Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * vacuumlazy.c
4 : * Concurrent ("lazy") vacuuming.
5 : *
6 : * Heap relations are vacuumed in three main phases. In phase I, vacuum scans
7 : * relation pages, pruning and freezing tuples and saving dead tuples' TIDs in
8 : * a TID store. If that TID store fills up or vacuum finishes scanning the
9 : * relation, it progresses to phase II: index vacuuming. Index vacuuming
10 : * deletes the dead index entries referenced in the TID store. In phase III,
11 : * vacuum scans the blocks of the relation referred to by the TIDs in the TID
12 : * store and reaps the corresponding dead items, freeing that space for future
13 : * tuples.
14 : *
15 : * If there are no indexes or index scanning is disabled, phase II may be
16 : * skipped. If phase I identified very few dead index entries or if vacuum's
17 : * failsafe mechanism has triggered (to avoid transaction ID wraparound),
18 : * vacuum may skip phases II and III.
19 : *
20 : * If the TID store fills up in phase I, vacuum suspends phase I and proceeds
21 : * to phases II and III, cleaning up the dead tuples referenced in the current
22 : * TID store. This empties the TID store, allowing vacuum to resume phase I.
23 : *
24 : * In a way, the phases are more like states in a state machine, but they have
25 : * been referred to colloquially as phases for so long that they are referred
26 : * to as such here.
27 : *
28 : * Manually invoked VACUUMs may scan indexes during phase II in parallel. For
29 : * more information on this, see the comment at the top of vacuumparallel.c.
30 : *
31 : * In between phases, vacuum updates the freespace map (every
32 : * VACUUM_FSM_EVERY_PAGES).
33 : *
34 : * After completing all three phases, vacuum may truncate the relation if it
35 : * has emptied pages at the end. Finally, vacuum updates relation statistics
36 : * in pg_class and the cumulative statistics subsystem.
37 : *
38 : * Relation Scanning:
39 : *
40 : * Vacuum scans the heap relation, starting at the beginning and progressing
41 : * to the end, skipping pages as permitted by their visibility status, vacuum
42 : * options, and various other requirements.
43 : *
44 : * Vacuums are either aggressive or normal. Aggressive vacuums must scan every
45 : * unfrozen tuple in order to advance relfrozenxid and avoid transaction ID
46 : * wraparound. Normal vacuums may scan otherwise skippable pages for one of
47 : * two reasons:
48 : *
49 : * When page skipping is not disabled, a normal vacuum may scan pages that are
50 : * marked all-visible (and even all-frozen) in the visibility map if the range
51 : * of skippable pages is below SKIP_PAGES_THRESHOLD. This is primarily for the
52 : * benefit of kernel readahead (see comment in heap_vac_scan_next_block()).
53 : *
54 : * A normal vacuum may also scan skippable pages in an effort to freeze them
55 : * and decrease the backlog of all-visible but not all-frozen pages that have
56 : * to be processed by the next aggressive vacuum. These are referred to as
57 : * eagerly scanned pages. Pages scanned due to SKIP_PAGES_THRESHOLD do not
58 : * count as eagerly scanned pages.
59 : *
60 : * Eagerly scanned pages that are set all-frozen in the VM are successful
61 : * eager freezes and those not set all-frozen in the VM are failed eager
62 : * freezes.
63 : *
64 : * Because we want to amortize the overhead of freezing pages over multiple
65 : * vacuums, normal vacuums cap the number of successful eager freezes to
66 : * MAX_EAGER_FREEZE_SUCCESS_RATE of the number of all-visible but not
67 : * all-frozen pages at the beginning of the vacuum. Since eagerly frozen pages
68 : * may be unfrozen before the next aggressive vacuum, capping the number of
69 : * successful eager freezes also caps the downside of eager freezing:
70 : * potentially wasted work.
71 : *
72 : * Once the success cap has been hit, eager scanning is disabled for the
73 : * remainder of the vacuum of the relation.
74 : *
75 : * Success is capped globally because we don't want to limit our successes if
76 : * old data happens to be concentrated in a particular part of the table. This
77 : * is especially likely to happen for append-mostly workloads where the oldest
78 : * data is at the beginning of the unfrozen portion of the relation.
79 : *
80 : * On the assumption that different regions of the table are likely to contain
81 : * similarly aged data, normal vacuums use a localized eager freeze failure
82 : * cap. The failure count is reset for each region of the table -- comprised
83 : * of EAGER_SCAN_REGION_SIZE blocks. In each region, we tolerate
84 : * vacuum_max_eager_freeze_failure_rate of EAGER_SCAN_REGION_SIZE failures
85 : * before suspending eager scanning until the end of the region.
86 : * vacuum_max_eager_freeze_failure_rate is configurable both globally and per
87 : * table.
88 : *
89 : * Aggressive vacuums must examine every unfrozen tuple and thus are not
90 : * subject to any of the limits imposed by the eager scanning algorithm.
91 : *
92 : * Once vacuum has decided to scan a given block, it must read the block and
93 : * obtain a cleanup lock to prune tuples on the page. A non-aggressive vacuum
94 : * may choose to skip pruning and freezing if it cannot acquire a cleanup lock
95 : * on the buffer right away. In this case, it may miss cleaning up dead tuples
96 : * and their associated index entries (though it is free to reap any existing
97 : * dead items on the page).
98 : *
99 : * After pruning and freezing, pages that are newly all-visible and all-frozen
100 : * are marked as such in the visibility map.
101 : *
102 : * Dead TID Storage:
103 : *
104 : * The major space usage for vacuuming is storage for the dead tuple IDs that
105 : * are to be removed from indexes. We want to ensure we can vacuum even the
106 : * very largest relations with finite memory space usage. To do that, we set
107 : * upper bounds on the memory that can be used for keeping track of dead TIDs
108 : * at once.
109 : *
110 : * We are willing to use at most maintenance_work_mem (or perhaps
111 : * autovacuum_work_mem) memory space to keep track of dead TIDs. If the
112 : * TID store is full, we must call lazy_vacuum to vacuum indexes (and to vacuum
113 : * the pages that we've pruned). This frees up the memory space dedicated to
114 : * store dead TIDs.
115 : *
116 : * In practice VACUUM will often complete its initial pass over the target
117 : * heap relation without ever running out of space to store TIDs. This means
118 : * that there only needs to be one call to lazy_vacuum, after the initial pass
119 : * completes.
120 : *
121 : * Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group
122 : * Portions Copyright (c) 1994, Regents of the University of California
123 : *
124 : *
125 : * IDENTIFICATION
126 : * src/backend/access/heap/vacuumlazy.c
127 : *
128 : *-------------------------------------------------------------------------
129 : */
130 : #include "postgres.h"
131 :
132 : #include <math.h>
133 :
134 : #include "access/genam.h"
135 : #include "access/heapam.h"
136 : #include "access/htup_details.h"
137 : #include "access/multixact.h"
138 : #include "access/tidstore.h"
139 : #include "access/transam.h"
140 : #include "access/visibilitymap.h"
141 : #include "access/xloginsert.h"
142 : #include "catalog/storage.h"
143 : #include "commands/dbcommands.h"
144 : #include "commands/progress.h"
145 : #include "commands/vacuum.h"
146 : #include "common/int.h"
147 : #include "common/pg_prng.h"
148 : #include "executor/instrument.h"
149 : #include "miscadmin.h"
150 : #include "pgstat.h"
151 : #include "portability/instr_time.h"
152 : #include "postmaster/autovacuum.h"
153 : #include "storage/bufmgr.h"
154 : #include "storage/freespace.h"
155 : #include "storage/lmgr.h"
156 : #include "storage/read_stream.h"
157 : #include "utils/lsyscache.h"
158 : #include "utils/pg_rusage.h"
159 : #include "utils/timestamp.h"
160 :
161 :
162 : /*
163 : * Space/time tradeoff parameters: do these need to be user-tunable?
164 : *
165 : * To consider truncating the relation, we want there to be at least
166 : * REL_TRUNCATE_MINIMUM or (relsize / REL_TRUNCATE_FRACTION) (whichever
167 : * is less) potentially-freeable pages.
168 : */
169 : #define REL_TRUNCATE_MINIMUM 1000
170 : #define REL_TRUNCATE_FRACTION 16
171 :
172 : /*
173 : * Timing parameters for truncate locking heuristics.
174 : *
175 : * These were not exposed as user tunable GUC values because it didn't seem
176 : * that the potential for improvement was great enough to merit the cost of
177 : * supporting them.
178 : */
179 : #define VACUUM_TRUNCATE_LOCK_CHECK_INTERVAL 20 /* ms */
180 : #define VACUUM_TRUNCATE_LOCK_WAIT_INTERVAL 50 /* ms */
181 : #define VACUUM_TRUNCATE_LOCK_TIMEOUT 5000 /* ms */
182 :
183 : /*
184 : * Threshold that controls whether we bypass index vacuuming and heap
185 : * vacuuming as an optimization
186 : */
187 : #define BYPASS_THRESHOLD_PAGES 0.02 /* i.e. 2% of rel_pages */
188 :
189 : /*
190 : * Perform a failsafe check each time we scan another 4GB of pages.
191 : * (Note that this is deliberately kept to a power-of-two, usually 2^19.)
192 : */
193 : #define FAILSAFE_EVERY_PAGES \
194 : ((BlockNumber) (((uint64) 4 * 1024 * 1024 * 1024) / BLCKSZ))
195 :
196 : /*
197 : * When a table has no indexes, vacuum the FSM after every 8GB, approximately
198 : * (it won't be exact because we only vacuum FSM after processing a heap page
199 : * that has some removable tuples). When there are indexes, this is ignored,
200 : * and we vacuum FSM after each index/heap cleaning pass.
201 : */
202 : #define VACUUM_FSM_EVERY_PAGES \
203 : ((BlockNumber) (((uint64) 8 * 1024 * 1024 * 1024) / BLCKSZ))
204 :
205 : /*
206 : * Before we consider skipping a page that's marked as clean in
207 : * visibility map, we must've seen at least this many clean pages.
208 : */
209 : #define SKIP_PAGES_THRESHOLD ((BlockNumber) 32)
210 :
211 : /*
212 : * Size of the prefetch window for lazy vacuum backwards truncation scan.
213 : * Needs to be a power of 2.
214 : */
215 : #define PREFETCH_SIZE ((BlockNumber) 32)
216 :
217 : /*
218 : * Macro to check if we are in a parallel vacuum. If true, we are in the
219 : * parallel mode and the DSM segment is initialized.
220 : */
221 : #define ParallelVacuumIsActive(vacrel) ((vacrel)->pvs != NULL)
222 :
223 : /* Phases of vacuum during which we report error context. */
224 : typedef enum
225 : {
226 : VACUUM_ERRCB_PHASE_UNKNOWN,
227 : VACUUM_ERRCB_PHASE_SCAN_HEAP,
228 : VACUUM_ERRCB_PHASE_VACUUM_INDEX,
229 : VACUUM_ERRCB_PHASE_VACUUM_HEAP,
230 : VACUUM_ERRCB_PHASE_INDEX_CLEANUP,
231 : VACUUM_ERRCB_PHASE_TRUNCATE,
232 : } VacErrPhase;
233 :
234 : /*
235 : * An eager scan of a page that is set all-frozen in the VM is considered
236 : * "successful". To spread out freezing overhead across multiple normal
237 : * vacuums, we limit the number of successful eager page freezes. The maximum
238 : * number of eager page freezes is calculated as a ratio of the all-visible
239 : * but not all-frozen pages at the beginning of the vacuum.
240 : */
241 : #define MAX_EAGER_FREEZE_SUCCESS_RATE 0.2
242 :
243 : /*
244 : * On the assumption that different regions of the table tend to have
245 : * similarly aged data, once vacuum fails to freeze
246 : * vacuum_max_eager_freeze_failure_rate of the blocks in a region of size
247 : * EAGER_SCAN_REGION_SIZE, it suspends eager scanning until it has progressed
248 : * to another region of the table with potentially older data.
249 : */
250 : #define EAGER_SCAN_REGION_SIZE 4096
251 :
252 : /*
253 : * heap_vac_scan_next_block() sets these flags to communicate information
254 : * about the block it read to the caller.
255 : */
256 : #define VAC_BLK_WAS_EAGER_SCANNED (1 << 0)
257 : #define VAC_BLK_ALL_VISIBLE_ACCORDING_TO_VM (1 << 1)
258 :
259 : typedef struct LVRelState
260 : {
261 : /* Target heap relation and its indexes */
262 : Relation rel;
263 : Relation *indrels;
264 : int nindexes;
265 :
266 : /* Buffer access strategy and parallel vacuum state */
267 : BufferAccessStrategy bstrategy;
268 : ParallelVacuumState *pvs;
269 :
270 : /* Aggressive VACUUM? (must set relfrozenxid >= FreezeLimit) */
271 : bool aggressive;
272 : /* Use visibility map to skip? (disabled by DISABLE_PAGE_SKIPPING) */
273 : bool skipwithvm;
274 : /* Consider index vacuuming bypass optimization? */
275 : bool consider_bypass_optimization;
276 :
277 : /* Doing index vacuuming, index cleanup, rel truncation? */
278 : bool do_index_vacuuming;
279 : bool do_index_cleanup;
280 : bool do_rel_truncate;
281 :
282 : /* VACUUM operation's cutoffs for freezing and pruning */
283 : struct VacuumCutoffs cutoffs;
284 : GlobalVisState *vistest;
285 : /* Tracks oldest extant XID/MXID for setting relfrozenxid/relminmxid */
286 : TransactionId NewRelfrozenXid;
287 : MultiXactId NewRelminMxid;
288 : bool skippedallvis;
289 :
290 : /* Error reporting state */
291 : char *dbname;
292 : char *relnamespace;
293 : char *relname;
294 : char *indname; /* Current index name */
295 : BlockNumber blkno; /* used only for heap operations */
296 : OffsetNumber offnum; /* used only for heap operations */
297 : VacErrPhase phase;
298 : bool verbose; /* VACUUM VERBOSE? */
299 :
300 : /*
301 : * dead_items stores TIDs whose index tuples are deleted by index
302 : * vacuuming. Each TID points to an LP_DEAD line pointer from a heap page
303 : * that has been processed by lazy_scan_prune. Also needed by
304 : * lazy_vacuum_heap_rel, which marks the same LP_DEAD line pointers as
305 : * LP_UNUSED during second heap pass.
306 : *
307 : * Both dead_items and dead_items_info are allocated in shared memory in
308 : * parallel vacuum cases.
309 : */
310 : TidStore *dead_items; /* TIDs whose index tuples we'll delete */
311 : VacDeadItemsInfo *dead_items_info;
312 :
313 : BlockNumber rel_pages; /* total number of pages */
314 : BlockNumber scanned_pages; /* # pages examined (not skipped via VM) */
315 :
316 : /*
317 : * Count of all-visible blocks eagerly scanned (for logging only). This
318 : * does not include skippable blocks scanned due to SKIP_PAGES_THRESHOLD.
319 : */
320 : BlockNumber eager_scanned_pages;
321 :
322 : BlockNumber removed_pages; /* # pages removed by relation truncation */
323 : BlockNumber new_frozen_tuple_pages; /* # pages with newly frozen tuples */
324 :
325 : /* # pages newly set all-visible in the VM */
326 : BlockNumber vm_new_visible_pages;
327 :
328 : /*
329 : * # pages newly set all-visible and all-frozen in the VM. This is a
330 : * subset of vm_new_visible_pages. That is, vm_new_visible_pages includes
331 : * all pages set all-visible, but vm_new_visible_frozen_pages includes
332 : * only those which were also set all-frozen.
333 : */
334 : BlockNumber vm_new_visible_frozen_pages;
335 :
336 : /* # all-visible pages newly set all-frozen in the VM */
337 : BlockNumber vm_new_frozen_pages;
338 :
339 : BlockNumber lpdead_item_pages; /* # pages with LP_DEAD items */
340 : BlockNumber missed_dead_pages; /* # pages with missed dead tuples */
341 : BlockNumber nonempty_pages; /* actually, last nonempty page + 1 */
342 :
343 : /* Statistics output by us, for table */
344 : double new_rel_tuples; /* new estimated total # of tuples */
345 : double new_live_tuples; /* new estimated total # of live tuples */
346 : /* Statistics output by index AMs */
347 : IndexBulkDeleteResult **indstats;
348 :
349 : /* Instrumentation counters */
350 : int num_index_scans;
351 : /* Counters that follow are only for scanned_pages */
352 : int64 tuples_deleted; /* # deleted from table */
353 : int64 tuples_frozen; /* # newly frozen */
354 : int64 lpdead_items; /* # deleted from indexes */
355 : int64 live_tuples; /* # live tuples remaining */
356 : int64 recently_dead_tuples; /* # dead, but not yet removable */
357 : int64 missed_dead_tuples; /* # removable, but not removed */
358 :
359 : /* State maintained by heap_vac_scan_next_block() */
360 : BlockNumber current_block; /* last block returned */
361 : BlockNumber next_unskippable_block; /* next unskippable block */
362 : bool next_unskippable_allvis; /* its visibility status */
363 : bool next_unskippable_eager_scanned; /* if it was eagerly scanned */
364 : Buffer next_unskippable_vmbuffer; /* buffer containing its VM bit */
365 :
366 : /* State related to managing eager scanning of all-visible pages */
367 :
368 : /*
369 : * A normal vacuum that has failed to freeze too many eagerly scanned
370 : * blocks in a region suspends eager scanning.
371 : * next_eager_scan_region_start is the block number of the first block
372 : * eligible for resumed eager scanning.
373 : *
374 : * When eager scanning is permanently disabled, either initially
375 : * (including for aggressive vacuum) or due to hitting the success cap,
376 : * this is set to InvalidBlockNumber.
