindex]);
- buf = ReadBufferExtended(onerel, MAIN_FORKNUM, tblk, RBM_NORMAL,
- vac_strategy);
- if (!ConditionalLockBufferForCleanup(buf))
+ while (tupindex < num_dead_tuples)
{
- ReleaseBuffer(buf);
- ++tupindex;
- continue;
- }
- tupindex = lazy_vacuum_page(onerel, tblk, buf, tupindex, vacrelstats,
- &vmbuffer);
+ BlockNumber tblk;
+ Buffer buf;
+ Page page;
+ Size freespace;
- /* Now that we've compacted the page, record its available space */
- page = BufferGetPage(buf);
- freespace = PageGetHeapFreeSpace(page);
+ vacuum_delay_point();
- UnlockReleaseBuffer(buf);
- RecordPageWithFreeSpace(onerel, tblk, freespace);
- npages++;
+ tblk = ItemPointerGetBlockNumber(&seg->dt_tids[tupindex]);
+ buf = ReadBufferExtended(onerel, MAIN_FORKNUM, tblk, RBM_NORMAL,
+ vac_strategy);
+ if (!ConditionalLockBufferForCleanup(buf))
+ {
+ ReleaseBuffer(buf);
+ ++tupindex;
+ continue;
+ }
+ tupindex = lazy_vacuum_page(onerel, tblk, buf, tupindex, vacrelstats,
+ seg, &vmbuffer);
+
+ /* Now that we've compacted the page, record its available space */
+ page = BufferGetPage(buf);
+ freespace = PageGetHeapFreeSpace(page);
+
+ UnlockReleaseBuffer(buf);
+ RecordPageWithFreeSpace(onerel, tblk, freespace);
+ npages++;
+ }
+ tottuples += tupindex;
}
if (BufferIsValid(vmbuffer))
@@ -1447,7 +1514,7 @@ lazy_vacuum_heap(Relation onerel, LVRelStats *vacrelstats)
ereport(elevel,
(errmsg("\"%s\": removed %d row versions in %d pages",
RelationGetRelationName(onerel),
- tupindex, npages),
+ tottuples, npages),
errdetail_internal("%s", pg_rusage_show(&ru0))));
}
@@ -1457,34 +1524,36 @@ lazy_vacuum_heap(Relation onerel, LVRelStats *vacrelstats)
*
* Caller must hold pin and buffer cleanup lock on the buffer.
*
- * tupindex is the index in vacrelstats->dead_tuples of the first dead
+ * tupindex is the index in seg->dt_tids 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, Buffer *vmbuffer)
+ int tupindex, LVRelStats *vacrelstats, DeadTuplesSegment * seg, Buffer *vmbuffer)
{
Page page = BufferGetPage(buffer);
OffsetNumber unused[MaxOffsetNumber];
int uncnt = 0;
TransactionId visibility_cutoff_xid;
bool all_frozen;
+ ItemPointer dead_tuples = seg->dt_tids;
+ int num_dead_tuples = seg->num_dead_tuples;
pgstat_progress_update_param(PROGRESS_VACUUM_HEAP_BLKS_VACUUMED, blkno);
START_CRIT_SECTION();
- for (; tupindex < vacrelstats->num_dead_tuples; tupindex++)
+ for (; tupindex < num_dead_tuples; tupindex++)
{
BlockNumber tblk;
OffsetNumber toff;
ItemId itemid;
- tblk = ItemPointerGetBlockNumber(&vacrelstats->dead_tuples[tupindex]);
+ tblk = ItemPointerGetBlockNumber(&dead_tuples[tupindex]);
if (tblk != blkno)
break; /* past end of tuples for this block */
- toff = ItemPointerGetOffsetNumber(&vacrelstats->dead_tuples[tupindex]);
+ toff = ItemPointerGetOffsetNumber(&dead_tuples[tupindex]);
itemid = PageGetItemId(page, toff);
ItemIdSetUnused(itemid);
unused[uncnt++] = toff;
@@ -1618,6 +1687,8 @@ lazy_vacuum_index(Relation indrel,
