On Thu, Sep 21, 2023 at 3:53 PM Robert Haas <robertmh...@gmail.com> wrote: > > On Wed, Sep 13, 2023 at 1:29 PM Andres Freund <and...@anarazel.de> wrote: > > > > /* > > > * The criteria for counting a tuple as live in this block > > > need to > > > @@ -1682,7 +1664,7 @@ retry: > > > * (Cases where we bypass index vacuuming will violate this > > > optimistic > > > * assumption, but the overall impact of that should be > > > negligible.) > > > */ > > > - switch (res) > > > + switch ((HTSV_Result) presult.htsv[offnum]) > > > { > > > case HEAPTUPLE_LIVE: > > > > I think we should assert that we have a valid HTSV_Result here, i.e. not > > -1. You could wrap the cast and Assert into an inline funciton as well. > > This isn't a bad idea, although I don't find it completely necessary either.
Attached v5 does this. Even though a value of -1 would hit the default switch case and error out, I can see the argument for this validation -- as all other places switching on an HTSV_Result are doing so on a value which was always an HTSV_Result. Once I started writing the function comment, however, I felt a bit awkward. In order to make the function available to both pruneheap.c and vacuumlazy.c, I had to put it in a header file. Writing a function, available to anyone including heapam.h, which takes an int and returns an HTSV_Result feels a bit odd. Do we want it to be common practice to use an int value outside the valid enum range to store "status not yet computed" for HTSV_Results? Anyway, on a tactical note, I added the inline function to heapam.h below the PruneResult definition since it is fairly tightly coupled to the htsv array in PruneResult. All of the function prototypes are under a comment that says "function prototypes for heap access method" -- which didn't feel like an accurate description of this function. I wonder if it makes sense to have pruneheap.c include vacuum.h and move pruning specific stuff like this helper and PruneResult over there? I can't remember why I didn't do this before, but maybe there is a reason not to? I also wasn't sure if I needed to forward declare the inline function or not. Oh, and, one more note. I've dropped the former patch 0001 which changed the function comment about off_loc above heap_page_prune(). I have plans to write a separate patch adding an error context callback for HOT pruning with the offset number and would include such a change in that patch. - Melanie
From aa1ec4fa1a689f9c6328d02fd13af1a42ecb0396 Mon Sep 17 00:00:00 2001 From: Melanie Plageman <melanieplage...@gmail.com> Date: Wed, 6 Sep 2023 16:54:41 -0400 Subject: [PATCH v5 2/2] Reuse heap_page_prune() tuple visibility statuses heap_page_prune() obtains the HTSV_Result (tuple visibility status) returned from HeapTupleSatisfiesVacuum() for every tuple on the page and stores them in an array. By making this array available to heap_page_prune()'s caller lazy_scan_prune(), we can avoid an additional call to HeapTupleSatisfiesVacuum() when freezing the tuples and recording LP_DEAD items for vacuum. This saves resources and eliminates the possibility that vacuuming corrupted data results in a hang due to endless retry looping. This replaces the retry mechanism introduced in 8523492d4 to handle cases in which a tuple's inserting transaction aborted between the visibility check in heap_page_prune() and lazy_scan_prune()'s call to HeapTupleSatisfiesVacuum() -- rendering it dead but without a dead line pointer. We can instead reuse the tuple's original visibility status, circumventing any disagreements. Discussion: https://postgr.es/m/CAAKRu_br124qsGJieuYA0nGjywEukhK1dKBfRdby_4yY3E9SXA%40mail.gmail.com --- src/backend/access/heap/pruneheap.c | 36 +++++++++++------------- src/backend/access/heap/vacuumlazy.c | 42 ++++++++-------------------- src/include/access/heapam.h | 25 +++++++++++++++++ 3 files changed, 54 insertions(+), 49 deletions(-) diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c index d5892a2db4..c5f1abd95a 100644 --- a/src/backend/access/heap/pruneheap.c +++ b/src/backend/access/heap/pruneheap.c @@ -53,16 +53,6 @@ typedef struct * 1. Otherwise every access would need to subtract 1. */ bool marked[MaxHeapTuplesPerPage + 1]; - - /* - * Tuple visibility is only computed once for each tuple, for correctness - * and efficiency reasons; see comment in heap_page_prune() for details. - * This is of type int8[], instead of HTSV_Result[], so we can use -1 to - * indicate no visibility has been computed, e.g. for LP_DEAD items. - * - * Same indexing as ->marked. - */ - int8 htsv[MaxHeapTuplesPerPage + 1]; } PruneState; /* Local functions */ @@ -71,6 +61,7 @@ static HTSV_Result heap_prune_satisfies_vacuum(PruneState *prstate, Buffer buffer); static int heap_prune_chain(Buffer buffer, OffsetNumber rootoffnum, + int8 *htsv, PruneState *prstate); static void heap_prune_record_prunable(PruneState *prstate, TransactionId xid); static void heap_prune_record_redirect(PruneState *prstate, @@ -240,6 +231,10 @@ heap_page_prune(Relation relation, Buffer buffer, prstate.nredirected = prstate.ndead = prstate.nunused = 0; memset(prstate.marked, 0, sizeof(prstate.marked)); + /* + * presult->htsv is not initialized here because all ntuple spots in the + * array will be set either to a valid HTSV_Result value or -1. + */ presult->ndeleted = 0; presult->nnewlpdead = 0; @@ -276,7 +271,7 @@ heap_page_prune(Relation relation, Buffer buffer, /* Nothing to do if slot doesn't contain a tuple */ if (!ItemIdIsNormal(itemid)) { - prstate.htsv[offnum] = -1; + presult->htsv[offnum] = -1; continue; } @@ -292,8 +287,8 @@ heap_page_prune(Relation relation, Buffer buffer, if (off_loc) *off_loc = offnum; - prstate.htsv[offnum] = heap_prune_satisfies_vacuum(&prstate, &tup, - buffer); + presult->htsv[offnum] = heap_prune_satisfies_vacuum(&prstate, &tup, + buffer); } /* Scan the page */ @@ -317,7 +312,8 @@ heap_page_prune(Relation relation, Buffer buffer, continue; /* Process this item or chain of items */ - presult->ndeleted += heap_prune_chain(buffer, offnum, &prstate); + presult->ndeleted += heap_prune_chain(buffer, offnum, + presult->htsv, &prstate); } /* Clear the offset information once we have processed the given page. */ @@ -446,6 +442,8 @@ heap_prune_satisfies_vacuum(PruneState *prstate, HeapTuple tup, Buffer buffer) /* * Prune specified line pointer or a HOT chain originating at line pointer. * + * Tuple visibility information is provided in htsv. + * * If the item is an index-referenced tuple (i.e. not a heap-only tuple), * the HOT chain is pruned by removing all DEAD tuples at the start of the HOT * chain. We also prune any RECENTLY_DEAD tuples preceding a DEAD tuple. @@ -473,7 +471,8 @@ heap_prune_satisfies_vacuum(PruneState *prstate, HeapTuple tup, Buffer buffer) * Returns the number of tuples (to be) deleted from the page. */ static int -heap_prune_chain(Buffer buffer, OffsetNumber rootoffnum, PruneState *prstate) +heap_prune_chain(Buffer buffer, OffsetNumber rootoffnum, + int8 *htsv, PruneState *prstate) { int ndeleted = 0; Page dp = (Page) BufferGetPage(buffer); @@ -494,7 +493,7 @@ heap_prune_chain(Buffer buffer, OffsetNumber rootoffnum, PruneState *prstate) */ if (ItemIdIsNormal(rootlp)) { - Assert(prstate->htsv[rootoffnum] != -1); + Assert(htsv[rootoffnum] != -1); htup = (HeapTupleHeader) PageGetItem(dp, rootlp); if (HeapTupleHeaderIsHeapOnly(htup)) @@ -517,7 +516,7 @@ heap_prune_chain(Buffer buffer, OffsetNumber rootoffnum, PruneState *prstate) * either here or while following a chain below. Whichever path * gets there first will mark the tuple unused. */ - if (prstate->htsv[rootoffnum] == HEAPTUPLE_DEAD && + if (htsv[rootoffnum] == HEAPTUPLE_DEAD && !HeapTupleHeaderIsHotUpdated(htup)) { heap_prune_record_unused(prstate, rootoffnum); @@ -585,7 +584,6 @@ heap_prune_chain(Buffer buffer, OffsetNumber rootoffnum, PruneState *prstate) break; Assert(ItemIdIsNormal(lp)); - Assert(prstate->htsv[offnum] != -1); htup = (HeapTupleHeader) PageGetItem(dp, lp); /* @@ -605,7 +603,7 @@ heap_prune_chain(Buffer buffer, OffsetNumber rootoffnum, PruneState *prstate) */ tupdead = recent_dead = false; - switch ((HTSV_Result) prstate->htsv[offnum]) + switch (htsv_get_valid_status(htsv[offnum])) { case HEAPTUPLE_DEAD: tupdead = true; diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c index fa77ef7f4a..b8268a5cef 100644 --- a/src/backend/access/heap/vacuumlazy.c +++ b/src/backend/access/heap/vacuumlazy.c @@ -1524,12 +1524,12 @@ lazy_scan_new_or_empty(LVRelState *vacrel, Buffer buf, BlockNumber blkno, * of complexity just so we could deal with tuples that were DEAD to VACUUM, * but nevertheless were left with storage after pruning. * - * The approach we take now is to restart pruning when the race condition is - * detected. This allows heap_page_prune() to prune the tuples inserted by - * the now-aborted transaction. This is a little crude, but it guarantees - * that any items that make it into the dead_items array are simple LP_DEAD - * line pointers, and that every remaining item with tuple storage is - * considered as a candidate for freezing. + * As of Postgres 17, we circumvent this problem altogether by reusing the + * result of heap_page_prune()'s visibility check. Without the second call to + * HeapTupleSatisfiesVacuum(), there is no new HTSV_Result and there can be no + * disagreement. The tuple's TID won't be added to the array of dead tuple TIDs + * for vacuum, thus vacuum will never be tasked with reaping a tuple with + * storage. */ static void lazy_scan_prune(LVRelState *vacrel, @@ -1542,8 +1542,6 @@ lazy_scan_prune(LVRelState *vacrel, OffsetNumber offnum, maxoff; ItemId itemid; - HeapTupleData tuple; - HTSV_Result res; PruneResult presult; int tuples_frozen, lpdead_items, @@ -1563,8 +1561,6 @@ lazy_scan_prune(LVRelState *vacrel, */ maxoff = PageGetMaxOffsetNumber(page); -retry: - /* Initialize (or reset) page-level state */ pagefrz.freeze_required = false; pagefrz.FreezePageRelfrozenXid = vacrel->NewRelfrozenXid; @@ -1600,6 +1596,7 @@ retry: offnum <= maxoff; offnum = OffsetNumberNext(offnum)) { + HeapTupleHeader htup; bool totally_frozen; /* @@ -1642,22 +1639,7 @@ retry: Assert(ItemIdIsNormal(itemid)); - ItemPointerSet(&(tuple.t_self), blkno, offnum); - tuple.t_data = (HeapTupleHeader) PageGetItem(page, itemid); - tuple.t_len = ItemIdGetLength(itemid); - tuple.t_tableOid = RelationGetRelid(rel); - - /* - * DEAD tuples are almost always pruned into LP_DEAD line pointers by - * heap_page_prune(), but it's possible that the tuple state changed - * since heap_page_prune() looked. Handle that here by restarting. - * (See comments at the top of function for a full explanation.) - */ - res = HeapTupleSatisfiesVacuum(&tuple, vacrel->cutoffs.OldestXmin, - buf); - - if (unlikely(res == HEAPTUPLE_DEAD)) - goto retry; + htup = (HeapTupleHeader) PageGetItem(page, itemid); /* * The criteria for counting a tuple as live in this block need to @@ -1678,7 +1660,7 @@ retry: * (Cases where we bypass index vacuuming will violate this optimistic * assumption, but the overall impact of that should be negligible.) */ - switch (res) + switch (htsv_get_valid_status(presult.htsv[offnum])) { case HEAPTUPLE_LIVE: @@ -1700,7 +1682,7 @@ retry: { TransactionId xmin; - if (!HeapTupleHeaderXminCommitted(tuple.t_data)) + if (!HeapTupleHeaderXminCommitted(htup)) { prunestate->all_visible = false; break; @@ -1710,7 +1692,7 @@ retry: * The inserter definitely committed. But is it old enough * that everyone sees it as committed? */ - xmin = HeapTupleHeaderGetXmin(tuple.t_data); + xmin = HeapTupleHeaderGetXmin(htup); if (!TransactionIdPrecedes(xmin, vacrel->cutoffs.OldestXmin)) { @@ -1764,7 +1746,7 @@ retry: prunestate->hastup = true; /* page makes rel truncation unsafe */ /* Tuple with storage -- consider need to freeze */ - if (heap_prepare_freeze_tuple(tuple.t_data, &vacrel->cutoffs, &pagefrz, + if (heap_prepare_freeze_tuple(htup, &vacrel->cutoffs, &pagefrz, &frozen[tuples_frozen], &totally_frozen)) { /* Save prepared freeze plan for later */ diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h index 2d3f149e4f..62fac1d5d2 100644 --- a/src/include/access/heapam.h +++ b/src/include/access/heapam.h @@ -198,8 +198,33 @@ typedef struct PruneResult { int ndeleted; /* Number of tuples deleted from the page */ int nnewlpdead; /* Number of newly LP_DEAD items */ + + /* + * Tuple visibility is only computed once for each tuple, for correctness + * and efficiency reasons; see comment in heap_page_prune() for details. + * This is of type int8[], instead of HTSV_Result[], so we can use -1 to + * indicate no visibility has been computed, e.g. for LP_DEAD items. + * + * This needs to be MaxHeapTuplesPerPage + 1 long as FirstOffsetNumber is + * 1. Otherwise every access would need to subtract 1. + */ + int8 htsv[MaxHeapTuplesPerPage + 1]; } PruneResult; +/* + * Pruning calculates tuple visibility once and saves the results in an array + * of int8. See PruneResult.htsv for details. This helper function is meant to + * guard against examining visibility status array members which have not yet + * been computed. + */ +static inline HTSV_Result +htsv_get_valid_status(int status) +{ + Assert(status >= HEAPTUPLE_DEAD && + status <= HEAPTUPLE_DELETE_IN_PROGRESS); + return (HTSV_Result) status; +} + /* ---------------- * function prototypes for heap access method * -- 2.37.2
From 430ff3458f0ac523f1c623d9bc2669981cc0e536 Mon Sep 17 00:00:00 2001 From: Melanie Plageman <melanieplage...@gmail.com> Date: Wed, 6 Sep 2023 14:57:20 -0400 Subject: [PATCH v5 1/2] Move heap_page_prune output parameters into struct Add PruneResult, a structure containing the output parameters and result of heap_page_prune(). Reorganizing the results of heap_page_prune() into a struct simplifies the function signature and provides a location for future commits to store additional output parameters. Discussion: https://postgr.es/m/CAAKRu_br124qsGJieuYA0nGjywEukhK1dKBfRdby_4yY3E9SXA%40mail.gmail.com --- src/backend/access/heap/pruneheap.c | 33 +++++++++++++--------------- src/backend/access/heap/vacuumlazy.c | 17 +++++--------- src/include/access/heapam.h | 13 +++++++++-- src/tools/pgindent/typedefs.list | 1 + 4 files changed, 33 insertions(+), 31 deletions(-) diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c index 18193efa23..d5892a2db4 100644 --- a/src/backend/access/heap/pruneheap.c +++ b/src/backend/access/heap/pruneheap.c @@ -155,15 +155,13 @@ heap_page_prune_opt(Relation relation, Buffer buffer) */ if (PageIsFull(page) || PageGetHeapFreeSpace(page) < minfree) { - int ndeleted, - nnewlpdead; + PruneResult presult; - ndeleted = heap_page_prune(relation, buffer, vistest, - &nnewlpdead, NULL); + heap_page_prune(relation, buffer, vistest, &presult, NULL); /* * Report the number of tuples reclaimed to pgstats. This is - * ndeleted minus the number of newly-LP_DEAD-set items. + * presult.ndeleted minus the number of newly-LP_DEAD-set items. * * We derive the number of dead tuples like this to avoid totally * forgetting about items that were set to LP_DEAD, since they @@ -175,9 +173,9 @@ heap_page_prune_opt(Relation relation, Buffer buffer) * tracks ndeleted, since it will set the same LP_DEAD items to * LP_UNUSED separately. */ - if (ndeleted > nnewlpdead) + if (presult.ndeleted > presult.nnewlpdead) pgstat_update_heap_dead_tuples(relation, - ndeleted - nnewlpdead); + presult.ndeleted - presult.nnewlpdead); } /* And release buffer lock */ @@ -204,21 +202,19 @@ heap_page_prune_opt(Relation relation, Buffer buffer) * (see heap_prune_satisfies_vacuum and * HeapTupleSatisfiesVacuum). * - * Sets *nnewlpdead for caller, indicating the number of items that were - * newly set LP_DEAD during prune operation. - * * off_loc is the offset location required by the caller to use in error * callback. * - * Returns the number of tuples deleted from the page during this call. + * presult contains output parameters needed by callers such as the number of + * tuples removed and the number of line pointers newly marked LP_DEAD. + * heap_page_prune() is responsible for initializing it. */ -int +void heap_page_prune(Relation relation, Buffer buffer, GlobalVisState *vistest, - int *nnewlpdead, + PruneResult *presult, OffsetNumber *off_loc) { - int ndeleted = 0; Page page = BufferGetPage(buffer); BlockNumber blockno = BufferGetBlockNumber(buffer); OffsetNumber offnum, @@ -244,6 +240,9 @@ heap_page_prune(Relation relation, Buffer buffer, prstate.nredirected = prstate.ndead = prstate.nunused = 0; memset(prstate.marked, 0, sizeof(prstate.marked)); + presult->ndeleted = 0; + presult->nnewlpdead = 0; + maxoff = PageGetMaxOffsetNumber(page); tup.t_tableOid = RelationGetRelid(prstate.rel); @@ -318,7 +317,7 @@ heap_page_prune(Relation relation, Buffer buffer, continue; /* Process this item or chain of items */ - ndeleted += heap_prune_chain(buffer, offnum, &prstate); + presult->ndeleted += heap_prune_chain(buffer, offnum, &prstate); } /* Clear the offset information once we have processed the given page. */ @@ -419,9 +418,7 @@ heap_page_prune(Relation relation, Buffer buffer, END_CRIT_SECTION(); /* Record number of newly-set-LP_DEAD items for caller */ - *nnewlpdead = prstate.ndead; - - return ndeleted; + presult->nnewlpdead = prstate.ndead; } diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c index b3c24d68f0..fa77ef7f4a 100644 --- a/src/backend/access/heap/vacuumlazy.c +++ b/src/backend/access/heap/vacuumlazy.c @@ -1544,12 +1544,11 @@ lazy_scan_prune(LVRelState *vacrel, ItemId itemid; HeapTupleData tuple; HTSV_Result res; - int tuples_deleted, - tuples_frozen, + PruneResult presult; + int tuples_frozen, lpdead_items, live_tuples, recently_dead_tuples; - int nnewlpdead; HeapPageFreeze pagefrz; int64 fpi_before = pgWalUsage.wal_fpi; OffsetNumber deadoffsets[MaxHeapTuplesPerPage]; @@ -1572,7 +1571,6 @@ retry: pagefrz.FreezePageRelminMxid = vacrel->NewRelminMxid; pagefrz.NoFreezePageRelfrozenXid = vacrel->NewRelfrozenXid; pagefrz.NoFreezePageRelminMxid = vacrel->NewRelminMxid; - tuples_deleted = 0; tuples_frozen = 0; lpdead_items = 0; live_tuples = 0; @@ -1581,15 +1579,12 @@ retry: /* * Prune all HOT-update chains in this page. * - * We count tuples removed by the pruning step as tuples_deleted. Its - * final value can be thought of as the number of tuples that have been - * deleted from the table. It should not be confused with lpdead_items; + * We count the number of tuples removed from the page by the pruning step + * in presult.ndeleted. It should not be confused with lpdead_items; * lpdead_items's final value can be thought of as the number of tuples * that were deleted from indexes. */ - tuples_deleted = heap_page_prune(rel, buf, vacrel->vistest, - &nnewlpdead, - &vacrel->offnum); + heap_page_prune(rel, buf, vacrel->vistest, &presult, &vacrel->offnum); /* * Now scan the page to collect LP_DEAD items and check for tuples @@ -1929,7 +1924,7 @@ retry: } /* Finally, add page-local counts to whole-VACUUM counts */ - vacrel->tuples_deleted += tuples_deleted; + vacrel->tuples_deleted += presult.ndeleted; vacrel->tuples_frozen += tuples_frozen; vacrel->lpdead_items += lpdead_items; vacrel->live_tuples += live_tuples; diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h index 6598c4d7d8..2d3f149e4f 100644 --- a/src/include/access/heapam.h +++ b/src/include/access/heapam.h @@ -191,6 +191,15 @@ typedef struct HeapPageFreeze } HeapPageFreeze; +/* + * Per-page state returned from pruning + */ +typedef struct PruneResult +{ + int ndeleted; /* Number of tuples deleted from the page */ + int nnewlpdead; /* Number of newly LP_DEAD items */ +} PruneResult; + /* ---------------- * function prototypes for heap access method * @@ -284,9 +293,9 @@ extern TransactionId heap_index_delete_tuples(Relation rel, /* in heap/pruneheap.c */ struct GlobalVisState; extern void heap_page_prune_opt(Relation relation, Buffer buffer); -extern int heap_page_prune(Relation relation, Buffer buffer, +extern void heap_page_prune(Relation relation, Buffer buffer, struct GlobalVisState *vistest, - int *nnewlpdead, + PruneResult *presult, OffsetNumber *off_loc); extern void heap_page_prune_execute(Buffer buffer, OffsetNumber *redirected, int nredirected, diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index b5bbdd1608..8de90c4958 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -2151,6 +2151,7 @@ ProjectionPath PromptInterruptContext ProtocolVersion PrsStorage +PruneResult PruneState PruneStepResult PsqlScanCallbacks -- 2.37.2