377 : */
378 : BlockNumber next_eager_scan_region_start;
379 :
380 : /*
381 : * The remaining number of blocks a normal vacuum will consider eager
382 : * scanning when it is successful. When eager scanning is enabled, this is
383 : * initialized to MAX_EAGER_FREEZE_SUCCESS_RATE of the total number of
384 : * all-visible but not all-frozen pages. For each eager freeze success,
385 : * this is decremented. Once it hits 0, eager scanning is permanently
386 : * disabled. It is initialized to 0 if eager scanning starts out disabled
387 : * (including for aggressive vacuum).
388 : */
389 : BlockNumber eager_scan_remaining_successes;
390 :
391 : /*
392 : * The maximum number of blocks which may be eagerly scanned and not
393 : * frozen before eager scanning is temporarily suspended. This is
394 : * configurable both globally, via the
395 : * vacuum_max_eager_freeze_failure_rate GUC, and per table, with a table
396 : * storage parameter of the same name. It is calculated as
397 : * vacuum_max_eager_freeze_failure_rate of EAGER_SCAN_REGION_SIZE blocks.
398 : * It is 0 when eager scanning is disabled.
399 : */
400 : BlockNumber eager_scan_max_fails_per_region;
401 :
402 : /*
403 : * The number of eagerly scanned blocks vacuum failed to freeze (due to
404 : * age) in the current eager scan region. Vacuum resets it to
405 : * eager_scan_max_fails_per_region each time it enters a new region of the
406 : * relation. If eager_scan_remaining_fails hits 0, eager scanning is
407 : * suspended until the next region. It is also 0 if eager scanning has
408 : * been permanently disabled.
409 : */
410 : BlockNumber eager_scan_remaining_fails;
411 : } LVRelState;
412 :
413 :
414 : /* Struct for saving and restoring vacuum error information. */
415 : typedef struct LVSavedErrInfo
416 : {
417 : BlockNumber blkno;
418 : OffsetNumber offnum;
419 : VacErrPhase phase;
420 : } LVSavedErrInfo;
421 :
422 :
423 : /* non-export function prototypes */
424 : static void lazy_scan_heap(LVRelState *vacrel);
425 : static void heap_vacuum_eager_scan_setup(LVRelState *vacrel,
426 : VacuumParams *params);
427 : static BlockNumber heap_vac_scan_next_block(ReadStream *stream,
428 : void *callback_private_data,
429 : void *per_buffer_data);
430 : static void find_next_unskippable_block(LVRelState *vacrel, bool *skipsallvis);
431 : static bool lazy_scan_new_or_empty(LVRelState *vacrel, Buffer buf,
432 : BlockNumber blkno, Page page,
433 : bool sharelock, Buffer vmbuffer);
434 : static void lazy_scan_prune(LVRelState *vacrel, Buffer buf,
435 : BlockNumber blkno, Page page,
436 : Buffer vmbuffer, bool all_visible_according_to_vm,
437 : bool *has_lpdead_items, bool *vm_page_frozen);
438 : static bool lazy_scan_noprune(LVRelState *vacrel, Buffer buf,
439 : BlockNumber blkno, Page page,
440 : bool *has_lpdead_items);
441 : static void lazy_vacuum(LVRelState *vacrel);
442 : static bool lazy_vacuum_all_indexes(LVRelState *vacrel);
443 : static void lazy_vacuum_heap_rel(LVRelState *vacrel);
444 : static void lazy_vacuum_heap_page(LVRelState *vacrel, BlockNumber blkno,
445 : Buffer buffer, OffsetNumber *deadoffsets,
446 : int num_offsets, Buffer vmbuffer);
447 : static bool lazy_check_wraparound_failsafe(LVRelState *vacrel);
448 : static void lazy_cleanup_all_indexes(LVRelState *vacrel);
449 : static IndexBulkDeleteResult *lazy_vacuum_one_index(Relation indrel,
450 : IndexBulkDeleteResult *istat,
451 : double reltuples,
452 : LVRelState *vacrel);
453 : static IndexBulkDeleteResult *lazy_cleanup_one_index(Relation indrel,
454 : IndexBulkDeleteResult *istat,
455 : double reltuples,
456 : bool estimated_count,
457 : LVRelState *vacrel);
458 : static bool should_attempt_truncation(LVRelState *vacrel);
459 : static void lazy_truncate_heap(LVRelState *vacrel);
460 : static BlockNumber count_nondeletable_pages(LVRelState *vacrel,
461 : bool *lock_waiter_detected);
462 : static void dead_items_alloc(LVRelState *vacrel, int nworkers);
463 : static void dead_items_add(LVRelState *vacrel, BlockNumber blkno, OffsetNumber *offsets,
464 : int num_offsets);
465 : static void dead_items_reset(LVRelState *vacrel);
466 : static void dead_items_cleanup(LVRelState *vacrel);
467 : static bool heap_page_is_all_visible(LVRelState *vacrel, Buffer buf,
468 : TransactionId *visibility_cutoff_xid, bool *all_frozen);
469 : static void update_relstats_all_indexes(LVRelState *vacrel);
470 : static void vacuum_error_callback(void *arg);
471 : static void update_vacuum_error_info(LVRelState *vacrel,
472 : LVSavedErrInfo *saved_vacrel,
473 : int phase, BlockNumber blkno,
474 : OffsetNumber offnum);
475 : static void restore_vacuum_error_info(LVRelState *vacrel,
476 : const LVSavedErrInfo *saved_vacrel);
477 :
478 :
479 :
480 : /*
481 : * Helper to set up the eager scanning state for vacuuming a single relation.
482 : * Initializes the eager scan management related members of the LVRelState.
483 : *
484 : * Caller provides whether or not an aggressive vacuum is required due to
485 : * vacuum options or for relfrozenxid/relminmxid advancement.
486 : */
487 : static void
488 129846 : heap_vacuum_eager_scan_setup(LVRelState *vacrel, VacuumParams *params)
489 : {
490 : uint32 randseed;
491 : BlockNumber allvisible;
492 : BlockNumber allfrozen;
493 : float first_region_ratio;
494 129846 : bool oldest_unfrozen_before_cutoff = false;
495 :
496 : /*
497 : * Initialize eager scan management fields to their disabled values.
498 : * Aggressive vacuums, normal vacuums of small tables, and normal vacuums
499 : * of tables without sufficiently old tuples disable eager scanning.
500 : */
501 129846 : vacrel->next_eager_scan_region_start = InvalidBlockNumber;
502 129846 : vacrel->eager_scan_max_fails_per_region = 0;
503 129846 : vacrel->eager_scan_remaining_fails = 0;
504 129846 : vacrel->eager_scan_remaining_successes = 0;
505 :
506 : /* If eager scanning is explicitly disabled, just return. */
507 129846 : if (params->max_eager_freeze_failure_rate == 0)
508 129846 : return;
509 :
510 : /*
511 : * The caller will have determined whether or not an aggressive vacuum is
512 : * required by either the vacuum parameters or the relative age of the
513 : * oldest unfrozen transaction IDs. An aggressive vacuum must scan every
514 : * all-visible page to safely advance the relfrozenxid and/or relminmxid,
515 : * so scans of all-visible pages are not considered eager.
516 : */
517 129846 : if (vacrel->aggressive)
518 118998 : return;
519 :
520 : /*
521 : * Aggressively vacuuming a small relation shouldn't take long, so it
522 : * isn't worth amortizing. We use two times the region size as the size
523 : * cutoff because the eager scan start block is a random spot somewhere in
524 : * the first region, making the second region the first to be eager
525 : * scanned normally.
526 : */
527 10848 : if (vacrel->rel_pages < 2 * EAGER_SCAN_REGION_SIZE)
528 10848 : return;
529 :
530 : /*
531 : * We only want to enable eager scanning if we are likely to be able to
532 : * freeze some of the pages in the relation.
533 : *
534 : * Tuples with XIDs older than OldestXmin or MXIDs older than OldestMxact
535 : * are technically freezable, but we won't freeze them unless the criteria
536 : * for opportunistic freezing is met. Only tuples with XIDs/MXIDs older
537 : * than the FreezeLimit/MultiXactCutoff are frozen in the common case.
538 : *
539 : * So, as a heuristic, we wait until the FreezeLimit has advanced past the
540 : * relfrozenxid or the MultiXactCutoff has advanced past the relminmxid to
541 : * enable eager scanning.
542 : */
543 0 : if (TransactionIdIsNormal(vacrel->cutoffs.relfrozenxid) &&
544 0 : TransactionIdPrecedes(vacrel->cutoffs.relfrozenxid,
545 : vacrel->cutoffs.FreezeLimit))
546 0 : oldest_unfrozen_before_cutoff = true;
547 :
548 0 : if (!oldest_unfrozen_before_cutoff &&
549 0 : MultiXactIdIsValid(vacrel->cutoffs.relminmxid) &&
550 0 : MultiXactIdPrecedes(vacrel->cutoffs.relminmxid,
551 : vacrel->cutoffs.MultiXactCutoff))
552 0 : oldest_unfrozen_before_cutoff = true;
553 :
554 0 : if (!oldest_unfrozen_before_cutoff)
555 0 : return;
556 :
557 : /* We have met the criteria to eagerly scan some pages. */
558 :
559 : /*
560 : * Our success cap is MAX_EAGER_FREEZE_SUCCESS_RATE of the number of
561 : * all-visible but not all-frozen blocks in the relation.
562 : */
563 0 : visibilitymap_count(vacrel->rel, &allvisible, &allfrozen);
564 :
565 0 : vacrel->eager_scan_remaining_successes =
566 0 : (BlockNumber) (MAX_EAGER_FREEZE_SUCCESS_RATE *
567 0 : (allvisible - allfrozen));
568 :
569 : /* If every all-visible page is frozen, eager scanning is disabled. */
570 0 : if (vacrel->eager_scan_remaining_successes == 0)
571 0 : return;
572 :
573 : /*
574 : * Now calculate the bounds of the first eager scan region. Its end block
575 : * will be a random spot somewhere in the first EAGER_SCAN_REGION_SIZE
576 : * blocks. This affects the bounds of all subsequent regions and avoids
577 : * eager scanning and failing to freeze the same blocks each vacuum of the
578 : * relation.
579 : */
580 0 : randseed = pg_prng_uint32(&pg_global_prng_state);
581 :
582 0 : vacrel->next_eager_scan_region_start = randseed % EAGER_SCAN_REGION_SIZE;
583 :
584 : Assert(params->max_eager_freeze_failure_rate > 0 &&
585 : params->max_eager_freeze_failure_rate <= 1);
586 :
587 0 : vacrel->eager_scan_max_fails_per_region =
588 0 : params->max_eager_freeze_failure_rate *
589 : EAGER_SCAN_REGION_SIZE;
590 :
591 : /*
592 : * The first region will be smaller than subsequent regions. As such,
593 : * adjust the eager freeze failures tolerated for this region.
594 : */
595 0 : first_region_ratio = 1 - (float) vacrel->next_eager_scan_region_start /
596 : EAGER_SCAN_REGION_SIZE;
597 :
598 0 : vacrel->eager_scan_remaining_fails =
599 0 : vacrel->eager_scan_max_fails_per_region *
600 : first_region_ratio;
601 : }
602 :
603 : /*
604 : * heap_vacuum_rel() -- perform VACUUM for one heap relation
605 : *
606 : * This routine sets things up for and then calls lazy_scan_heap, where
607 : * almost all work actually takes place. Finalizes everything after call
608 : * returns by managing relation truncation and updating rel's pg_class
609 : * entry. (Also updates pg_class entries for any indexes that need it.)
610 : *
611 : * At entry, we have already established a transaction and opened
612 : * and locked the relation.
613 : */
614 : void
615 129846 : heap_vacuum_rel(Relation rel, VacuumParams *params,
616 : BufferAccessStrategy bstrategy)
617 : {
618 : LVRelState *vacrel;
619 : bool verbose,
620 : instrument,
621 : skipwithvm,
622 : frozenxid_updated,
623 : minmulti_updated;
624 : BlockNumber orig_rel_pages,
625 : new_rel_pages,
626 : new_rel_allvisible,
627 : new_rel_allfrozen;
628 : PGRUsage ru0;
629 129846 : TimestampTz starttime = 0;
630 129846 : PgStat_Counter startreadtime = 0,
631 129846 : startwritetime = 0;
632 129846 : WalUsage startwalusage = pgWalUsage;
633 129846 : BufferUsage startbufferusage = pgBufferUsage;
634 : ErrorContextCallback errcallback;
635 129846 : char **indnames = NULL;
636 :
637 129846 : verbose = (params->options & VACOPT_VERBOSE) != 0;
638 233570 : instrument = (verbose || (AmAutoVacuumWorkerProcess() &&
639 103724 : params->log_min_duration >= 0));
640 129846 : if (instrument)
641 : {
642 103744 : pg_rusage_init(&ru0);
643 103744 : if (track_io_timing)
644 : {
645 0 : startreadtime = pgStatBlockReadTime;
646 0 : startwritetime = pgStatBlockWriteTime;
647 : }
648 : }
649 :
650 : /* Used for instrumentation and stats report */
651 129846 : starttime = GetCurrentTimestamp();
652 :
653 129846 : pgstat_progress_start_command(PROGRESS_COMMAND_VACUUM,
654 : RelationGetRelid(rel));
655 :
656 : /*
657 : * Setup error traceback support for ereport() first. The idea is to set
658 : * up an error context callback to display additional information on any
659 : * error during a vacuum. During different phases of vacuum, we update
660 : * the state so that the error context callback always display current
661 : * information.
662 : *
663 : * Copy the names of heap rel into local memory for error reporting
664 : * purposes, too. It isn't always safe to assume that we can get the name
665 : * of each rel. It's convenient for code in lazy_scan_heap to always use
666 : * these temp copies.
667 : */
668 129846 : vacrel = (LVRelState *) palloc0(sizeof(LVRelState));
669 129846 : vacrel->dbname = get_database_name(MyDatabaseId);
670 129846 : vacrel->relnamespace = get_namespace_name(RelationGetNamespace(rel));
671 129846 : vacrel->relname = pstrdup(RelationGetRelationName(rel));
672 129846 : vacrel->indname = NULL;
673 129846 : vacrel->phase = VACUUM_ERRCB_PHASE_UNKNOWN;
674 129846 : vacrel->verbose = verbose;
675 129846 : errcallback.callback = vacuum_error_callback;
676 129846 : errcallback.arg = vacrel;
677 129846 : errcallback.previous = error_context_stack;
678 129846 : error_context_stack = &errcallback;
679 :
680 : /* Set up high level stuff about rel and its indexes */
681 129846 : vacrel->rel = rel;
682 129846 : vac_open_indexes(vacrel->rel, RowExclusiveLock, &vacrel->nindexes,
683 : &vacrel->indrels);
684 129846 : vacrel->bstrategy = bstrategy;
685 129846 : if (instrument && vacrel->nindexes > 0)
686 : {
687 : /* Copy index names used by instrumentation (not error reporting) */
688 99482 : indnames = palloc(sizeof(char *) * vacrel->nindexes);
689 255186 : for (int i = 0; i < vacrel->nindexes; i++)
690 155704 : indnames[i] = pstrdup(RelationGetRelationName(vacrel->indrels[i]));
691 : }
692 :
693 : /*
694 : * The index_cleanup param either disables index vacuuming and cleanup or
695 : * forces it to go ahead when we would otherwise apply the index bypass
696 : * optimization. The default is 'auto', which leaves the final decision
697 : * up to lazy_vacuum().
698 : *
699 : * The truncate param allows user to avoid attempting relation truncation,
700 : * though it can't force truncation to happen.
701 : */
702 : Assert(params->index_cleanup != VACOPTVALUE_UNSPECIFIED);
703 : Assert(params->truncate != VACOPTVALUE_UNSPECIFIED &&
704 : params->truncate != VACOPTVALUE_AUTO);
705 :
706 : /*
707 : * While VacuumFailSafeActive is reset to false before calling this, we
708 : * still need to reset it here due to recursive calls.
709 : */
710 129846 : VacuumFailsafeActive = false;
711 129846 : vacrel->consider_bypass_optimization = true;
712 129846 : vacrel->do_index_vacuuming = true;
713 129846 : vacrel->do_index_cleanup = true;
714 129846 : vacrel->do_rel_truncate = (params->truncate != VACOPTVALUE_DISABLED);
715 129846 : if (params->index_cleanup == VACOPTVALUE_DISABLED)
716 : {
717 : /* Force disable index vacuuming up-front */
718 264 : vacrel->do_index_vacuuming = false;
719 264 : vacrel->do_index_cleanup = false;
720 : }
721 129582 : else if (params->index_cleanup == VACOPTVALUE_ENABLED)
722 : {
723 : /* Force index vacuuming. Note that failsafe can still bypass. */
724 32 : vacrel->consider_bypass_optimization = false;
725 : }
726 : else
727 : {
728 : /* Default/auto, make all decisions dynamically */
729 : Assert(params->index_cleanup == VACOPTVALUE_AUTO);
730 : }
731 :
732 : /* Initialize page counters explicitly (be tidy) */
733 129846 : vacrel->scanned_pages = 0;
734 129846 : vacrel->eager_scanned_pages = 0;
735 129846 : vacrel->removed_pages = 0;
736 129846 : vacrel->new_frozen_tuple_pages = 0;
737 129846 : vacrel->lpdead_item_pages = 0;
738 129846 : vacrel->missed_dead_pages = 0;
739 129846 : vacrel->nonempty_pages = 0;
740 : /* dead_items_alloc allocates vacrel->dead_items later on */
741 :
742 : /* Allocate/initialize output statistics state */
743 129846 : vacrel->new_rel_tuples = 0;
744 129846 : vacrel->new_live_tuples = 0;
745 129846 : vacrel->indstats = (IndexBulkDeleteResult **)
746 129846 : palloc0(vacrel->nindexes * sizeof(IndexBulkDeleteResult *));
747 :
748 : /* Initialize remaining counters (be tidy) */
749 129846 : vacrel->num_index_scans = 0;
750 129846 : vacrel->tuples_deleted = 0;
751 129846 : vacrel->tuples_frozen = 0;
752 129846 : vacrel->lpdead_items = 0;
753 129846 : vacrel->live_tuples = 0;
754 129846 : vacrel->recently_dead_tuples = 0;
755 129846 : vacrel->missed_dead_tuples = 0;
756 :
757 129846 : vacrel->vm_new_visible_pages = 0;
758 129846 : vacrel->vm_new_visible_frozen_pages = 0;
759 129846 : vacrel->vm_new_frozen_pages = 0;
760 129846 : vacrel->rel_pages = orig_rel_pages = RelationGetNumberOfBlocks(rel);
761 :
762 : /*
763 : * Get cutoffs that determine which deleted tuples are considered DEAD,
764 : * not just RECENTLY_DEAD, and which XIDs/MXIDs to freeze. Then determine
765 : * the extent of the blocks that we'll scan in lazy_scan_heap. It has to
766 : * happen in this order to ensure that the OldestXmin cutoff field works
767 : * as an upper bound on the XIDs stored in the pages we'll actually scan
768 : * (NewRelfrozenXid tracking must never be allowed to miss unfrozen XIDs).