{
IndexVacuumInfo ivinfo;
PGRUsage ru0;
+ DeadTuplesSegment *seg;
+ int n;
pg_rusage_init(&ru0);
@@ -1628,6 +1699,20 @@ lazy_vacuum_index(Relation indrel,
ivinfo.num_heap_tuples = vacrelstats->old_rel_tuples;
ivinfo.strategy = vac_strategy;
+ /* If uninitialized, we have no tuples to delete from the indexes */
+ if (GetNumDeadTuplesSegments(vacrelstats) == 0)
+ {
+ return;
+ }
+
+ /* Finalize all segments by setting their upper bound dead tuple */
+ for (n = 0; n <= vacrelstats->dead_tuples.last_seg; n++)
+ {
+ seg = GetDeadTuplesSegment(vacrelstats, n);
+ if (seg->num_dead_tuples > 0)
+ seg->last_dead_tuple = seg->dt_tids[seg->num_dead_tuples - 1];
+ }
+
/* Do bulk deletion */
*stats = index_bulk_delete(&ivinfo, *stats,
lazy_tid_reaped, (void *) vacrelstats);
@@ -1635,7 +1720,7 @@ lazy_vacuum_index(Relation indrel,
ereport(elevel,
(errmsg("scanned index \"%s\" to remove %d row versions",
RelationGetRelationName(indrel),
- vacrelstats->num_dead_tuples),
+ vacrelstats->dead_tuples.num_entries),
errdetail_internal("%s", pg_rusage_show(&ru0))));
}
@@ -2012,7 +2097,6 @@ lazy_space_alloc(LVRelStats *vacrelstats, BlockNumber relblocks)
{
maxtuples = (vac_work_mem * 1024L) / sizeof(ItemPointerData);
maxtuples = Min(maxtuples, INT_MAX);
- maxtuples = Min(maxtuples, MaxAllocSize / sizeof(ItemPointerData));
/* curious coding here to ensure the multiplication can't overflow */
if ((BlockNumber) (maxtuples / LAZY_ALLOC_TUPLES) > relblocks)
@@ -2026,10 +2110,11 @@ lazy_space_alloc(LVRelStats *vacrelstats, BlockNumber relblocks)
maxtuples = MaxHeapTuplesPerPage;
}
- vacrelstats->num_dead_tuples = 0;
- vacrelstats->max_dead_tuples = (int) maxtuples;
- vacrelstats->dead_tuples = (ItemPointer)
- palloc(maxtuples * sizeof(ItemPointerData));
+ vacrelstats->dead_tuples.num_entries = 0;
+ vacrelstats->dead_tuples.max_entries = (int) maxtuples;
+ vacrelstats->dead_tuples.num_segs = 0;
+ vacrelstats->dead_tuples.last_seg = 0;
+ vacrelstats->dead_tuples.dt_segments = NULL;
}
/*
@@ -2039,46 +2124,165 @@ static void
lazy_record_dead_tuple(LVRelStats *vacrelstats,
ItemPointer itemptr)
{
+ int mintuples;
+
+ /* Initialize multiarray if needed */
+ if (GetNumDeadTuplesSegments(vacrelstats) == 0)
+ {
+ mintuples = Min(LAZY_INIT_TUPLES, vacrelstats->dead_tuples.max_entries);
+
+ vacrelstats->dead_tuples.num_segs = 1;
+ vacrelstats->dead_tuples.dt_segments = (DeadTuplesSegment *)
+ palloc(sizeof(DeadTuplesSegment));
+ vacrelstats->dead_tuples.dt_segments[0].dt_tids = (ItemPointer)
+ palloc(mintuples * sizeof(ItemPointerData));
+ vacrelstats->dead_tuples.dt_segments[0].max_dead_tuples = mintuples;
+ vacrelstats->dead_tuples.dt_segments[0].num_dead_tuples = 0;
+ }
+
/*
* The array shouldn't overflow under normal behavior, but perhaps it
* could if we are given a really small maintenance_work_mem. In that
* case, just forget the last few tuples (we'll get 'em next time).