769 : *
770 : * Next acquire vistest, a related cutoff that's used in pruning. We use
771 : * vistest in combination with OldestXmin to ensure that
772 : * heap_page_prune_and_freeze() always removes any deleted tuple whose
773 : * xmax is < OldestXmin. lazy_scan_prune must never become confused about
774 : * whether a tuple should be frozen or removed. (In the future we might
775 : * want to teach lazy_scan_prune to recompute vistest from time to time,
776 : * to increase the number of dead tuples it can prune away.)
777 : */
778 129846 : vacrel->aggressive = vacuum_get_cutoffs(rel, params, &vacrel->cutoffs);
779 129846 : vacrel->vistest = GlobalVisTestFor(rel);
780 : /* Initialize state used to track oldest extant XID/MXID */
781 129846 : vacrel->NewRelfrozenXid = vacrel->cutoffs.OldestXmin;
782 129846 : vacrel->NewRelminMxid = vacrel->cutoffs.OldestMxact;
783 :
784 : /*
785 : * Initialize state related to tracking all-visible page skipping. This is
786 : * very important to determine whether or not it is safe to advance the
787 : * relfrozenxid/relminmxid.
788 : */
789 129846 : vacrel->skippedallvis = false;
790 129846 : skipwithvm = true;
791 129846 : if (params->options & VACOPT_DISABLE_PAGE_SKIPPING)
792 : {
793 : /*
794 : * Force aggressive mode, and disable skipping blocks using the
795 : * visibility map (even those set all-frozen)
796 : */
797 346 : vacrel->aggressive = true;
798 346 : skipwithvm = false;
799 : }
800 :
801 129846 : vacrel->skipwithvm = skipwithvm;
802 :
803 : /*
804 : * Set up eager scan tracking state. This must happen after determining
805 : * whether or not the vacuum must be aggressive, because only normal
806 : * vacuums use the eager scan algorithm.
807 : */
808 129846 : heap_vacuum_eager_scan_setup(vacrel, params);
809 :
810 129846 : if (verbose)
811 : {
812 20 : if (vacrel->aggressive)
813 0 : ereport(INFO,
814 : (errmsg("aggressively vacuuming \"%s.%s.%s\"",
815 : vacrel->dbname, vacrel->relnamespace,
816 : vacrel->relname)));
817 : else
818 20 : ereport(INFO,
819 : (errmsg("vacuuming \"%s.%s.%s\"",
820 : vacrel->dbname, vacrel->relnamespace,
821 : vacrel->relname)));
822 : }
823 :
824 : /*
825 : * Allocate dead_items memory using dead_items_alloc. This handles
826 : * parallel VACUUM initialization as part of allocating shared memory
827 : * space used for dead_items. (But do a failsafe precheck first, to
828 : * ensure that parallel VACUUM won't be attempted at all when relfrozenxid
829 : * is already dangerously old.)
830 : */
831 129846 : lazy_check_wraparound_failsafe(vacrel);
832 129846 : dead_items_alloc(vacrel, params->nworkers);
833 :
834 : /*
835 : * Call lazy_scan_heap to perform all required heap pruning, index
836 : * vacuuming, and heap vacuuming (plus related processing)
837 : */
838 129846 : lazy_scan_heap(vacrel);
839 :
840 : /*
841 : * Free resources managed by dead_items_alloc. This ends parallel mode in
842 : * passing when necessary.
843 : */
844 129846 : dead_items_cleanup(vacrel);
845 : Assert(!IsInParallelMode());
846 :
847 : /*
848 : * Update pg_class entries for each of rel's indexes where appropriate.
849 : *
850 : * Unlike the later update to rel's pg_class entry, this is not critical.
851 : * Maintains relpages/reltuples statistics used by the planner only.
852 : */
853 129846 : if (vacrel->do_index_cleanup)
854 107880 : update_relstats_all_indexes(vacrel);
855 :
856 : /* Done with rel's indexes */
857 129846 : vac_close_indexes(vacrel->nindexes, vacrel->indrels, NoLock);
858 :
859 : /* Optionally truncate rel */
860 129846 : if (should_attempt_truncation(vacrel))
861 312 : lazy_truncate_heap(vacrel);
862 :
863 : /* Pop the error context stack */
864 129846 : error_context_stack = errcallback.previous;
865 :
866 : /* Report that we are now doing final cleanup */
867 129846 : pgstat_progress_update_param(PROGRESS_VACUUM_PHASE,
868 : PROGRESS_VACUUM_PHASE_FINAL_CLEANUP);
869 :
870 : /*
871 : * Prepare to update rel's pg_class entry.
872 : *
873 : * Aggressive VACUUMs must always be able to advance relfrozenxid to a
874 : * value >= FreezeLimit, and relminmxid to a value >= MultiXactCutoff.
875 : * Non-aggressive VACUUMs may advance them by any amount, or not at all.
876 : */
877 : Assert(vacrel->NewRelfrozenXid == vacrel->cutoffs.OldestXmin ||
878 : TransactionIdPrecedesOrEquals(vacrel->aggressive ? vacrel->cutoffs.FreezeLimit :
879 : vacrel->cutoffs.relfrozenxid,
880 : vacrel->NewRelfrozenXid));
881 : Assert(vacrel->NewRelminMxid == vacrel->cutoffs.OldestMxact ||
882 : MultiXactIdPrecedesOrEquals(vacrel->aggressive ? vacrel->cutoffs.MultiXactCutoff :
883 : vacrel->cutoffs.relminmxid,
884 : vacrel->NewRelminMxid));
885 129846 : if (vacrel->skippedallvis)
886 : {
887 : /*
888 : * Must keep original relfrozenxid in a non-aggressive VACUUM that
889 : * chose to skip an all-visible page range. The state that tracks new
890 : * values will have missed unfrozen XIDs from the pages we skipped.
891 : */
892 : Assert(!vacrel->aggressive);
893 50 : vacrel->NewRelfrozenXid = InvalidTransactionId;
894 50 : vacrel->NewRelminMxid = InvalidMultiXactId;
895 : }
896 :
897 : /*
898 : * For safety, clamp relallvisible to be not more than what we're setting
899 : * pg_class.relpages to
900 : */
901 129846 : new_rel_pages = vacrel->rel_pages; /* After possible rel truncation */
902 129846 : visibilitymap_count(rel, &new_rel_allvisible, &new_rel_allfrozen);
903 129846 : if (new_rel_allvisible > new_rel_pages)
904 0 : new_rel_allvisible = new_rel_pages;
905 :
906 : /*
907 : * An all-frozen block _must_ be all-visible. As such, clamp the count of
908 : * all-frozen blocks to the count of all-visible blocks. This matches the
909 : * clamping of relallvisible above.
910 : */
911 129846 : if (new_rel_allfrozen > new_rel_allvisible)
912 0 : new_rel_allfrozen = new_rel_allvisible;
913 :
914 : /*
915 : * Now actually update rel's pg_class entry.
916 : *
917 : * In principle new_live_tuples could be -1 indicating that we (still)
918 : * don't know the tuple count. In practice that can't happen, since we
919 : * scan every page that isn't skipped using the visibility map.
920 : */
921 129846 : vac_update_relstats(rel, new_rel_pages, vacrel->new_live_tuples,
922 : new_rel_allvisible, new_rel_allfrozen,
923 129846 : vacrel->nindexes > 0,
924 : vacrel->NewRelfrozenXid, vacrel->NewRelminMxid,
925 : &frozenxid_updated, &minmulti_updated, false);
926 :
927 : /*
928 : * Report results to the cumulative stats system, too.
929 : *
930 : * Deliberately avoid telling the stats system about LP_DEAD items that
931 : * remain in the table due to VACUUM bypassing index and heap vacuuming.
932 : * ANALYZE will consider the remaining LP_DEAD items to be dead "tuples".
933 : * It seems like a good idea to err on the side of not vacuuming again too
934 : * soon in cases where the failsafe prevented significant amounts of heap
935 : * vacuuming.
936 : */
937 78000 : pgstat_report_vacuum(RelationGetRelid(rel),
938 129846 : rel->rd_rel->relisshared,
939 51846 : Max(vacrel->new_live_tuples, 0),
940 129846 : vacrel->recently_dead_tuples +
941 129846 : vacrel->missed_dead_tuples,
942 : starttime);
943 129846 : pgstat_progress_end_command();
944 :
945 129846 : if (instrument)
946 : {
947 103744 : TimestampTz endtime = GetCurrentTimestamp();
948 :
949 103776 : if (verbose || params->log_min_duration == 0 ||
950 32 : TimestampDifferenceExceeds(starttime, endtime,
951 : params->log_min_duration))
952 : {
953 : long secs_dur;
954 : int usecs_dur;
955 : WalUsage walusage;
956 : BufferUsage bufferusage;
957 : StringInfoData buf;
958 : char *msgfmt;
959 : int32 diff;
960 103712 : double read_rate = 0,
961 103712 : write_rate = 0;
962 : int64 total_blks_hit;
963 : int64 total_blks_read;
964 : int64 total_blks_dirtied;
965 :
966 103712 : TimestampDifference(starttime, endtime, &secs_dur, &usecs_dur);
967 103712 : memset(&walusage, 0, sizeof(WalUsage));
968 103712 : WalUsageAccumDiff(&walusage, &pgWalUsage, &startwalusage);
969 103712 : memset(&bufferusage, 0, sizeof(BufferUsage));
970 103712 : BufferUsageAccumDiff(&bufferusage, &pgBufferUsage, &startbufferusage);
971 :
972 103712 : total_blks_hit = bufferusage.shared_blks_hit +
973 103712 : bufferusage.local_blks_hit;
974 103712 : total_blks_read = bufferusage.shared_blks_read +
975 103712 : bufferusage.local_blks_read;
976 103712 : total_blks_dirtied = bufferusage.shared_blks_dirtied +
977 103712 : bufferusage.local_blks_dirtied;
978 :
979 103712 : initStringInfo(&buf);
980 103712 : if (verbose)
981 : {
982 : /*
983 : * Aggressiveness already reported earlier, in dedicated
984 : * VACUUM VERBOSE ereport
985 : */
986 : Assert(!params->is_wraparound);
987 20 : msgfmt = _("finished vacuuming \"%s.%s.%s\": index scans: %d\n");
988 : }
989 103692 : else if (params->is_wraparound)
990 : {
991 : /*
992 : * While it's possible for a VACUUM to be both is_wraparound
993 : * and !aggressive, that's just a corner-case -- is_wraparound
994 : * implies aggressive. Produce distinct output for the corner
995 : * case all the same, just in case.
996 : */
997 103576 : if (vacrel->aggressive)
998 103554 : msgfmt = _("automatic aggressive vacuum to prevent wraparound of table \"%s.%s.%s\": index scans: %d\n");
999 : else
1000 22 : msgfmt = _("automatic vacuum to prevent wraparound of table \"%s.%s.%s\": index scans: %d\n");
1001 : }
1002 : else
1003 : {
1004 116 : if (vacrel->aggressive)
1005 28 : msgfmt = _("automatic aggressive vacuum of table \"%s.%s.%s\": index scans: %d\n");
1006 : else
1007 88 : msgfmt = _("automatic vacuum of table \"%s.%s.%s\": index scans: %d\n");
1008 : }
1009 103712 : appendStringInfo(&buf, msgfmt,
1010 : vacrel->dbname,
1011 : vacrel->relnamespace,
1012 : vacrel->relname,
1013 : vacrel->num_index_scans);
1014 144852 : appendStringInfo(&buf, _("pages: %u removed, %u remain, %u scanned (%.2f%% of total), %u eagerly scanned\n"),
1015 : vacrel->removed_pages,
1016 : new_rel_pages,
1017 : vacrel->scanned_pages,
1018 : orig_rel_pages == 0 ? 100.0 :
1019 41140 : 100.0 * vacrel->scanned_pages /
1020 : orig_rel_pages,
1021 : vacrel->eager_scanned_pages);
1022 103712 : appendStringInfo(&buf,
1023 103712 : _("tuples: %" PRId64 " removed, %" PRId64 " remain, %" PRId64 " are dead but not yet removable\n"),
1024 : vacrel->tuples_deleted,
1025 103712 : (int64) vacrel->new_rel_tuples,
1026 : vacrel->recently_dead_tuples);
1027 103712 : if (vacrel->missed_dead_tuples > 0)
1028 0 : appendStringInfo(&buf,
1029 0 : _("tuples missed: %" PRId64 " dead from %u pages not removed due to cleanup lock contention\n"),
1030 : vacrel->missed_dead_tuples,
1031 : vacrel->missed_dead_pages);
1032 103712 : diff = (int32) (ReadNextTransactionId() -
1033 103712 : vacrel->cutoffs.OldestXmin);
1034 103712 : appendStringInfo(&buf,
1035 103712 : _("removable cutoff: %u, which was %d XIDs old when operation ended\n"),
1036 : vacrel->cutoffs.OldestXmin, diff);
1037 103712 : if (frozenxid_updated)
1038 : {
1039 36062 : diff = (int32) (vacrel->NewRelfrozenXid -
1040 36062 : vacrel->cutoffs.relfrozenxid);
1041 36062 : appendStringInfo(&buf,
1042 36062 : _("new relfrozenxid: %u, which is %d XIDs ahead of previous value\n"),
1043 : vacrel->NewRelfrozenXid, diff);
1044 : }
1045 103712 : if (minmulti_updated)
1046 : {
1047 84 : diff = (int32) (vacrel->NewRelminMxid -
1048 84 : vacrel->cutoffs.relminmxid);
1049 84 : appendStringInfo(&buf,
1050 84 : _("new relminmxid: %u, which is %d MXIDs ahead of previous value\n"),
1051 : vacrel->NewRelminMxid, diff);
1052 : }
1053 144852 : appendStringInfo(&buf, _("frozen: %u pages from table (%.2f%% of total) had %" PRId64 " tuples frozen\n"),
1054 : vacrel->new_frozen_tuple_pages,
1055 : orig_rel_pages == 0 ? 100.0 :
1056 41140 : 100.0 * vacrel->new_frozen_tuple_pages /
1057 : orig_rel_pages,
1058 : vacrel->tuples_frozen);
1059 :
1060 103712 : appendStringInfo(&buf,
1061 103712 : _("visibility map: %u pages set all-visible, %u pages set all-frozen (%u were all-visible)\n"),
1062 : vacrel->vm_new_visible_pages,
1063 103712 : vacrel->vm_new_visible_frozen_pages +
1064 103712 : vacrel->vm_new_frozen_pages,
1065 : vacrel->vm_new_frozen_pages);
1066 103712 : if (vacrel->do_index_vacuuming)
1067 : {
1068 82228 : if (vacrel->nindexes == 0 || vacrel->num_index_scans == 0)
1069 82146 : appendStringInfoString(&buf, _("index scan not needed: "));
1070 : else
1071 82 : appendStringInfoString(&buf, _("index scan needed: "));
1072 :
1073 82228 : msgfmt = _("%u pages from table (%.2f%% of total) had %" PRId64 " dead item identifiers removed\n");
1074 : }
1075 : else
1076 : {
1077 21484 : if (!VacuumFailsafeActive)
1078 2 : appendStringInfoString(&buf, _("index scan bypassed: "));
1079 : else
1080 21482 : appendStringInfoString(&buf, _("index scan bypassed by failsafe: "));
1081 :
1082 21484 : msgfmt = _("%u pages from table (%.2f%% of total) have %" PRId64 " dead item identifiers\n");
1083 : }
1084 144852 : appendStringInfo(&buf, msgfmt,
1085 : vacrel->lpdead_item_pages,
1086 : orig_rel_pages == 0 ? 100.0 :
1087 41140 : 100.0 * vacrel->lpdead_item_pages / orig_rel_pages,
1088 : vacrel->lpdead_items);
1089 259346 : for (int i = 0; i < vacrel->nindexes; i++)
1090 : {
1091 155634 : IndexBulkDeleteResult *istat = vacrel->indstats[i];
1092 :
1093 155634 : if (!istat)
1094 155458 : continue;
1095 :
1096 176 : appendStringInfo(&buf,
1097 176 : _("index \"%s\": pages: %u in total, %u newly deleted, %u currently deleted, %u reusable\n"),
1098 176 : indnames[i],
1099 : istat->num_pages,
1100 : istat->pages_newly_deleted,
1101 : istat->pages_deleted,
1102 : istat->pages_free);
1103 : }
1104 103712 : if (track_cost_delay_timing)
1105 : {
1106 : /*
1107 : * We bypass the changecount mechanism because this value is
1108 : * only updated by the calling process. We also rely on the
1109 : * above call to pgstat_progress_end_command() to not clear
1110 : * the st_progress_param array.
1111 : */
1112 0 : appendStringInfo(&buf, _("delay time: %.3f ms\n"),
1113 0 : (double) MyBEEntry->st_progress_param[PROGRESS_VACUUM_DELAY_TIME] / 1000000.0);
1114 : }
1115 103712 : if (track_io_timing)
1116 : {
1117 0 : double read_ms = (double) (pgStatBlockReadTime - startreadtime) / 1000;
1118 0 : double write_ms = (double) (pgStatBlockWriteTime - startwritetime) / 1000;
1119 :
1120 0 : appendStringInfo(&buf, _("I/O timings: read: %.3f ms, write: %.3f ms\n"),
1121 : read_ms, write_ms);
1122 : }
1123 103712 : if (secs_dur > 0 || usecs_dur > 0)
1124 : {
1125 103712 : read_rate = (double) BLCKSZ * total_blks_read /
1126 103712 : (1024 * 1024) / (secs_dur + usecs_dur / 1000000.0);
1127 103712 : write_rate = (double) BLCKSZ * total_blks_dirtied /
1128 103712 : (1024 * 1024) / (secs_dur + usecs_dur / 1000000.0);
1129 : }
1130 103712 : appendStringInfo(&buf, _("avg read rate: %.3f MB/s, avg write rate: %.3f MB/s\n"),
1131 : read_rate, write_rate);
1132 103712 : appendStringInfo(&buf,
1133 103712 : _("buffer usage: %" PRId64 " hits, %" PRId64 " reads, %" PRId64 " dirtied\n"),
1134 : total_blks_hit,
1135 : total_blks_read,
1136 : total_blks_dirtied);
1137 103712 : appendStringInfo(&buf,
1138 103712 : _("WAL usage: %" PRId64 " records, %" PRId64 " full page images, %" PRIu64 " bytes, %" PRId64 " buffers full\n"),
1139 : walusage.wal_records,
1140 : walusage.wal_fpi,
1141 : walusage.wal_bytes,
1142 : walusage.wal_buffers_full);
1143 103712 : appendStringInfo(&buf, _("system usage: %s"), pg_rusage_show(&ru0));
1144 :
1145 103712 : ereport(verbose ? INFO : LOG,
1146 : (errmsg_internal("%s", buf.data)));
1147 103712 : pfree(buf.data);
1148 : }
1149 : }
1150 :
1151 : /* Cleanup index statistics and index names */
1152 322936 : for (int i = 0; i < vacrel->nindexes; i++)
1153 : {
1154 193090 : if (vacrel->indstats[i])
1155 2512 : pfree(vacrel->indstats[i]);
1156 :
1157 193090 : if (instrument)
1158 155704 : pfree(indnames[i]);
1159 : }
1160 129846 : }
1161 :
1162 : /*
1163 : * lazy_scan_heap() -- workhorse function for VACUUM
1164 : *
1165 : * This routine prunes each page in the heap, and considers the need to
1166 : * freeze remaining tuples with storage (not including pages that can be
1167 : * skipped using the visibility map). Also performs related maintenance
1168 : * of the FSM and visibility map. These steps all take place during an
1169 : * initial pass over the target heap relation.
1170 : *
1171 : * Also invokes lazy_vacuum_all_indexes to vacuum indexes, which largely
1172 : * consists of deleting index tuples that point to LP_DEAD items left in
1173 : * heap pages following pruning. Earlier initial pass over the heap will
1174 : * have collected the TIDs whose index tuples need to be removed.
1175 : *
1176 : * Finally, invokes lazy_vacuum_heap_rel to vacuum heap pages, which
1177 : * largely consists of marking LP_DEAD items (from vacrel->dead_items)
1178 : * as LP_UNUSED. This has to happen in a second, final pass over the
1179 : * heap, to preserve a basic invariant that all index AMs rely on: no
1180 : * extant index tuple can ever be allowed to contain a TID that points to
1181 : * an LP_UNUSED line pointer in the heap. We must disallow premature
1182 : * recycling of line pointers to avoid index scans that get confused
1183 : * about which TID points to which tuple immediately after recycling.
1184 : * (Actually, this isn't a concern when target heap relation happens to
1185 : * have no indexes, which allows us to safely apply the one-pass strategy
1186 : * as an optimization).
1187 : *
1188 : * In practice we often have enough space to fit all TIDs, and so won't
1189 : * need to call lazy_vacuum more than once, after our initial pass over
1190 : * the heap has totally finished. Otherwise things are slightly more
1191 : * complicated: our "initial pass" over the heap applies only to those
1192 : * pages that were pruned before we needed to call lazy_vacuum, and our
1193 : * "final pass" over the heap only vacuums these same heap pages.
1194 : * However, we process indexes in full every time lazy_vacuum is called,
1195 : * which makes index processing very inefficient when memory is in short
1196 : * supply.
1197 : */
1198 : static void
1199 129846 : lazy_scan_heap(LVRelState *vacrel)
1200 : {
1201 : ReadStream *stream;
1202 129846 : BlockNumber rel_pages = vacrel->rel_pages,
1203 129846 : blkno = 0,
1204 129846 : next_fsm_block_to_vacuum = 0;
1205 129846 : BlockNumber orig_eager_scan_success_limit =
1206 : vacrel->eager_scan_remaining_successes; /* for logging */
1207 129846 : Buffer vmbuffer = InvalidBuffer;
1208 129846 : const int initprog_index[] = {
1209 : PROGRESS_VACUUM_PHASE,
1210 : PROGRESS_VACUUM_TOTAL_HEAP_BLKS,
1211 : PROGRESS_VACUUM_MAX_DEAD_TUPLE_BYTES
1212 : };
1213 : int64 initprog_val[3];
1214 :
1215 : /* Report that we're scanning the heap, advertising total # of blocks */
1216 129846 : initprog_val[0] = PROGRESS_VACUUM_PHASE_SCAN_HEAP;
1217 129846 : initprog_val[1] = rel_pages;
1218 129846 : initprog_val[2] = vacrel->dead_items_info->max_bytes;
1219 129846 : pgstat_progress_update_multi_param(3, initprog_index, initprog_val);
1220 :
1221 : /* Initialize for the first heap_vac_scan_next_block() call */
1222 129846 : vacrel->current_block = InvalidBlockNumber;
1223 129846 : vacrel->next_unskippable_block = InvalidBlockNumber;
1224 129846 : vacrel->next_unskippable_allvis = false;
1225 129846 : vacrel->next_unskippable_eager_scanned = false;
1226 129846 : vacrel->next_unskippable_vmbuffer = InvalidBuffer;
1227 :
1228 : /*
1229 : * Set up the read stream for vacuum's first pass through the heap.
1230 : *
1231 : * This could be made safe for READ_STREAM_USE_BATCHING, but only with
1232 : * explicit work in heap_vac_scan_next_block.
1233 : */
1234 129846 : stream = read_stream_begin_relation(READ_STREAM_MAINTENANCE,
1235 : vacrel->bstrategy,
1236 : vacrel->rel,
1237 : MAIN_FORKNUM,
1238 : heap_vac_scan_next_block,
1239 : vacrel,
1240 : sizeof(uint8));
1241 :
1242 : while (true)
1243 767182 : {
1244 : Buffer buf;
1245 : Page page;
1246 897028 : uint8 blk_info = 0;
1247 : bool has_lpdead_items;
1248 897028 : void *per_buffer_data = NULL;
1249 897028 : bool vm_page_frozen = false;
1250 897028 : bool got_cleanup_lock = false;
1251 :
1252 897028 : vacuum_delay_point(false);
1253 :
1254 : /*
1255 : * Regularly check if wraparound failsafe should trigger.
1256 : *
1257 : * There is a similar check inside lazy_vacuum_all_indexes(), but
1258 : * relfrozenxid might start to look dangerously old before we reach
1259 : * that point. This check also provides failsafe coverage for the
1260 : * one-pass strategy, and the two-pass strategy with the index_cleanup
1261 : * param set to 'off'.
1262 : */
1263 897028 : if (vacrel->scanned_pages > 0 &&
1264 767182 : vacrel->scanned_pages % FAILSAFE_EVERY_PAGES == 0)
1265 0 : lazy_check_wraparound_failsafe(vacrel);
1266 :
1267 : /*
1268 : * Consider if we definitely have enough space to process TIDs on page
1269 : * already. If we are close to overrunning the available space for
1270 : * dead_items TIDs, pause and do a cycle of vacuuming before we tackle
1271 : * this page. However, let's force at least one page-worth of tuples
1272 : * to be stored as to ensure we do at least some work when the memory
1273 : * configured is so low that we run out before storing anything.
1274 : */
1275 897028 : if (vacrel->dead_items_info->num_items > 0 &&
1276 44190 : TidStoreMemoryUsage(vacrel->dead_items) > vacrel->dead_items_info->max_bytes)
1277 : {
1278 : /*
1279 : * Before beginning index vacuuming, we release any pin we may
1280 : * hold on the visibility map page. This isn't necessary for
1281 : * correctness, but we do it anyway to avoid holding the pin
1282 : * across a lengthy, unrelated operation.
1283 : */
1284 0 : if (BufferIsValid(vmbuffer))
1285 : {
1286 0 : ReleaseBuffer(vmbuffer);
1287 0 : vmbuffer = InvalidBuffer;
1288 : }
1289 :
1290 : /* Perform a round of index and heap vacuuming */
1291 0 : vacrel->consider_bypass_optimization = false;
1292 0 : lazy_vacuum(vacrel);
1293 :
1294 : /*
1295 : * Vacuum the Free Space Map to make newly-freed space visible on
1296 : * upper-level FSM pages. Note that blkno is the previously
1297 : * processed block.
1298 : */
1299 0 : FreeSpaceMapVacuumRange(vacrel->rel, next_fsm_block_to_vacuum,
1300 : blkno + 1);
1301 0 : next_fsm_block_to_vacuum = blkno;
1302 :
1303 : /* Report that we are once again scanning the heap */
1304 0 : pgstat_progress_update_param(PROGRESS_VACUUM_PHASE,
1305 : PROGRESS_VACUUM_PHASE_SCAN_HEAP);
1306 : }
1307 :
1308 897028 : buf = read_stream_next_buffer(stream, &per_buffer_data);
1309 :
1310 : /* The relation is exhausted. */
1311 897028 : if (!BufferIsValid(buf))
1312 129846 : break;
1313 :
1314 767182 : blk_info = *((uint8 *) per_buffer_data);
1315 767182 : CheckBufferIsPinnedOnce(buf);
1316 767182 : page = BufferGetPage(buf);
1317 767182 : blkno = BufferGetBlockNumber(buf);
1318 :
1319 767182 : vacrel->scanned_pages++;
1320 767182 : if (blk_info & VAC_BLK_WAS_EAGER_SCANNED)
1321 0 : vacrel->eager_scanned_pages++;
1322 :
1323 : /* Report as block scanned, update error traceback information */
1324 767182 : pgstat_progress_update_param(PROGRESS_VACUUM_HEAP_BLKS_SCANNED, blkno);
1325 767182 : update_vacuum_error_info(vacrel, NULL, VACUUM_ERRCB_PHASE_SCAN_HEAP,
1326 : blkno, InvalidOffsetNumber);
1327 :
1328 : /*
1329 : * Pin the visibility map page in case we need to mark the page
1330 : * all-visible. In most cases this will be very cheap, because we'll
1331 : * already have the correct page pinned anyway.
1332 : */
1333 767182 : visibilitymap_pin(vacrel->rel, blkno, &vmbuffer);
1334 :
1335 : /*
1336 : * We need a buffer cleanup lock to prune HOT chains and defragment
1337 : * the page in lazy_scan_prune. But when it's not possible to acquire
1338 : * a cleanup lock right away, we may be able to settle for reduced
1339 : * processing using lazy_scan_noprune.
1340 : */
1341 767182 : got_cleanup_lock = ConditionalLockBufferForCleanup(buf);
1342 :
1343 767182 : if (!got_cleanup_lock)
1344 18 : LockBuffer(buf, BUFFER_LOCK_SHARE);
1345 :
1346 : /* Check for new or empty pages before lazy_scan_[no]prune call */
1347 767182 : if (lazy_scan_new_or_empty(vacrel, buf, blkno, page, !got_cleanup_lock,
1348 767182 : vmbuffer))
1349 : {
1350 : /* Processed as new/empty page (lock and pin released) */
1351 1960 : continue;
1352 : }
1353 :
1354 : /*
1355 : * If we didn't get the cleanup lock, we can still collect LP_DEAD
1356 : * items in the dead_items area for later vacuuming, count live and
1357 : * recently dead tuples for vacuum logging, and determine if this
1358 : * block could later be truncated. If we encounter any xid/mxids that
1359 : * require advancing the relfrozenxid/relminxid, we'll have to wait
1360 : * for a cleanup lock and call lazy_scan_prune().
1361 : */
1362 765222 : if (!got_cleanup_lock &&
1363 18 : !lazy_scan_noprune(vacrel, buf, blkno, page, &has_lpdead_items))
1364 : {
1365 : /*
1366 : * lazy_scan_noprune could not do all required processing. Wait
1367 : * for a cleanup lock, and call lazy_scan_prune in the usual way.
1368 : */
1369 : Assert(vacrel->aggressive);
1370 0 : LockBuffer(buf, BUFFER_LOCK_UNLOCK);
1371 0 : LockBufferForCleanup(buf);
1372 0 : got_cleanup_lock = true;
1373 : }
1374 :
1375 : /*
1376 : * If we have a cleanup lock, we must now prune, freeze, and count
1377 : * tuples. We may have acquired the cleanup lock originally, or we may
1378 : * have gone back and acquired it after lazy_scan_noprune() returned
1379 : * false. Either way, the page hasn't been processed yet.
1380 : *
1381 : * Like lazy_scan_noprune(), lazy_scan_prune() will count
1382 : * recently_dead_tuples and live tuples for vacuum logging, determine
1383 : * if the block can later be truncated, and accumulate the details of
1384 : * remaining LP_DEAD line pointers on the page into dead_items. These
1385 : * dead items include those pruned by lazy_scan_prune() as well as
1386 : * line pointers previously marked LP_DEAD.
1387 : */
1388 765222 : if (got_cleanup_lock)
1389 765204 : lazy_scan_prune(vacrel, buf, blkno, page,
1390 : vmbuffer,
1391 765204 : blk_info & VAC_BLK_ALL_VISIBLE_ACCORDING_TO_VM,
1392 : &has_lpdead_items, &vm_page_frozen);
1393 :
1394 : /*
1395 : * Count an eagerly scanned page as a failure or a success.
1396 : *
1397 : * Only lazy_scan_prune() freezes pages, so if we didn't get the
1398 : * cleanup lock, we won't have frozen the page. However, we only count
1399 : * pages that were too new to require freezing as eager freeze
1400 : * failures.
1401 : *
1402 : * We could gather more information from lazy_scan_noprune() about
1403 : * whether or not there were tuples with XIDs or MXIDs older than the
1404 : * FreezeLimit or MultiXactCutoff. However, for simplicity, we simply
1405 : * exclude pages skipped due to cleanup lock contention from eager
1406 : * freeze algorithm caps.
1407 : */
1408 765222 : if (got_cleanup_lock &&
1409 765204 : (blk_info & VAC_BLK_WAS_EAGER_SCANNED))
1410 : {
1411 : /* Aggressive vacuums do not eager scan. */
1412 : Assert(!vacrel->aggressive);
1413 :
1414 0 : if (vm_page_frozen)
1415 : {
1416 : Assert(vacrel->eager_scan_remaining_successes > 0);
1417 0 : vacrel->eager_scan_remaining_successes--;
1418 :
1419 0 : if (vacrel->eager_scan_remaining_successes == 0)
1420 : {
1421 : /*
1422 : * If we hit our success cap, permanently disable eager
1423 : * scanning by setting the other eager scan management
1424 : * fields to their disabled values.
1425 : */
1426 0 : vacrel->eager_scan_remaining_fails = 0;
1427 0 : vacrel->next_eager_scan_region_start = InvalidBlockNumber;
1428 0 : vacrel->eager_scan_max_fails_per_region = 0;
1429 :
1430 0 : ereport(vacrel->verbose ? INFO : DEBUG2,
1431 : (errmsg("disabling eager scanning after freezing %u eagerly scanned blocks of \"%s.%s.%s\"",
1432 : orig_eager_scan_success_limit,
1433 : vacrel->dbname, vacrel->relnamespace,
1434 : vacrel->relname)));
1435 : }
1436 : }
1437 : else
1438 : {
1439 : Assert(vacrel->eager_scan_remaining_fails > 0);
1440 0 : vacrel->eager_scan_remaining_fails--;
1441 : }
1442 : }
1443 :
1444 : /*
1445 : * Now drop the buffer lock and, potentially, update the FSM.
1446 : *
1447 : * Our goal is to update the freespace map the last time we touch the
1448 : * page. If we'll process a block in the second pass, we may free up
1449 : * additional space on the page, so it is better to update the FSM
1450 : * after the second pass. If the relation has no indexes, or if index
1451 : * vacuuming is disabled, there will be no second heap pass; if this
1452 : * particular page has no dead items, the second heap pass will not
1453 : * touch this page. So, in those cases, update the FSM now.