*/
- if (vacrelstats->num_dead_tuples < vacrelstats->max_dead_tuples)
+ if (vacrelstats->dead_tuples.num_entries < vacrelstats->dead_tuples.max_entries)
{
- vacrelstats->dead_tuples[vacrelstats->num_dead_tuples] = *itemptr;
- vacrelstats->num_dead_tuples++;
+ DeadTuplesSegment *seg = DeadTuplesCurrentSegment(vacrelstats);
+
+ if (seg->num_dead_tuples >= seg->max_dead_tuples)
+ {
+ DeadTuplesMultiArray *dt = &vacrelstats->dead_tuples;
+
+ /*
+ * The segment is overflowing, so we must allocate a new segment.
+ * We could have a preallocated segment descriptor already, in
+ * which case we just reinitialize it, or we may need to repalloc
+ * the vacrelstats->dead_tuples array. In that case, seg will no
+ * longer be valid, so we must be careful about that.
+ */
+ Assert(seg->num_dead_tuples == seg->max_dead_tuples);
+ if (dt->last_seg + 1 >= dt->num_segs)
+ {
+ int new_num_segs = dt->num_segs * 2;
+ int new_segs_size = new_num_segs * sizeof(DeadTuplesSegment);
+
+ dt->dt_segments = (DeadTuplesSegment *) repalloc((void *) dt->dt_segments, new_segs_size);
+ while (dt->num_segs < new_num_segs)
+ {
+ /* Initialize as "unallocated" */
+ DeadTuplesSegment *nseg = &(dt->dt_segments[dt->num_segs]);
+
+ nseg->num_dead_tuples = 0;
+ nseg->max_dead_tuples = 0;
+ nseg->dt_tids = NULL;
+ dt->num_segs++;
+ }
+ }
+ dt->last_seg++;
+ seg = DeadTuplesCurrentSegment(vacrelstats);
+ if (seg->dt_tids == NULL)
+ {
+ /*
+ * If unallocated, allocate up to twice the amount of the
+ * previous segment
+ */
+ DeadTuplesSegment *pseg = seg - 1;
+ int numtuples = dt->max_entries - dt->num_entries;
+
+ numtuples = Max(numtuples, MaxHeapTuplesPerPage);
+ numtuples = Min(numtuples, INT_MAX / 2);
+ numtuples = Min(numtuples, 2 * pseg->max_dead_tuples);
+ numtuples = Min(numtuples, MaxAllocSize / sizeof(ItemPointerData));
+ seg->dt_tids = (ItemPointer) palloc(sizeof(ItemPointerData) * numtuples);
+ seg->max_dead_tuples = numtuples;
+ }
+ seg->num_dead_tuples = 0;
+ }
+ seg->dt_tids[seg->num_dead_tuples] = *itemptr;
+ vacrelstats->dead_tuples.num_entries++;
+ seg->num_dead_tuples++;
pgstat_progress_update_param(PROGRESS_VACUUM_NUM_DEAD_TUPLES,
- vacrelstats->num_dead_tuples);
+ vacrelstats->dead_tuples.num_entries);
}
}
/*
- * lazy_tid_reaped() -- is a particular tid deletable?
+ * lazy_clear_dead_tuples - reset all dead tuples segments
+ */
+static void
+lazy_clear_dead_tuples(LVRelStats *vacrelstats)
+{
+ int nseg;
+
+ for (nseg = 0; nseg < GetNumDeadTuplesSegments(vacrelstats); nseg++)
+ GetDeadTuplesSegment(vacrelstats, nseg)->num_dead_tuples = 0;
+ vacrelstats->dead_tuples.last_seg = 0;
+ vacrelstats->dead_tuples.num_entries = 0;
+}
+
+/*
+ * vac_itemptr_binsrch() -- search a sorted array of item pointers
*
- * This has the right signature to be an IndexBulkDeleteCallback.
+ * Returns the offset of the first item pointer greater than or
+ * equal to refvalue, or arr_elems if there is no such item pointer
*
- * Assumes dead_tuples array is in sorted order.