1454 : *
1455 : * Note: In corner cases, it's possible to miss updating the FSM
1456 : * entirely. If index vacuuming is currently enabled, we'll skip the
1457 : * FSM update now. But if failsafe mode is later activated, or there
1458 : * are so few dead tuples that index vacuuming is bypassed, there will
1459 : * also be no opportunity to update the FSM later, because we'll never
1460 : * revisit this page. Since updating the FSM is desirable but not
1461 : * absolutely required, that's OK.
1462 : */
1463 765222 : if (vacrel->nindexes == 0
1464 743164 : || !vacrel->do_index_vacuuming
1465 568386 : || !has_lpdead_items)
1466 742258 : {
1467 742258 : Size freespace = PageGetHeapFreeSpace(page);
1468 :
1469 742258 : UnlockReleaseBuffer(buf);
1470 742258 : RecordPageWithFreeSpace(vacrel->rel, blkno, freespace);
1471 :
1472 : /*
1473 : * Periodically perform FSM vacuuming to make newly-freed space
1474 : * visible on upper FSM pages. This is done after vacuuming if the
1475 : * table has indexes. There will only be newly-freed space if we
1476 : * held the cleanup lock and lazy_scan_prune() was called.
1477 : */
1478 742258 : if (got_cleanup_lock && vacrel->nindexes == 0 && has_lpdead_items &&
1479 0 : blkno - next_fsm_block_to_vacuum >= VACUUM_FSM_EVERY_PAGES)
1480 : {
1481 0 : FreeSpaceMapVacuumRange(vacrel->rel, next_fsm_block_to_vacuum,
1482 : blkno);
1483 0 : next_fsm_block_to_vacuum = blkno;
1484 : }
1485 : }
1486 : else
1487 22964 : UnlockReleaseBuffer(buf);
1488 : }
1489 :
1490 129846 : vacrel->blkno = InvalidBlockNumber;
1491 129846 : if (BufferIsValid(vmbuffer))
1492 52008 : ReleaseBuffer(vmbuffer);
1493 :
1494 : /*
1495 : * Report that everything is now scanned. We never skip scanning the last
1496 : * block in the relation, so we can pass rel_pages here.
1497 : */
1498 129846 : pgstat_progress_update_param(PROGRESS_VACUUM_HEAP_BLKS_SCANNED,
1499 : rel_pages);
1500 :
1501 : /* now we can compute the new value for pg_class.reltuples */
1502 259692 : vacrel->new_live_tuples = vac_estimate_reltuples(vacrel->rel, rel_pages,
1503 : vacrel->scanned_pages,
1504 129846 : vacrel->live_tuples);
1505 :
1506 : /*
1507 : * Also compute the total number of surviving heap entries. In the
1508 : * (unlikely) scenario that new_live_tuples is -1, take it as zero.
1509 : */
1510 129846 : vacrel->new_rel_tuples =
1511 129846 : Max(vacrel->new_live_tuples, 0) + vacrel->recently_dead_tuples +
1512 129846 : vacrel->missed_dead_tuples;
1513 :
1514 129846 : read_stream_end(stream);
1515 :
1516 : /*
1517 : * Do index vacuuming (call each index's ambulkdelete routine), then do
1518 : * related heap vacuuming
1519 : */
1520 129846 : if (vacrel->dead_items_info->num_items > 0)
1521 1200 : lazy_vacuum(vacrel);
1522 :
1523 : /*
1524 : * Vacuum the remainder of the Free Space Map. We must do this whether or
1525 : * not there were indexes, and whether or not we bypassed index vacuuming.
1526 : * We can pass rel_pages here because we never skip scanning the last
1527 : * block of the relation.
1528 : */
1529 129846 : if (rel_pages > next_fsm_block_to_vacuum)
1530 52008 : FreeSpaceMapVacuumRange(vacrel->rel, next_fsm_block_to_vacuum, rel_pages);
1531 :
1532 : /* report all blocks vacuumed */
1533 129846 : pgstat_progress_update_param(PROGRESS_VACUUM_HEAP_BLKS_VACUUMED, rel_pages);
1534 :
1535 : /* Do final index cleanup (call each index's amvacuumcleanup routine) */
1536 129846 : if (vacrel->nindexes > 0 && vacrel->do_index_cleanup)
1537 102466 : lazy_cleanup_all_indexes(vacrel);
1538 129846 : }
1539 :
1540 : /*
1541 : * heap_vac_scan_next_block() -- read stream callback to get the next block
1542 : * for vacuum to process
1543 : *
1544 : * Every time lazy_scan_heap() needs a new block to process during its first
1545 : * phase, it invokes read_stream_next_buffer() with a stream set up to call
1546 : * heap_vac_scan_next_block() to get the next block.
1547 : *
1548 : * heap_vac_scan_next_block() uses the visibility map, vacuum options, and
1549 : * various thresholds to skip blocks which do not need to be processed and
1550 : * returns the next block to process or InvalidBlockNumber if there are no
1551 : * remaining blocks.
1552 : *
1553 : * The visibility status of the next block to process and whether or not it
1554 : * was eager scanned is set in the per_buffer_data.
1555 : *
1556 : * callback_private_data contains a reference to the LVRelState, passed to the
1557 : * read stream API during stream setup. The LVRelState is an in/out parameter
1558 : * here (locally named `vacrel`). Vacuum options and information about the
1559 : * relation are read from it. vacrel->skippedallvis is set if we skip a block
1560 : * that's all-visible but not all-frozen (to ensure that we don't update
1561 : * relfrozenxid in that case). vacrel also holds information about the next
1562 : * unskippable block -- as bookkeeping for this function.
1563 : */
1564 : static BlockNumber
1565 897028 : heap_vac_scan_next_block(ReadStream *stream,
1566 : void *callback_private_data,
1567 : void *per_buffer_data)
1568 : {
1569 : BlockNumber next_block;
1570 897028 : LVRelState *vacrel = callback_private_data;
1571 897028 : uint8 blk_info = 0;
1572 :
1573 : /* relies on InvalidBlockNumber + 1 overflowing to 0 on first call */
1574 897028 : next_block = vacrel->current_block + 1;
1575 :
1576 : /* Have we reached the end of the relation? */
1577 897028 : if (next_block >= vacrel->rel_pages)
1578 : {
1579 129846 : if (BufferIsValid(vacrel->next_unskippable_vmbuffer))
1580 : {
1581 49148 : ReleaseBuffer(vacrel->next_unskippable_vmbuffer);
1582 49148 : vacrel->next_unskippable_vmbuffer = InvalidBuffer;
1583 : }
1584 129846 : return InvalidBlockNumber;
1585 : }
1586 :
1587 : /*
1588 : * We must be in one of the three following states:
1589 : */
1590 767182 : if (next_block > vacrel->next_unskippable_block ||
1591 200114 : vacrel->next_unskippable_block == InvalidBlockNumber)
1592 : {
1593 : /*
1594 : * 1. We have just processed an unskippable block (or we're at the
1595 : * beginning of the scan). Find the next unskippable block using the
1596 : * visibility map.
1597 : */
1598 : bool skipsallvis;
1599 :
1600 619076 : find_next_unskippable_block(vacrel, &skipsallvis);
1601 :
1602 : /*
1603 : * We now know the next block that we must process. It can be the
1604 : * next block after the one we just processed, or something further
1605 : * ahead. If it's further ahead, we can jump to it, but we choose to
1606 : * do so only if we can skip at least SKIP_PAGES_THRESHOLD consecutive
1607 : * pages. Since we're reading sequentially, the OS should be doing
1608 : * readahead for us, so there's no gain in skipping a page now and
1609 : * then. Skipping such a range might even discourage sequential
1610 : * detection.
1611 : *
1612 : * This test also enables more frequent relfrozenxid advancement
1613 : * during non-aggressive VACUUMs. If the range has any all-visible
1614 : * pages then skipping makes updating relfrozenxid unsafe, which is a
1615 : * real downside.
1616 : */
1617 619076 : if (vacrel->next_unskippable_block - next_block >= SKIP_PAGES_THRESHOLD)
1618 : {
1619 4466 : next_block = vacrel->next_unskippable_block;
1620 4466 : if (skipsallvis)
1621 50 : vacrel->skippedallvis = true;
1622 : }
1623 : }
1624 :
1625 : /* Now we must be in one of the two remaining states: */
1626 767182 : if (next_block < vacrel->next_unskippable_block)
1627 : {
1628 : /*
1629 : * 2. We are processing a range of blocks that we could have skipped
1630 : * but chose not to. We know that they are all-visible in the VM,
1631 : * otherwise they would've been unskippable.
1632 : */
1633 148106 : vacrel->current_block = next_block;
1634 148106 : blk_info |= VAC_BLK_ALL_VISIBLE_ACCORDING_TO_VM;
1635 148106 : *((uint8 *) per_buffer_data) = blk_info;
1636 148106 : return vacrel->current_block;
1637 : }
1638 : else
1639 : {
1640 : /*
1641 : * 3. We reached the next unskippable block. Process it. On next
1642 : * iteration, we will be back in state 1.
1643 : */
1644 : Assert(next_block == vacrel->next_unskippable_block);
1645 :
1646 619076 : vacrel->current_block = next_block;
1647 619076 : if (vacrel->next_unskippable_allvis)
1648 44860 : blk_info |= VAC_BLK_ALL_VISIBLE_ACCORDING_TO_VM;
1649 619076 : if (vacrel->next_unskippable_eager_scanned)
1650 0 : blk_info |= VAC_BLK_WAS_EAGER_SCANNED;
1651 619076 : *((uint8 *) per_buffer_data) = blk_info;
1652 619076 : return vacrel->current_block;
1653 : }
1654 : }
1655 :
1656 : /*
1657 : * Find the next unskippable block in a vacuum scan using the visibility map.
1658 : * The next unskippable block and its visibility information is updated in
1659 : * vacrel.
1660 : *
1661 : * Note: our opinion of which blocks can be skipped can go stale immediately.
1662 : * It's okay if caller "misses" a page whose all-visible or all-frozen marking
1663 : * was concurrently cleared, though. All that matters is that caller scan all
1664 : * pages whose tuples might contain XIDs < OldestXmin, or MXIDs < OldestMxact.
1665 : * (Actually, non-aggressive VACUUMs can choose to skip all-visible pages with
1666 : * older XIDs/MXIDs. The *skippedallvis flag will be set here when the choice
1667 : * to skip such a range is actually made, making everything safe.)
1668 : */
1669 : static void
1670 619076 : find_next_unskippable_block(LVRelState *vacrel, bool *skipsallvis)
1671 : {
1672 619076 : BlockNumber rel_pages = vacrel->rel_pages;
1673 619076 : BlockNumber next_unskippable_block = vacrel->next_unskippable_block + 1;
1674 619076 : Buffer next_unskippable_vmbuffer = vacrel->next_unskippable_vmbuffer;
1675 619076 : bool next_unskippable_eager_scanned = false;
1676 : bool next_unskippable_allvis;
1677 :
1678 619076 : *skipsallvis = false;
1679 :
1680 492448 : for (;; next_unskippable_block++)
1681 492448 : {
1682 1111524 : uint8 mapbits = visibilitymap_get_status(vacrel->rel,
1683 : next_unskippable_block,
1684 : &next_unskippable_vmbuffer);
1685 :
1686 1111524 : next_unskippable_allvis = (mapbits & VISIBILITYMAP_ALL_VISIBLE) != 0;
1687 :
1688 : /*
1689 : * At the start of each eager scan region, normal vacuums with eager
1690 : * scanning enabled reset the failure counter, allowing vacuum to
1691 : * resume eager scanning if it had been suspended in the previous
1692 : * region.
1693 : */
1694 1111524 : if (next_unskippable_block >= vacrel->next_eager_scan_region_start)
1695 : {
1696 0 : vacrel->eager_scan_remaining_fails =
1697 0 : vacrel->eager_scan_max_fails_per_region;
1698 0 : vacrel->next_eager_scan_region_start += EAGER_SCAN_REGION_SIZE;
1699 : }
1700 :
1701 : /*
1702 : * A block is unskippable if it is not all visible according to the
1703 : * visibility map.
1704 : */
1705 1111524 : if (!next_unskippable_allvis)
1706 : {
1707 : Assert((mapbits & VISIBILITYMAP_ALL_FROZEN) == 0);
1708 574216 : break;
1709 : }
1710 :
1711 : /*
1712 : * Caller must scan the last page to determine whether it has tuples
1713 : * (caller must have the opportunity to set vacrel->nonempty_pages).
1714 : * This rule avoids having lazy_truncate_heap() take access-exclusive
1715 : * lock on rel to attempt a truncation that fails anyway, just because
1716 : * there are tuples on the last page (it is likely that there will be
1717 : * tuples on other nearby pages as well, but those can be skipped).
1718 : *
1719 : * Implement this by always treating the last block as unsafe to skip.
1720 : */
1721 537308 : if (next_unskippable_block == rel_pages - 1)
1722 44062 : break;
1723 :
1724 : /* DISABLE_PAGE_SKIPPING makes all skipping unsafe */
1725 493246 : if (!vacrel->skipwithvm)
1726 798 : break;
1727 :
1728 : /*
1729 : * All-frozen pages cannot contain XIDs < OldestXmin (XIDs that aren't
1730 : * already frozen by now), so this page can be skipped.
1731 : */
1732 492448 : if ((mapbits & VISIBILITYMAP_ALL_FROZEN) != 0)
1733 487502 : continue;
1734 :
1735 : /*
1736 : * Aggressive vacuums cannot skip any all-visible pages that are not
1737 : * also all-frozen.
1738 : */
1739 4946 : if (vacrel->aggressive)
1740 0 : break;
1741 :
1742 : /*
1743 : * Normal vacuums with eager scanning enabled only skip all-visible
1744 : * but not all-frozen pages if they have hit the failure limit for the
1745 : * current eager scan region.
1746 : */
1747 4946 : if (vacrel->eager_scan_remaining_fails > 0)
1748 : {
1749 0 : next_unskippable_eager_scanned = true;
1750 0 : break;
1751 : }
1752 :
1753 : /*
1754 : * All-visible blocks are safe to skip in a normal vacuum. But
1755 : * remember that the final range contains such a block for later.
1756 : */
1757 4946 : *skipsallvis = true;
1758 : }
1759 :
1760 : /* write the local variables back to vacrel */
1761 619076 : vacrel->next_unskippable_block = next_unskippable_block;
1762 619076 : vacrel->next_unskippable_allvis = next_unskippable_allvis;
1763 619076 : vacrel->next_unskippable_eager_scanned = next_unskippable_eager_scanned;
1764 619076 : vacrel->next_unskippable_vmbuffer = next_unskippable_vmbuffer;
1765 619076 : }
1766 :
1767 : /*
1768 : * lazy_scan_new_or_empty() -- lazy_scan_heap() new/empty page handling.
1769 : *
1770 : * Must call here to handle both new and empty pages before calling
1771 : * lazy_scan_prune or lazy_scan_noprune, since they're not prepared to deal
1772 : * with new or empty pages.
1773 : *
1774 : * It's necessary to consider new pages as a special case, since the rules for
1775 : * maintaining the visibility map and FSM with empty pages are a little
1776 : * different (though new pages can be truncated away during rel truncation).
1777 : *
1778 : * Empty pages are not really a special case -- they're just heap pages that
1779 : * have no allocated tuples (including even LP_UNUSED items). You might
1780 : * wonder why we need to handle them here all the same. It's only necessary
1781 : * because of a corner-case involving a hard crash during heap relation
1782 : * extension. If we ever make relation-extension crash safe, then it should
1783 : * no longer be necessary to deal with empty pages here (or new pages, for
1784 : * that matter).
1785 : *
1786 : * Caller must hold at least a shared lock. We might need to escalate the
1787 : * lock in that case, so the type of lock caller holds needs to be specified
1788 : * using 'sharelock' argument.
1789 : *
1790 : * Returns false in common case where caller should go on to call
1791 : * lazy_scan_prune (or lazy_scan_noprune). Otherwise returns true, indicating
1792 : * that lazy_scan_heap is done processing the page, releasing lock on caller's
1793 : * behalf.
1794 : *
1795 : * No vm_page_frozen output parameter (like that passed to lazy_scan_prune())
1796 : * is passed here because neither empty nor new pages can be eagerly frozen.
1797 : * New pages are never frozen. Empty pages are always set frozen in the VM at
1798 : * the same time that they are set all-visible, and we don't eagerly scan
1799 : * frozen pages.
1800 : */
1801 : static bool
1802 767182 : lazy_scan_new_or_empty(LVRelState *vacrel, Buffer buf, BlockNumber blkno,
1803 : Page page, bool sharelock, Buffer vmbuffer)
1804 : {
1805 : Size freespace;
1806 :
1807 767182 : if (PageIsNew(page))
1808 : {
1809 : /*
1810 : * All-zeroes pages can be left over if either a backend extends the
1811 : * relation by a single page, but crashes before the newly initialized
1812 : * page has been written out, or when bulk-extending the relation
1813 : * (which creates a number of empty pages at the tail end of the
1814 : * relation), and then enters them into the FSM.
1815 : *
1816 : * Note we do not enter the page into the visibilitymap. That has the
1817 : * downside that we repeatedly visit this page in subsequent vacuums,
1818 : * but otherwise we'll never discover the space on a promoted standby.
1819 : * The harm of repeated checking ought to normally not be too bad. The
1820 : * space usually should be used at some point, otherwise there
1821 : * wouldn't be any regular vacuums.
1822 : *
1823 : * Make sure these pages are in the FSM, to ensure they can be reused.
1824 : * Do that by testing if there's any space recorded for the page. If
1825 : * not, enter it. We do so after releasing the lock on the heap page,
1826 : * the FSM is approximate, after all.
1827 : */
1828 1906 : UnlockReleaseBuffer(buf);
1829 :
1830 1906 : if (GetRecordedFreeSpace(vacrel->rel, blkno) == 0)
1831 : {
1832 950 : freespace = BLCKSZ - SizeOfPageHeaderData;
1833 :
1834 950 : RecordPageWithFreeSpace(vacrel->rel, blkno, freespace);
1835 : }
1836 :
1837 1906 : return true;
1838 : }
1839 :
1840 765276 : if (PageIsEmpty(page))
1841 : {
1842 : /*
1843 : * It seems likely that caller will always be able to get a cleanup
1844 : * lock on an empty page. But don't take any chances -- escalate to
1845 : * an exclusive lock (still don't need a cleanup lock, though).
1846 : */
1847 54 : if (sharelock)
1848 : {
1849 0 : LockBuffer(buf, BUFFER_LOCK_UNLOCK);
1850 0 : LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
1851 :
1852 0 : if (!PageIsEmpty(page))
1853 : {
1854 : /* page isn't new or empty -- keep lock and pin for now */
1855 0 : return false;
1856 : }
1857 : }
1858 : else
1859 : {
1860 : /* Already have a full cleanup lock (which is more than enough) */
1861 : }
1862 :
1863 : /*
1864 : * Unlike new pages, empty pages are always set all-visible and
1865 : * all-frozen.
1866 : */
1867 54 : if (!PageIsAllVisible(page))
1868 : {
1869 : uint8 old_vmbits;
1870 :
1871 0 : START_CRIT_SECTION();
1872 :
1873 : /* mark buffer dirty before writing a WAL record */
1874 0 : MarkBufferDirty(buf);
1875 :
1876 : /*
1877 : * It's possible that another backend has extended the heap,
1878 : * initialized the page, and then failed to WAL-log the page due
1879 : * to an ERROR. Since heap extension is not WAL-logged, recovery
1880 : * might try to replay our record setting the page all-visible and
1881 : * find that the page isn't initialized, which will cause a PANIC.
1882 : * To prevent that, check whether the page has been previously
1883 : * WAL-logged, and if not, do that now.
1884 : */
1885 0 : if (RelationNeedsWAL(vacrel->rel) &&
1886 0 : PageGetLSN(page) == InvalidXLogRecPtr)
1887 0 : log_newpage_buffer(buf, true);
1888 :
1889 0 : PageSetAllVisible(page);
1890 0 : old_vmbits = visibilitymap_set(vacrel->rel, blkno, buf,
1891 : InvalidXLogRecPtr,
1892 : vmbuffer, InvalidTransactionId,
1893 : VISIBILITYMAP_ALL_VISIBLE |
1894 : VISIBILITYMAP_ALL_FROZEN);
1895 0 : END_CRIT_SECTION();
1896 :
1897 : /*
1898 : * If the page wasn't already set all-visible and/or all-frozen in
1899 : * the VM, count it as newly set for logging.
1900 : */
1901 0 : if ((old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
1902 : {
1903 0 : vacrel->vm_new_visible_pages++;
1904 0 : vacrel->vm_new_visible_frozen_pages++;
1905 : }
1906 0 : else if ((old_vmbits & VISIBILITYMAP_ALL_FROZEN) == 0)
1907 0 : vacrel->vm_new_frozen_pages++;
1908 : }
1909 :
1910 54 : freespace = PageGetHeapFreeSpace(page);
1911 54 : UnlockReleaseBuffer(buf);
1912 54 : RecordPageWithFreeSpace(vacrel->rel, blkno, freespace);
1913 54 : return true;
1914 : }
1915 :
1916 : /* page isn't new or empty -- keep lock and pin */
1917 765222 : return false;
1918 : }
1919 :
1920 : /* qsort comparator for sorting OffsetNumbers */
1921 : static int
1922 5321426 : cmpOffsetNumbers(const void *a, const void *b)
1923 : {
1924 5321426 : return pg_cmp_u16(*(const OffsetNumber *) a, *(const OffsetNumber *) b);
1925 : }
1926 :
1927 : /*
1928 : * lazy_scan_prune() -- lazy_scan_heap() pruning and freezing.
1929 : *
1930 : * Caller must hold pin and buffer cleanup lock on the buffer.
1931 : *
1932 : * vmbuffer is the buffer containing the VM block with visibility information
1933 : * for the heap block, blkno. all_visible_according_to_vm is the saved
1934 : * visibility status of the heap block looked up earlier by the caller. We
1935 : * won't rely entirely on this status, as it may be out of date.
1936 : *
1937 : * *has_lpdead_items is set to true or false depending on whether, upon return
1938 : * from this function, any LP_DEAD items are still present on the page.
1939 : *
1940 : * *vm_page_frozen is set to true if the page is newly set all-frozen in the
1941 : * VM. The caller currently only uses this for determining whether an eagerly
1942 : * scanned page was successfully set all-frozen.
1943 : */
1944 : static void
1945 765204 : lazy_scan_prune(LVRelState *vacrel,
1946 : Buffer buf,
1947 : BlockNumber blkno,
1948 : Page page,
1949 : Buffer vmbuffer,
1950 : bool all_visible_according_to_vm,
1951 : bool *has_lpdead_items,
1952 : bool *vm_page_frozen)
1953 : {
1954 765204 : Relation rel = vacrel->rel;
1955 : PruneFreezeResult presult;
1956 765204 : int prune_options = 0;
1957 :
1958 : Assert(BufferGetBlockNumber(buf) == blkno);
1959 :
1960 : /*
1961 : * Prune all HOT-update chains and potentially freeze tuples on this page.
1962 : *
1963 : * If the relation has no indexes, we can immediately mark would-be dead
1964 : * items LP_UNUSED.
1965 : *
1966 : * The number of tuples removed from the page is returned in
1967 : * presult.ndeleted. It should not be confused with presult.lpdead_items;
1968 : * presult.lpdead_items's final value can be thought of as the number of
1969 : * tuples that were deleted from indexes.
1970 : *
1971 : * We will update the VM after collecting LP_DEAD items and freezing
1972 : * tuples. Pruning will have determined whether or not the page is
1973 : * all-visible.
1974 : */
1975 765204 : prune_options = HEAP_PAGE_PRUNE_FREEZE;
1976 765204 : if (vacrel->nindexes == 0)
1977 22058 : prune_options |= HEAP_PAGE_PRUNE_MARK_UNUSED_NOW;
1978 :
1979 765204 : heap_page_prune_and_freeze(rel, buf, vacrel->vistest, prune_options,
1980 : &vacrel->cutoffs, &presult, PRUNE_VACUUM_SCAN,
1981 : &vacrel->offnum,
1982 : &vacrel->NewRelfrozenXid, &vacrel->NewRelminMxid);
1983 :
1984 : Assert(MultiXactIdIsValid(vacrel->NewRelminMxid));
1985 : Assert(TransactionIdIsValid(vacrel->NewRelfrozenXid));
1986 :
1987 765204 : if (presult.nfrozen > 0)
1988 : {
1989 : /*
1990 : * We don't increment the new_frozen_tuple_pages instrumentation
1991 : * counter when nfrozen == 0, since it only counts pages with newly
1992 : * frozen tuples (don't confuse that with pages newly set all-frozen
1993 : * in VM).
1994 : */
1995 42660 : vacrel->new_frozen_tuple_pages++;
1996 : }
1997 :
1998 : /*
1999 : * VACUUM will call heap_page_is_all_visible() during the second pass over
2000 : * the heap to determine all_visible and all_frozen for the page -- this
2001 : * is a specialized version of the logic from this function. Now that
2002 : * we've finished pruning and freezing, make sure that we're in total
2003 : * agreement with heap_page_is_all_visible() using an assertion.
2004 : */
2005 : #ifdef USE_ASSERT_CHECKING
2006 : /* Note that all_frozen value does not matter when !all_visible */
2007 : if (presult.all_visible)
2008 : {
2009 : TransactionId debug_cutoff;
2010 : bool debug_all_frozen;
2011 :
2012 : Assert(presult.lpdead_items == 0);
2013 :
2014 : if (!heap_page_is_all_visible(vacrel, buf,
2015 : &debug_cutoff, &debug_all_frozen))
2016 : Assert(false);
2017 :
2018 : Assert(presult.all_frozen == debug_all_frozen);
2019 :
2020 : Assert(!TransactionIdIsValid(debug_cutoff) ||
2021 : debug_cutoff == presult.vm_conflict_horizon);
2022 : }
2023 : #endif
2024 :
2025 : /*
2026 : * Now save details of the LP_DEAD items from the page in vacrel
2027 : */
2028 765204 : if (presult.lpdead_items > 0)
2029 : {
2030 27578 : vacrel->lpdead_item_pages++;
2031 :
2032 : /*
2033 : * deadoffsets are collected incrementally in
2034 : * heap_page_prune_and_freeze() as each dead line pointer is recorded,
2035 : * with an indeterminate order, but dead_items_add requires them to be
2036 : * sorted.
2037 : */
2038 27578 : qsort(presult.deadoffsets, presult.lpdead_items, sizeof(OffsetNumber),
2039 : cmpOffsetNumbers);
2040 :
2041 27578 : dead_items_add(vacrel, blkno, presult.deadoffsets, presult.lpdead_items);
2042 : }
2043 :
2044 : /* Finally, add page-local counts to whole-VACUUM counts */
2045 765204 : vacrel->tuples_deleted += presult.ndeleted;
2046 765204 : vacrel->tuples_frozen += presult.nfrozen;
2047 765204 : vacrel->lpdead_items += presult.lpdead_items;
2048 765204 : vacrel->live_tuples += presult.live_tuples;
2049 765204 : vacrel->recently_dead_tuples += presult.recently_dead_tuples;
2050 :
2051 : /* Can't truncate this page */
2052 765204 : if (presult.hastup)
2053 751456 : vacrel->nonempty_pages = blkno + 1;
2054 :
2055 : /* Did we find LP_DEAD items? */
2056 765204 : *has_lpdead_items = (presult.lpdead_items > 0);
2057 :
2058 : Assert(!presult.all_visible || !(*has_lpdead_items));
2059 :
2060 : /*
2061 : * Handle setting visibility map bit based on information from the VM (as
2062 : * of last heap_vac_scan_next_block() call), and from all_visible and
2063 : * all_frozen variables
2064 : */
2065 765204 : if (!all_visible_according_to_vm && presult.all_visible)
2066 69258 : {
2067 : uint8 old_vmbits;
2068 69258 : uint8 flags = VISIBILITYMAP_ALL_VISIBLE;
2069 :
2070 69258 : if (presult.all_frozen)
2071 : {
2072 : Assert(!TransactionIdIsValid(presult.vm_conflict_horizon));
2073 51360 : flags |= VISIBILITYMAP_ALL_FROZEN;
2074 : }
2075 :
2076 : /*
2077 : * It should never be the case that the visibility map page is set
2078 : * while the page-level bit is clear, but the reverse is allowed (if
2079 : * checksums are not enabled). Regardless, set both bits so that we
2080 : * get back in sync.
2081 : *
2082 : * NB: If the heap page is all-visible but the VM bit is not set, we
2083 : * don't need to dirty the heap page. However, if checksums are
2084 : * enabled, we do need to make sure that the heap page is dirtied
2085 : * before passing it to visibilitymap_set(), because it may be logged.
2086 : * Given that this situation should only happen in rare cases after a
2087 : * crash, it is not worth optimizing.
2088 : */
2089 69258 : PageSetAllVisible(page);
2090 69258 : MarkBufferDirty(buf);
2091 69258 : old_vmbits = visibilitymap_set(vacrel->rel, blkno, buf,
2092 : InvalidXLogRecPtr,
2093 : vmbuffer, presult.vm_conflict_horizon,
2094 : flags);
2095 :
2096 : /*
2097 : * If the page wasn't already set all-visible and/or all-frozen in the
2098 : * VM, count it as newly set for logging.
2099 : */
2100 69258 : if ((old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
2101 : {
2102 69258 : vacrel->vm_new_visible_pages++;
2103 69258 : if (presult.all_frozen)
2104 : {
2105 51360 : vacrel->vm_new_visible_frozen_pages++;
2106 51360 : *vm_page_frozen = true;
2107 : }
2108 : }
2109 0 : else if ((old_vmbits & VISIBILITYMAP_ALL_FROZEN) == 0 &&
2110 0 : presult.all_frozen)
2111 : {
2112 0 : vacrel->vm_new_frozen_pages++;
2113 0 : *vm_page_frozen = true;
2114 : }
2115 : }
2116 :
2117 : /*
2118 : * As of PostgreSQL 9.2, the visibility map bit should never be set if the
2119 : * page-level bit is clear. However, it's possible that the bit got
2120 : * cleared after heap_vac_scan_next_block() was called, so we must recheck
2121 : * with buffer lock before concluding that the VM is corrupt.
2122 : */
2123 695946 : else if (all_visible_according_to_vm && !PageIsAllVisible(page) &&
2124 0 : visibilitymap_get_status(vacrel->rel, blkno, &vmbuffer) != 0)
2125 : {
2126 0 : elog(WARNING, "page is not marked all-visible but visibility map bit is set in relation \"%s\" page %u",
2127 : vacrel->relname, blkno);
2128 0 : visibilitymap_clear(vacrel->rel, blkno, vmbuffer,
2129 : VISIBILITYMAP_VALID_BITS);
2130 : }
2131 :
2132 : /*
2133 : * It's possible for the value returned by
2134 : * GetOldestNonRemovableTransactionId() to move backwards, so it's not
2135 : * wrong for us to see tuples that appear to not be visible to everyone
2136 : * yet, while PD_ALL_VISIBLE is already set. The real safe xmin value
2137 : * never moves backwards, but GetOldestNonRemovableTransactionId() is
2138 : * conservative and sometimes returns a value that's unnecessarily small,
2139 : * so if we see that contradiction it just means that the tuples that we
2140 : * think are not visible to everyone yet actually are, and the
2141 : * PD_ALL_VISIBLE flag is correct.
2142 : *
2143 : * There should never be LP_DEAD items on a page with PD_ALL_VISIBLE set,
2144 : * however.
2145 : */
2146 695946 : else if (presult.lpdead_items > 0 && PageIsAllVisible(page))
2147 : {
2148 0 : elog(WARNING, "page containing LP_DEAD items is marked as all-visible in relation \"%s\" page %u",
2149 : vacrel->relname, blkno);
2150 0 : PageClearAllVisible(page);
2151 0 : MarkBufferDirty(buf);
2152 0 : visibilitymap_clear(vacrel->rel, blkno, vmbuffer,
2153 : VISIBILITYMAP_VALID_BITS);
2154 : }
2155 :
2156 : /*
2157 : * If the all-visible page is all-frozen but not marked as such yet, mark
2158 : * it as all-frozen. Note that all_frozen is only valid if all_visible is
2159 : * true, so we must check both all_visible and all_frozen.
2160 : */
2161 695946 : else if (all_visible_according_to_vm && presult.all_visible &&
2162 192904 : presult.all_frozen && !VM_ALL_FROZEN(vacrel->rel, blkno, &vmbuffer))
2163 : {
2164 : uint8 old_vmbits;
2165 :
2166 : /*
2167 : * Avoid relying on all_visible_according_to_vm as a proxy for the
2168 : * page-level PD_ALL_VISIBLE bit being set, since it might have become
2169 : * stale -- even when all_visible is set
2170 : */
2171 42 : if (!PageIsAllVisible(page))
2172 : {
2173 0 : PageSetAllVisible(page);
2174 0 : MarkBufferDirty(buf);
2175 : }
2176 :
2177 : /*
2178 : * Set the page all-frozen (and all-visible) in the VM.
2179 : *
2180 : * We can pass InvalidTransactionId as our cutoff_xid, since a
2181 : * snapshotConflictHorizon sufficient to make everything safe for REDO
2182 : * was logged when the page's tuples were frozen.
2183 : */
2184 : Assert(!TransactionIdIsValid(presult.vm_conflict_horizon));
2185 42 : old_vmbits = visibilitymap_set(vacrel->rel, blkno, buf,
2186 : InvalidXLogRecPtr,
2187 : vmbuffer, InvalidTransactionId,
2188 : VISIBILITYMAP_ALL_VISIBLE |
2189 : VISIBILITYMAP_ALL_FROZEN);
2190 :
2191 : /*
2192 : * The page was likely already set all-visible in the VM. However,
2193 : * there is a small chance that it was modified sometime between
2194 : * setting all_visible_according_to_vm and checking the visibility
2195 : * during pruning. Check the return value of old_vmbits anyway to
2196 : * ensure the visibility map counters used for logging are accurate.
2197 : */
2198 42 : if ((old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
2199 : {
2200 0 : vacrel->vm_new_visible_pages++;
2201 0 : vacrel->vm_new_visible_frozen_pages++;
2202 0 : *vm_page_frozen = true;
2203 : }
2204 :
2205 : /*
2206 : * We already checked that the page was not set all-frozen in the VM
2207 : * above, so we don't need to test the value of old_vmbits.
2208 : */
2209 : else
2210 : {
2211 42 : vacrel->vm_new_frozen_pages++;
2212 42 : *vm_page_frozen = true;
2213 : }
2214 : }
2215 765204 : }
2216 :
2217 : /*
2218 : * lazy_scan_noprune() -- lazy_scan_prune() without pruning or freezing
2219 : *
2220 : * Caller need only hold a pin and share lock on the buffer, unlike
2221 : * lazy_scan_prune, which requires a full cleanup lock. While pruning isn't
2222 : * performed here, it's quite possible that an earlier opportunistic pruning
2223 : * operation left LP_DEAD items behind. We'll at least collect any such items
2224 : * in dead_items for removal from indexes.
2225 : *
2226 : * For aggressive VACUUM callers, we may return false to indicate that a full
2227 : * cleanup lock is required for processing by lazy_scan_prune. This is only
2228 : * necessary when the aggressive VACUUM needs to freeze some tuple XIDs from
2229 : * one or more tuples on the page. We always return true for non-aggressive
2230 : * callers.