+ * All item pointers in the array are assumed to be valid
+ *
+ * Within, vac_cmp_itemptr has been inlined to remove redundant
+ * validity checking (the dead tuples array contains only valid
+ * item pointers) and ItemPointerGetX invocations (the refvalue
+ * never changes). This makes the code easier to optimize for
+ * the compiler, and should improve performance
*/
-static bool
-lazy_tid_reaped(ItemPointer itemptr, void *state)
+static inline size_t vac_itemptr_binsrch(ItemPointer refvalue, void *arr,
+ size_t arr_elems, size_t arr_stride)
{
- LVRelStats *vacrelstats = (LVRelStats *) state;
- ItemPointer res;
+ BlockNumber refblk, blk;
+ OffsetNumber refoff, off;
+ ItemPointer value;
+ size_t left, right, mid;
- res = (ItemPointer) bsearch((void *) itemptr,
- (void *) vacrelstats->dead_tuples,
- vacrelstats->num_dead_tuples,
- sizeof(ItemPointerData),
- vac_cmp_itemptr);
+ if (arr_elems == 0 || !ItemPointerIsValid(refvalue))
+ return arr_elems;
+
+ refblk = ItemPointerGetBlockNumberNoCheck(refvalue);
+ refoff = ItemPointerGetOffsetNumberNoCheck(refvalue);
+
+ left = 0;
+ right = arr_elems - 1;
+ while (right > left)
+ {
+ mid = left + ((right - left) / 2);
+ value = (ItemPointer)((char*) arr + mid * arr_stride);
- return (res != NULL);
+ blk = ItemPointerGetBlockNumberNoCheck(value);
+ if (refblk < blk)
+ {
+ right = mid;
+ }
+ else if (refblk == blk)
+ {
+ off = ItemPointerGetOffsetNumberNoCheck(value);
+ if (refoff < off)
+ right = mid;
+ else if (refoff == off)
+ return mid;
+ else
+ left = mid + 1;
+ }
+ else
+ {
+ left = mid + 1;
+ }
+ }
+
+ return left;
}
/*
- * Comparator routines for use with qsort() and bsearch().
+ * Comparator routine for use within lazy_tid_reaped
*/
-static int
+static inline int
vac_cmp_itemptr(const void *left, const void *right)
{
BlockNumber lblk,
@@ -2106,6 +2310,56 @@ vac_cmp_itemptr(const void *left, const void *right)
}
/*
+ * lazy_tid_reaped() -- is a particular tid deletable?
+ *
+ * This has the right signature to be an IndexBulkDeleteCallback.
+ *
+ * Assumes the dead_tuples multiarray is in sorted order, both
+ * the segment list and each segment itself, and that all segments'
+ * last_dead_tuple fields up to date
+ */
+static bool
+lazy_tid_reaped(ItemPointer itemptr, void *state)
+{
+ LVRelStats *vacrelstats = (LVRelStats *) state;
+ DeadTuplesSegment *seg;
+ size_t iseg, itup;
+
+ if (GetNumDeadTuplesSegments(vacrelstats) == 0)
+ return false;
+
+ /* Quickly rule out by lower bound (should happen a lot) */
+ seg = vacrelstats->dead_tuples.dt_segments;
+ if (0 > vac_cmp_itemptr((void *) itemptr, (void *) seg->dt_tids))
+ return false;
+
+ /* Search for the segment likely to contain the item pointer */
+ iseg = vac_itemptr_binsrch(
+ (void *) itemptr,
+ (void *) &(seg->last_dead_tuple),
+ vacrelstats->dead_tuples.last_seg + 1,
+ sizeof(DeadTuplesSegment));
+
+ if (iseg > vacrelstats->dead_tuples.last_seg)