2231 : *
2232 : * If this function returns true, *has_lpdead_items gets set to true or false
2233 : * depending on whether, upon return from this function, any LP_DEAD items are
2234 : * present on the page. If this function returns false, *has_lpdead_items
2235 : * is not updated.
2236 : */
2237 : static bool
2238 18 : lazy_scan_noprune(LVRelState *vacrel,
2239 : Buffer buf,
2240 : BlockNumber blkno,
2241 : Page page,
2242 : bool *has_lpdead_items)
2243 : {
2244 : OffsetNumber offnum,
2245 : maxoff;
2246 : int lpdead_items,
2247 : live_tuples,
2248 : recently_dead_tuples,
2249 : missed_dead_tuples;
2250 : bool hastup;
2251 : HeapTupleHeader tupleheader;
2252 18 : TransactionId NoFreezePageRelfrozenXid = vacrel->NewRelfrozenXid;
2253 18 : MultiXactId NoFreezePageRelminMxid = vacrel->NewRelminMxid;
2254 : OffsetNumber deadoffsets[MaxHeapTuplesPerPage];
2255 :
2256 : Assert(BufferGetBlockNumber(buf) == blkno);
2257 :
2258 18 : hastup = false; /* for now */
2259 :
2260 18 : lpdead_items = 0;
2261 18 : live_tuples = 0;
2262 18 : recently_dead_tuples = 0;
2263 18 : missed_dead_tuples = 0;
2264 :
2265 18 : maxoff = PageGetMaxOffsetNumber(page);
2266 444 : for (offnum = FirstOffsetNumber;
2267 : offnum <= maxoff;
2268 426 : offnum = OffsetNumberNext(offnum))
2269 : {
2270 : ItemId itemid;
2271 : HeapTupleData tuple;
2272 :
2273 426 : vacrel->offnum = offnum;
2274 426 : itemid = PageGetItemId(page, offnum);
2275 :
2276 426 : if (!ItemIdIsUsed(itemid))
2277 38 : continue;
2278 :
2279 390 : if (ItemIdIsRedirected(itemid))
2280 : {
2281 2 : hastup = true;
2282 2 : continue;
2283 : }
2284 :
2285 388 : if (ItemIdIsDead(itemid))
2286 : {
2287 : /*
2288 : * Deliberately don't set hastup=true here. See same point in
2289 : * lazy_scan_prune for an explanation.
2290 : */
2291 0 : deadoffsets[lpdead_items++] = offnum;
2292 0 : continue;
2293 : }
2294 :
2295 388 : hastup = true; /* page prevents rel truncation */
2296 388 : tupleheader = (HeapTupleHeader) PageGetItem(page, itemid);
2297 388 : if (heap_tuple_should_freeze(tupleheader, &vacrel->cutoffs,
2298 : &NoFreezePageRelfrozenXid,
2299 : &NoFreezePageRelminMxid))
2300 : {
2301 : /* Tuple with XID < FreezeLimit (or MXID < MultiXactCutoff) */
2302 128 : if (vacrel->aggressive)
2303 : {
2304 : /*
2305 : * Aggressive VACUUMs must always be able to advance rel's
2306 : * relfrozenxid to a value >= FreezeLimit (and be able to
2307 : * advance rel's relminmxid to a value >= MultiXactCutoff).
2308 : * The ongoing aggressive VACUUM won't be able to do that
2309 : * unless it can freeze an XID (or MXID) from this tuple now.
2310 : *
2311 : * The only safe option is to have caller perform processing
2312 : * of this page using lazy_scan_prune. Caller might have to
2313 : * wait a while for a cleanup lock, but it can't be helped.
2314 : */
2315 0 : vacrel->offnum = InvalidOffsetNumber;
2316 0 : return false;
2317 : }
2318 :
2319 : /*
2320 : * Non-aggressive VACUUMs are under no obligation to advance
2321 : * relfrozenxid (even by one XID). We can be much laxer here.
2322 : *
2323 : * Currently we always just accept an older final relfrozenxid
2324 : * and/or relminmxid value. We never make caller wait or work a
2325 : * little harder, even when it likely makes sense to do so.
2326 : */
2327 : }
2328 :
2329 388 : ItemPointerSet(&(tuple.t_self), blkno, offnum);
2330 388 : tuple.t_data = (HeapTupleHeader) PageGetItem(page, itemid);
2331 388 : tuple.t_len = ItemIdGetLength(itemid);
2332 388 : tuple.t_tableOid = RelationGetRelid(vacrel->rel);
2333 :
2334 388 : switch (HeapTupleSatisfiesVacuum(&tuple, vacrel->cutoffs.OldestXmin,
2335 : buf))
2336 : {
2337 380 : case HEAPTUPLE_DELETE_IN_PROGRESS:
2338 : case HEAPTUPLE_LIVE:
2339 :
2340 : /*
2341 : * Count both cases as live, just like lazy_scan_prune
2342 : */
2343 380 : live_tuples++;
2344 :
2345 380 : break;
2346 4 : case HEAPTUPLE_DEAD:
2347 :
2348 : /*
2349 : * There is some useful work for pruning to do, that won't be
2350 : * done due to failure to get a cleanup lock.
2351 : */
2352 4 : missed_dead_tuples++;
2353 4 : break;
2354 4 : case HEAPTUPLE_RECENTLY_DEAD:
2355 :
2356 : /*
2357 : * Count in recently_dead_tuples, just like lazy_scan_prune
2358 : */
2359 4 : recently_dead_tuples++;
2360 4 : break;
2361 0 : case HEAPTUPLE_INSERT_IN_PROGRESS:
2362 :
2363 : /*
2364 : * Do not count these rows as live, just like lazy_scan_prune
2365 : */
2366 0 : break;
2367 0 : default:
2368 0 : elog(ERROR, "unexpected HeapTupleSatisfiesVacuum result");
2369 : break;
2370 : }
2371 : }
2372 :
2373 18 : vacrel->offnum = InvalidOffsetNumber;
2374 :
2375 : /*
2376 : * By here we know for sure that caller can put off freezing and pruning
2377 : * this particular page until the next VACUUM. Remember its details now.
2378 : * (lazy_scan_prune expects a clean slate, so we have to do this last.)
2379 : */
2380 18 : vacrel->NewRelfrozenXid = NoFreezePageRelfrozenXid;
2381 18 : vacrel->NewRelminMxid = NoFreezePageRelminMxid;
2382 :
2383 : /* Save any LP_DEAD items found on the page in dead_items */
2384 18 : if (vacrel->nindexes == 0)
2385 : {
2386 : /* Using one-pass strategy (since table has no indexes) */
2387 0 : if (lpdead_items > 0)
2388 : {
2389 : /*
2390 : * Perfunctory handling for the corner case where a single pass
2391 : * strategy VACUUM cannot get a cleanup lock, and it turns out
2392 : * that there is one or more LP_DEAD items: just count the LP_DEAD
2393 : * items as missed_dead_tuples instead. (This is a bit dishonest,
2394 : * but it beats having to maintain specialized heap vacuuming code
2395 : * forever, for vanishingly little benefit.)
2396 : */
2397 0 : hastup = true;
2398 0 : missed_dead_tuples += lpdead_items;
2399 : }
2400 : }
2401 18 : else if (lpdead_items > 0)
2402 : {
2403 : /*
2404 : * Page has LP_DEAD items, and so any references/TIDs that remain in
2405 : * indexes will be deleted during index vacuuming (and then marked
2406 : * LP_UNUSED in the heap)
2407 : */
2408 0 : vacrel->lpdead_item_pages++;
2409 :
2410 0 : dead_items_add(vacrel, blkno, deadoffsets, lpdead_items);
2411 :
2412 0 : vacrel->lpdead_items += lpdead_items;
2413 : }
2414 :
2415 : /*
2416 : * Finally, add relevant page-local counts to whole-VACUUM counts
2417 : */
2418 18 : vacrel->live_tuples += live_tuples;
2419 18 : vacrel->recently_dead_tuples += recently_dead_tuples;
2420 18 : vacrel->missed_dead_tuples += missed_dead_tuples;
2421 18 : if (missed_dead_tuples > 0)
2422 4 : vacrel->missed_dead_pages++;
2423 :
2424 : /* Can't truncate this page */
2425 18 : if (hastup)
2426 18 : vacrel->nonempty_pages = blkno + 1;
2427 :
2428 : /* Did we find LP_DEAD items? */
2429 18 : *has_lpdead_items = (lpdead_items > 0);
2430 :
2431 : /* Caller won't need to call lazy_scan_prune with same page */
2432 18 : return true;
2433 : }
2434 :
2435 : /*
2436 : * Main entry point for index vacuuming and heap vacuuming.
2437 : *
2438 : * Removes items collected in dead_items from table's indexes, then marks the
2439 : * same items LP_UNUSED in the heap. See the comments above lazy_scan_heap
2440 : * for full details.
2441 : *
2442 : * Also empties dead_items, freeing up space for later TIDs.
2443 : *
2444 : * We may choose to bypass index vacuuming at this point, though only when the
2445 : * ongoing VACUUM operation will definitely only have one index scan/round of
2446 : * index vacuuming.
2447 : */
2448 : static void
2449 1200 : lazy_vacuum(LVRelState *vacrel)
2450 : {
2451 : bool bypass;
2452 :
2453 : /* Should not end up here with no indexes */
2454 : Assert(vacrel->nindexes > 0);
2455 : Assert(vacrel->lpdead_item_pages > 0);
2456 :
2457 1200 : if (!vacrel->do_index_vacuuming)
2458 : {
2459 : Assert(!vacrel->do_index_cleanup);
2460 26 : dead_items_reset(vacrel);
2461 26 : return;
2462 : }
2463 :
2464 : /*
2465 : * Consider bypassing index vacuuming (and heap vacuuming) entirely.
2466 : *
2467 : * We currently only do this in cases where the number of LP_DEAD items
2468 : * for the entire VACUUM operation is close to zero. This avoids sharp
2469 : * discontinuities in the duration and overhead of successive VACUUM
2470 : * operations that run against the same table with a fixed workload.
2471 : * Ideally, successive VACUUM operations will behave as if there are
2472 : * exactly zero LP_DEAD items in cases where there are close to zero.
2473 : *
2474 : * This is likely to be helpful with a table that is continually affected
2475 : * by UPDATEs that can mostly apply the HOT optimization, but occasionally
2476 : * have small aberrations that lead to just a few heap pages retaining
2477 : * only one or two LP_DEAD items. This is pretty common; even when the
2478 : * DBA goes out of their way to make UPDATEs use HOT, it is practically
2479 : * impossible to predict whether HOT will be applied in 100% of cases.
2480 : * It's far easier to ensure that 99%+ of all UPDATEs against a table use
2481 : * HOT through careful tuning.
2482 : */
2483 1174 : bypass = false;
2484 1174 : if (vacrel->consider_bypass_optimization && vacrel->rel_pages > 0)
2485 : {
2486 : BlockNumber threshold;
2487 :
2488 : Assert(vacrel->num_index_scans == 0);
2489 : Assert(vacrel->lpdead_items == vacrel->dead_items_info->num_items);
2490 : Assert(vacrel->do_index_vacuuming);
2491 : Assert(vacrel->do_index_cleanup);
2492 :
2493 : /*
2494 : * This crossover point at which we'll start to do index vacuuming is
2495 : * expressed as a percentage of the total number of heap pages in the
2496 : * table that are known to have at least one LP_DEAD item. This is
2497 : * much more important than the total number of LP_DEAD items, since
2498 : * it's a proxy for the number of heap pages whose visibility map bits
2499 : * cannot be set on account of bypassing index and heap vacuuming.
2500 : *
2501 : * We apply one further precautionary test: the space currently used
2502 : * to store the TIDs (TIDs that now all point to LP_DEAD items) must
2503 : * not exceed 32MB. This limits the risk that we will bypass index
2504 : * vacuuming again and again until eventually there is a VACUUM whose
2505 : * dead_items space is not CPU cache resident.
2506 : *
2507 : * We don't take any special steps to remember the LP_DEAD items (such
2508 : * as counting them in our final update to the stats system) when the
2509 : * optimization is applied. Though the accounting used in analyze.c's
2510 : * acquire_sample_rows() will recognize the same LP_DEAD items as dead
2511 : * rows in its own stats report, that's okay. The discrepancy should
2512 : * be negligible. If this optimization is ever expanded to cover more
2513 : * cases then this may need to be reconsidered.
2514 : */
2515 1154 : threshold = (double) vacrel->rel_pages * BYPASS_THRESHOLD_PAGES;
2516 1156 : bypass = (vacrel->lpdead_item_pages < threshold &&
2517 2 : TidStoreMemoryUsage(vacrel->dead_items) < 32 * 1024 * 1024);
2518 : }
2519 :
2520 1174 : if (bypass)
2521 : {
2522 : /*
2523 : * There are almost zero TIDs. Behave as if there were precisely
2524 : * zero: bypass index vacuuming, but do index cleanup.
2525 : *
2526 : * We expect that the ongoing VACUUM operation will finish very
2527 : * quickly, so there is no point in considering speeding up as a
2528 : * failsafe against wraparound failure. (Index cleanup is expected to
2529 : * finish very quickly in cases where there were no ambulkdelete()
2530 : * calls.)
2531 : */
2532 2 : vacrel->do_index_vacuuming = false;
2533 : }
2534 1172 : else if (lazy_vacuum_all_indexes(vacrel))
2535 : {
2536 : /*
2537 : * We successfully completed a round of index vacuuming. Do related
2538 : * heap vacuuming now.
2539 : */
2540 1172 : lazy_vacuum_heap_rel(vacrel);
2541 : }
2542 : else
2543 : {
2544 : /*
2545 : * Failsafe case.
2546 : *
2547 : * We attempted index vacuuming, but didn't finish a full round/full
2548 : * index scan. This happens when relfrozenxid or relminmxid is too
2549 : * far in the past.
2550 : *
2551 : * From this point on the VACUUM operation will do no further index
2552 : * vacuuming or heap vacuuming. This VACUUM operation won't end up
2553 : * back here again.
2554 : */
2555 : Assert(VacuumFailsafeActive);
2556 : }
2557 :
2558 : /*
2559 : * Forget the LP_DEAD items that we just vacuumed (or just decided to not
2560 : * vacuum)
2561 : */
2562 1174 : dead_items_reset(vacrel);
2563 : }
2564 :
2565 : /*
2566 : * lazy_vacuum_all_indexes() -- Main entry for index vacuuming
2567 : *
2568 : * Returns true in the common case when all indexes were successfully
2569 : * vacuumed. Returns false in rare cases where we determined that the ongoing
2570 : * VACUUM operation is at risk of taking too long to finish, leading to
2571 : * wraparound failure.
2572 : */
2573 : static bool
2574 1172 : lazy_vacuum_all_indexes(LVRelState *vacrel)
2575 : {
2576 1172 : bool allindexes = true;
2577 1172 : double old_live_tuples = vacrel->rel->rd_rel->reltuples;
2578 1172 : const int progress_start_index[] = {
2579 : PROGRESS_VACUUM_PHASE,
2580 : PROGRESS_VACUUM_INDEXES_TOTAL
2581 : };
2582 1172 : const int progress_end_index[] = {
2583 : PROGRESS_VACUUM_INDEXES_TOTAL,
2584 : PROGRESS_VACUUM_INDEXES_PROCESSED,
2585 : PROGRESS_VACUUM_NUM_INDEX_VACUUMS
2586 : };
2587 : int64 progress_start_val[2];
2588 : int64 progress_end_val[3];
2589 :
2590 : Assert(vacrel->nindexes > 0);
2591 : Assert(vacrel->do_index_vacuuming);
2592 : Assert(vacrel->do_index_cleanup);
2593 :
2594 : /* Precheck for XID wraparound emergencies */
2595 1172 : if (lazy_check_wraparound_failsafe(vacrel))
2596 : {
2597 : /* Wraparound emergency -- don't even start an index scan */
2598 0 : return false;
2599 : }
2600 :
2601 : /*
2602 : * Report that we are now vacuuming indexes and the number of indexes to
2603 : * vacuum.
2604 : */
2605 1172 : progress_start_val[0] = PROGRESS_VACUUM_PHASE_VACUUM_INDEX;
2606 1172 : progress_start_val[1] = vacrel->nindexes;
2607 1172 : pgstat_progress_update_multi_param(2, progress_start_index, progress_start_val);
2608 :
2609 1172 : if (!ParallelVacuumIsActive(vacrel))
2610 : {
2611 3398 : for (int idx = 0; idx < vacrel->nindexes; idx++)
2612 : {
2613 2236 : Relation indrel = vacrel->indrels[idx];
2614 2236 : IndexBulkDeleteResult *istat = vacrel->indstats[idx];
2615 :
2616 2236 : vacrel->indstats[idx] = lazy_vacuum_one_index(indrel, istat,
2617 : old_live_tuples,
2618 : vacrel);
2619 :
2620 : /* Report the number of indexes vacuumed */
2621 2236 : pgstat_progress_update_param(PROGRESS_VACUUM_INDEXES_PROCESSED,
2622 2236 : idx + 1);
2623 :
2624 2236 : if (lazy_check_wraparound_failsafe(vacrel))
2625 : {
2626 : /* Wraparound emergency -- end current index scan */
2627 0 : allindexes = false;
2628 0 : break;
2629 : }
2630 : }
2631 : }
2632 : else
2633 : {
2634 : /* Outsource everything to parallel variant */
2635 10 : parallel_vacuum_bulkdel_all_indexes(vacrel->pvs, old_live_tuples,
2636 : vacrel->num_index_scans);
2637 :
2638 : /*
2639 : * Do a postcheck to consider applying wraparound failsafe now. Note
2640 : * that parallel VACUUM only gets the precheck and this postcheck.