+ return false;
+
+ seg = GetDeadTuplesSegment(vacrelstats, iseg);
+ if (seg->num_dead_tuples == 0)
+ return false;
+
+ /* Search within the segment for the right item pointer */
+ itup = vac_itemptr_binsrch((void *) itemptr,
+ (void *) seg->dt_tids,
+ seg->num_dead_tuples,
+ sizeof(ItemPointerData));
+ if (itup >= seg->num_dead_tuples)
+ return false;
+ else
+ return 0 == vac_cmp_itemptr((void *) itemptr,
+ (void *) (&seg->dt_tids[itup]));
+}
+
+/*
* Check if every tuple in the given page is visible to all current and future
* transactions. Also return the visibility_cutoff_xid which is the highest
* xmin amongst the visible tuples. Set *all_frozen to true if every tuple
diff --git a/src/test/regress/expected/vacuum.out b/src/test/regress/expected/vacuum.out
index c440c7e..29a5242 100644
--- a/src/test/regress/expected/vacuum.out
+++ b/src/test/regress/expected/vacuum.out
@@ -1,7 +1,17 @@
--
-- VACUUM
--
-CREATE TABLE vactst (i INT);
+CREATE FUNCTION wait_barrier() RETURNS void LANGUAGE plpgsql AS $$
+DECLARE vxids text[];
+BEGIN
+ -- wait for all transactions to commit to make deleted tuples vacuumable
+ SELECT array_agg(DISTINCT virtualtransaction) FROM pg_locks WHERE pid <> pg_backend_pid() INTO vxids;
+ WHILE (SELECT count(virtualtransaction) FROM pg_locks WHERE pid <> pg_backend_pid() AND virtualtransaction = ANY(vxids)) > 0 LOOP
+ PERFORM pg_sleep(0.1);
+ END LOOP;
+END
+$$;
+CREATE TABLE vactst (i INT) WITH (autovacuum_enabled = false);
INSERT INTO vactst VALUES (1);
INSERT INTO vactst SELECT * FROM vactst;
INSERT INTO vactst SELECT * FROM vactst;
@@ -22,6 +32,12 @@ SELECT count(*) FROM vactst;
(1 row)
DELETE FROM vactst WHERE i != 0;
+SELECT wait_barrier();
+ wait_barrier
+--------------
+
+(1 row)
+
SELECT * FROM vactst;
i
---
@@ -49,8 +65,20 @@ SELECT count(*) FROM vactst;
(1 row)
DELETE FROM vactst WHERE i != 0;
+SELECT wait_barrier();
+ wait_barrier
+--------------
+
+(1 row)
+
VACUUM (FULL) vactst;
DELETE FROM vactst;
+SELECT wait_barrier();
+ wait_barrier
+--------------
+
+(1 row)
+
SELECT * FROM vactst;
i
---
@@ -112,6 +140,48 @@ ANALYZE vactst, does_not_exist, vacparted;
ERROR: relation "does_not_exist" does not exist
ANALYZE vactst (i), vacparted (does_not_exist);
ERROR: column "does_not_exist" of relation "vacparted" does not exist
+-- large mwm vacuum runs
+CREATE UNLOGGED TABLE vactst2 (i INT) WITH (autovacuum_enabled = false);
+INSERT INTO vactst2 SELECT * from generate_series(1,300000);
+CREATE INDEX ix_vactst2 ON vactst2 (i);
+DELETE FROM vactst2 WHERE i % 4 != 1;
+SET maintenance_work_mem = 1024;
+SELECT wait_barrier();
+ wait_barrier
+--------------
+
+(1 row)
+
+VACUUM vactst2;
+SET maintenance_work_mem TO DEFAULT;
+DROP INDEX ix_vactst2;
+TRUNCATE TABLE vactst2;
+INSERT INTO vactst2 SELECT * from generate_series(1,40);
+CREATE INDEX ix_vactst2 ON vactst2 (i);
+DELETE FROM vactst2;
+SELECT wait_barrier();
+ wait_barrier
+--------------
+
+(1 row)
+
+VACUUM vactst2;
+EXPLAIN (ANALYZE, BUFFERS, COSTS off, TIMING off, SUMMARY off) SELECT * FROM vactst2;
+ QUERY PLAN
+---------------------------------------------
+ Seq Scan on vactst2 (actual rows=0 loops=1)
+(1 row)
+
+SELECT count(*) FROM vactst2;
+ count
+-------
+ 0
+(1 row)
+
+DROP INDEX ix_vactst2;
+TRUNCATE TABLE vactst2;
DROP TABLE vaccluster;
+DROP TABLE vactst2;
DROP TABLE vactst;
DROP TABLE vacparted;
+DROP FUNCTION wait_barrier();
diff --git a/src/test/regress/sql/vacuum.sql b/src/test/regress/sql/vacuum.sql
index 92eaca2..4470967 100644
--- a/src/test/regress/sql/vacuum.sql
+++ b/src/test/regress/sql/vacuum.sql
@@ -2,7 +2,18 @@
-- VACUUM
--
-CREATE TABLE vactst (i INT);
+CREATE FUNCTION wait_barrier() RETURNS void LANGUAGE plpgsql AS $$
+DECLARE vxids text[];
+BEGIN
+ -- wait for all transactions to commit to make deleted tuples vacuumable
+ SELECT array_agg(DISTINCT virtualtransaction) FROM pg_locks WHERE pid <> pg_backend_pid() INTO vxids;
+ WHILE (SELECT count(virtualtransaction) FROM pg_locks WHERE pid <> pg_backend_pid() AND virtualtransaction = ANY(vxids)) > 0 LOOP
+ PERFORM pg_sleep(0.1);
+ END LOOP;
+END
+$$;
+
+CREATE TABLE vactst (i INT) WITH (autovacuum_enabled = false);
INSERT INTO vactst VALUES (1);
INSERT INTO vactst SELECT * FROM vactst;
INSERT INTO vactst SELECT * FROM vactst;
@@ -18,6 +29,7 @@ INSERT INTO vactst SELECT * FROM vactst;
INSERT INTO vactst VALUES (0);
SELECT count(*) FROM vactst;
DELETE FROM vactst WHERE i != 0;
+SELECT wait_barrier();
SELECT * FROM vactst;
VACUUM FULL vactst;
UPDATE vactst SET i = i + 1;
@@ -35,8 +47,10 @@ INSERT INTO vactst SELECT * FROM vactst;
INSERT INTO vactst VALUES (0);
SELECT count(*) FROM vactst;
DELETE FROM vactst WHERE i != 0;
+SELECT wait_barrier();
VACUUM (FULL) vactst;
DELETE FROM vactst;
+SELECT wait_barrier();
SELECT * FROM vactst;
VACUUM (FULL, FREEZE) vactst;
@@ -89,6 +103,30 @@ ANALYZE vacparted (b), vactst;
ANALYZE vactst, does_not_exist, vacparted;
ANALYZE vactst (i), vacparted (does_not_exist);
+-- large mwm vacuum runs
+CREATE UNLOGGED TABLE vactst2 (i INT) WITH (autovacuum_enabled = false);
+INSERT INTO vactst2 SELECT * from generate_series(1,300000);
+CREATE INDEX ix_vactst2 ON vactst2 (i);
+DELETE FROM vactst2 WHERE i % 4 != 1;
+SET maintenance_work_mem = 1024;
+SELECT wait_barrier();
+VACUUM vactst2;
+SET maintenance_work_mem TO DEFAULT;
+DROP INDEX ix_vactst2;
+TRUNCATE TABLE vactst2;
+
+INSERT INTO vactst2 SELECT * from generate_series(1,40);
+CREATE INDEX ix_vactst2 ON vactst2 (i);
+DELETE FROM vactst2;
+SELECT wait_barrier();
+VACUUM vactst2;
+EXPLAIN (ANALYZE, BUFFERS, COSTS off, TIMING off, SUMMARY off) SELECT * FROM vactst2;
+SELECT count(*) FROM vactst2;
+DROP INDEX ix_vactst2;
+TRUNCATE TABLE vactst2;
+
DROP TABLE vaccluster;
+DROP TABLE vactst2;
DROP TABLE vactst;
DROP TABLE vacparted;
+DROP FUNCTION wait_barrier();
--
1.8.4.5