2641 : */
2642 10 : if (lazy_check_wraparound_failsafe(vacrel))
2643 0 : allindexes = false;
2644 : }
2645 :
2646 : /*
2647 : * We delete all LP_DEAD items from the first heap pass in all indexes on
2648 : * each call here (except calls where we choose to do the failsafe). This
2649 : * makes the next call to lazy_vacuum_heap_rel() safe (except in the event
2650 : * of the failsafe triggering, which prevents the next call from taking
2651 : * place).
2652 : */
2653 : Assert(vacrel->num_index_scans > 0 ||
2654 : vacrel->dead_items_info->num_items == vacrel->lpdead_items);
2655 : Assert(allindexes || VacuumFailsafeActive);
2656 :
2657 : /*
2658 : * Increase and report the number of index scans. Also, we reset
2659 : * PROGRESS_VACUUM_INDEXES_TOTAL and PROGRESS_VACUUM_INDEXES_PROCESSED.
2660 : *
2661 : * We deliberately include the case where we started a round of bulk
2662 : * deletes that we weren't able to finish due to the failsafe triggering.
2663 : */
2664 1172 : vacrel->num_index_scans++;
2665 1172 : progress_end_val[0] = 0;
2666 1172 : progress_end_val[1] = 0;
2667 1172 : progress_end_val[2] = vacrel->num_index_scans;
2668 1172 : pgstat_progress_update_multi_param(3, progress_end_index, progress_end_val);
2669 :
2670 1172 : return allindexes;
2671 : }
2672 :
2673 : /*
2674 : * Read stream callback for vacuum's third phase (second pass over the heap).
2675 : * Gets the next block from the TID store and returns it or InvalidBlockNumber
2676 : * if there are no further blocks to vacuum.
2677 : *
2678 : * NB: Assumed to be safe to use with READ_STREAM_USE_BATCHING.
2679 : */
2680 : static BlockNumber
2681 24134 : vacuum_reap_lp_read_stream_next(ReadStream *stream,
2682 : void *callback_private_data,
2683 : void *per_buffer_data)
2684 : {
2685 24134 : TidStoreIter *iter = callback_private_data;
2686 : TidStoreIterResult *iter_result;
2687 :
2688 24134 : iter_result = TidStoreIterateNext(iter);
2689 24134 : if (iter_result == NULL)
2690 1172 : return InvalidBlockNumber;
2691 :
2692 : /*
2693 : * Save the TidStoreIterResult for later, so we can extract the offsets.
2694 : * It is safe to copy the result, according to TidStoreIterateNext().
2695 : */
2696 22962 : memcpy(per_buffer_data, iter_result, sizeof(*iter_result));
2697 :
2698 22962 : return iter_result->blkno;
2699 : }
2700 :
2701 : /*
2702 : * lazy_vacuum_heap_rel() -- second pass over the heap for two pass strategy
2703 : *
2704 : * This routine marks LP_DEAD items in vacrel->dead_items as LP_UNUSED. Pages
2705 : * that never had lazy_scan_prune record LP_DEAD items are not visited at all.
2706 : *
2707 : * We may also be able to truncate the line pointer array of the heap pages we
2708 : * visit. If there is a contiguous group of LP_UNUSED items at the end of the
2709 : * array, it can be reclaimed as free space. These LP_UNUSED items usually
2710 : * start out as LP_DEAD items recorded by lazy_scan_prune (we set items from
2711 : * each page to LP_UNUSED, and then consider if it's possible to truncate the
2712 : * page's line pointer array).
2713 : *
2714 : * Note: the reason for doing this as a second pass is we cannot remove the
2715 : * tuples until we've removed their index entries, and we want to process
2716 : * index entry removal in batches as large as possible.
2717 : */
2718 : static void
2719 1172 : lazy_vacuum_heap_rel(LVRelState *vacrel)
2720 : {
2721 : ReadStream *stream;
2722 1172 : BlockNumber vacuumed_pages = 0;
2723 1172 : Buffer vmbuffer = InvalidBuffer;
2724 : LVSavedErrInfo saved_err_info;
2725 : TidStoreIter *iter;
2726 :
2727 : Assert(vacrel->do_index_vacuuming);
2728 : Assert(vacrel->do_index_cleanup);
2729 : Assert(vacrel->num_index_scans > 0);
2730 :
2731 : /* Report that we are now vacuuming the heap */
2732 1172 : pgstat_progress_update_param(PROGRESS_VACUUM_PHASE,
2733 : PROGRESS_VACUUM_PHASE_VACUUM_HEAP);
2734 :
2735 : /* Update error traceback information */
2736 1172 : update_vacuum_error_info(vacrel, &saved_err_info,
2737 : VACUUM_ERRCB_PHASE_VACUUM_HEAP,
2738 : InvalidBlockNumber, InvalidOffsetNumber);
2739 :
2740 1172 : iter = TidStoreBeginIterate(vacrel->dead_items);
2741 :
2742 : /*
2743 : * Set up the read stream for vacuum's second pass through the heap.
2744 : *
2745 : * It is safe to use batchmode, as vacuum_reap_lp_read_stream_next() does
2746 : * not need to wait for IO and does not perform locking. Once we support
2747 : * parallelism it should still be fine, as presumably the holder of locks
2748 : * would never be blocked by IO while holding the lock.
2749 : */
2750 1172 : stream = read_stream_begin_relation(READ_STREAM_MAINTENANCE |
2751 : READ_STREAM_USE_BATCHING,
2752 : vacrel->bstrategy,
2753 : vacrel->rel,
2754 : MAIN_FORKNUM,
2755 : vacuum_reap_lp_read_stream_next,
2756 : iter,
2757 : sizeof(TidStoreIterResult));
2758 :
2759 : while (true)
2760 22962 : {
2761 : BlockNumber blkno;
2762 : Buffer buf;
2763 : Page page;
2764 : TidStoreIterResult *iter_result;
2765 : Size freespace;
2766 : OffsetNumber offsets[MaxOffsetNumber];
2767 : int num_offsets;
2768 :
2769 24134 : vacuum_delay_point(false);
2770 :
2771 24134 : buf = read_stream_next_buffer(stream, (void **) &iter_result);
2772 :
2773 : /* The relation is exhausted */
2774 24134 : if (!BufferIsValid(buf))
2775 1172 : break;
2776 :
2777 22962 : vacrel->blkno = blkno = BufferGetBlockNumber(buf);
2778 :
2779 : Assert(iter_result);
2780 22962 : num_offsets = TidStoreGetBlockOffsets(iter_result, offsets, lengthof(offsets));
2781 : Assert(num_offsets <= lengthof(offsets));
2782 :
2783 : /*
2784 : * Pin the visibility map page in case we need to mark the page
2785 : * all-visible. In most cases this will be very cheap, because we'll
2786 : * already have the correct page pinned anyway.
2787 : */
2788 22962 : visibilitymap_pin(vacrel->rel, blkno, &vmbuffer);
2789 :
2790 : /* We need a non-cleanup exclusive lock to mark dead_items unused */
2791 22962 : LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
2792 22962 : lazy_vacuum_heap_page(vacrel, blkno, buf, offsets,
2793 : num_offsets, vmbuffer);
2794 :
2795 : /* Now that we've vacuumed the page, record its available space */
2796 22962 : page = BufferGetPage(buf);
2797 22962 : freespace = PageGetHeapFreeSpace(page);
2798 :
2799 22962 : UnlockReleaseBuffer(buf);
2800 22962 : RecordPageWithFreeSpace(vacrel->rel, blkno, freespace);
2801 22962 : vacuumed_pages++;
2802 : }
2803 :
2804 1172 : read_stream_end(stream);
2805 1172 : TidStoreEndIterate(iter);
2806 :
2807 1172 : vacrel->blkno = InvalidBlockNumber;
2808 1172 : if (BufferIsValid(vmbuffer))
2809 1172 : ReleaseBuffer(vmbuffer);
2810 :
2811 : /*
2812 : * We set all LP_DEAD items from the first heap pass to LP_UNUSED during
2813 : * the second heap pass. No more, no less.
2814 : */
2815 : Assert(vacrel->num_index_scans > 1 ||
2816 : (vacrel->dead_items_info->num_items == vacrel->lpdead_items &&
2817 : vacuumed_pages == vacrel->lpdead_item_pages));
2818 :
2819 1172 : ereport(DEBUG2,
2820 : (errmsg("table \"%s\": removed %" PRId64 " dead item identifiers in %u pages",
2821 : vacrel->relname, vacrel->dead_items_info->num_items,
2822 : vacuumed_pages)));
2823 :
2824 : /* Revert to the previous phase information for error traceback */
2825 1172 : restore_vacuum_error_info(vacrel, &saved_err_info);
2826 1172 : }
2827 :
2828 : /*
2829 : * lazy_vacuum_heap_page() -- free page's LP_DEAD items listed in the
2830 : * vacrel->dead_items store.
2831 : *
2832 : * Caller must have an exclusive buffer lock on the buffer (though a full
2833 : * cleanup lock is also acceptable). vmbuffer must be valid and already have
2834 : * a pin on blkno's visibility map page.
2835 : */
2836 : static void
2837 22962 : lazy_vacuum_heap_page(LVRelState *vacrel, BlockNumber blkno, Buffer buffer,
2838 : OffsetNumber *deadoffsets, int num_offsets,
2839 : Buffer vmbuffer)
2840 : {
2841 22962 : Page page = BufferGetPage(buffer);
2842 : OffsetNumber unused[MaxHeapTuplesPerPage];
2843 22962 : int nunused = 0;
2844 : TransactionId visibility_cutoff_xid;
2845 : bool all_frozen;
2846 : LVSavedErrInfo saved_err_info;
2847 :
2848 : Assert(vacrel->do_index_vacuuming);
2849 :
2850 22962 : pgstat_progress_update_param(PROGRESS_VACUUM_HEAP_BLKS_VACUUMED, blkno);
2851 :
2852 : /* Update error traceback information */
2853 22962 : update_vacuum_error_info(vacrel, &saved_err_info,
2854 : VACUUM_ERRCB_PHASE_VACUUM_HEAP, blkno,
2855 : InvalidOffsetNumber);
2856 :
2857 22962 : START_CRIT_SECTION();
2858 :
2859 1509192 : for (int i = 0; i < num_offsets; i++)
2860 : {
2861 : ItemId itemid;
2862 1486230 : OffsetNumber toff = deadoffsets[i];
2863 :
2864 1486230 : itemid = PageGetItemId(page, toff);
2865 :
2866 : Assert(ItemIdIsDead(itemid) && !ItemIdHasStorage(itemid));
2867 1486230 : ItemIdSetUnused(itemid);
2868 1486230 : unused[nunused++] = toff;
2869 : }
2870 :
2871 : Assert(nunused > 0);
2872 :
2873 : /* Attempt to truncate line pointer array now */
2874 22962 : PageTruncateLinePointerArray(page);
2875 :
2876 : /*
2877 : * Mark buffer dirty before we write WAL.
2878 : */
2879 22962 : MarkBufferDirty(buffer);
2880 :
2881 : /* XLOG stuff */
2882 22962 : if (RelationNeedsWAL(vacrel->rel))
2883 : {
2884 21272 : log_heap_prune_and_freeze(vacrel->rel, buffer,
2885 : InvalidTransactionId,
2886 : false, /* no cleanup lock required */
2887 : PRUNE_VACUUM_CLEANUP,
2888 : NULL, 0, /* frozen */
2889 : NULL, 0, /* redirected */
2890 : NULL, 0, /* dead */
2891 : unused, nunused);
2892 : }
2893 :
2894 : /*
2895 : * End critical section, so we safely can do visibility tests (which
2896 : * possibly need to perform IO and allocate memory!). If we crash now the
2897 : * page (including the corresponding vm bit) might not be marked all
2898 : * visible, but that's fine. A later vacuum will fix that.
2899 : */
2900 22962 : END_CRIT_SECTION();
2901 :
2902 : /*
2903 : * Now that we have removed the LP_DEAD items from the page, once again
2904 : * check if the page has become all-visible. The page is already marked
2905 : * dirty, exclusively locked, and, if needed, a full page image has been
2906 : * emitted.
2907 : */
2908 : Assert(!PageIsAllVisible(page));
2909 22962 : if (heap_page_is_all_visible(vacrel, buffer, &visibility_cutoff_xid,
2910 : &all_frozen))
2911 : {
2912 : uint8 old_vmbits;
2913 22902 : uint8 flags = VISIBILITYMAP_ALL_VISIBLE;
2914 :
2915 22902 : if (all_frozen)
2916 : {
2917 : Assert(!TransactionIdIsValid(visibility_cutoff_xid));
2918 17502 : flags |= VISIBILITYMAP_ALL_FROZEN;
2919 : }
2920 :
2921 22902 : PageSetAllVisible(page);
2922 22902 : old_vmbits = visibilitymap_set(vacrel->rel, blkno, buffer,
2923 : InvalidXLogRecPtr,
2924 : vmbuffer, visibility_cutoff_xid,
2925 : flags);
2926 :
2927 : /*
2928 : * If the page wasn't already set all-visible and/or all-frozen in the
2929 : * VM, count it as newly set for logging.
2930 : */
2931 22902 : if ((old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
2932 : {
2933 22902 : vacrel->vm_new_visible_pages++;
2934 22902 : if (all_frozen)
2935 17502 : vacrel->vm_new_visible_frozen_pages++;
2936 : }
2937 :
2938 0 : else if ((old_vmbits & VISIBILITYMAP_ALL_FROZEN) == 0 &&
2939 : all_frozen)
2940 0 : vacrel->vm_new_frozen_pages++;
2941 : }
2942 :
2943 : /* Revert to the previous phase information for error traceback */
2944 22962 : restore_vacuum_error_info(vacrel, &saved_err_info);
2945 22962 : }
2946 :
2947 : /*
2948 : * Trigger the failsafe to avoid wraparound failure when vacrel table has a
2949 : * relfrozenxid and/or relminmxid that is dangerously far in the past.
2950 : * Triggering the failsafe makes the ongoing VACUUM bypass any further index
2951 : * vacuuming and heap vacuuming. Truncating the heap is also bypassed.
2952 : *
2953 : * Any remaining work (work that VACUUM cannot just bypass) is typically sped
2954 : * up when the failsafe triggers. VACUUM stops applying any cost-based delay
2955 : * that it started out with.
2956 : *
2957 : * Returns true when failsafe has been triggered.
2958 : */
2959 : static bool
2960 133264 : lazy_check_wraparound_failsafe(LVRelState *vacrel)
2961 : {
2962 : /* Don't warn more than once per VACUUM */
2963 133264 : if (VacuumFailsafeActive)
2964 0 : return true;
2965 :
2966 133264 : if (unlikely(vacuum_xid_failsafe_check(&vacrel->cutoffs)))
2967 : {
2968 21702 : const int progress_index[] = {
2969 : PROGRESS_VACUUM_INDEXES_TOTAL,
2970 : PROGRESS_VACUUM_INDEXES_PROCESSED
2971 : };
2972 21702 : int64 progress_val[2] = {0, 0};
2973 :
2974 21702 : VacuumFailsafeActive = true;
2975 :
2976 : /*
2977 : * Abandon use of a buffer access strategy to allow use of all of
2978 : * shared buffers. We assume the caller who allocated the memory for
2979 : * the BufferAccessStrategy will free it.
2980 : */
2981 21702 : vacrel->bstrategy = NULL;
2982 :
2983 : /* Disable index vacuuming, index cleanup, and heap rel truncation */
2984 21702 : vacrel->do_index_vacuuming = false;
2985 21702 : vacrel->do_index_cleanup = false;
2986 21702 : vacrel->do_rel_truncate = false;
2987 :
2988 : /* Reset the progress counters */
2989 21702 : pgstat_progress_update_multi_param(2, progress_index, progress_val);
2990 :
2991 21702 : ereport(WARNING,
2992 : (errmsg("bypassing nonessential maintenance of table \"%s.%s.%s\" as a failsafe after %d index scans",
2993 : vacrel->dbname, vacrel->relnamespace, vacrel->relname,
2994 : vacrel->num_index_scans),
2995 : errdetail("The table's relfrozenxid or relminmxid is too far in the past."),
2996 : errhint("Consider increasing configuration parameter \"maintenance_work_mem\" or \"autovacuum_work_mem\".\n"
2997 : "You might also need to consider other ways for VACUUM to keep up with the allocation of transaction IDs.")));
2998 :
2999 : /* Stop applying cost limits from this point on */
3000 21702 : VacuumCostActive = false;
3001 21702 : VacuumCostBalance = 0;
3002 :
3003 21702 : return true;
3004 : }
3005 :
3006 111562 : return false;
3007 : }
3008 :
3009 : /*
3010 : * lazy_cleanup_all_indexes() -- cleanup all indexes of relation.
3011 : */
3012 : static void
3013 102466 : lazy_cleanup_all_indexes(LVRelState *vacrel)
3014 : {
3015 102466 : double reltuples = vacrel->new_rel_tuples;
3016 102466 : bool estimated_count = vacrel->scanned_pages < vacrel->rel_pages;
3017 102466 : const int progress_start_index[] = {
3018 : PROGRESS_VACUUM_PHASE,
3019 : PROGRESS_VACUUM_INDEXES_TOTAL
3020 : };
3021 102466 : const int progress_end_index[] = {
3022 : PROGRESS_VACUUM_INDEXES_TOTAL,
3023 : PROGRESS_VACUUM_INDEXES_PROCESSED
3024 : };
3025 : int64 progress_start_val[2];
3026 102466 : int64 progress_end_val[2] = {0, 0};
3027 :
3028 : Assert(vacrel->do_index_cleanup);
3029 : Assert(vacrel->nindexes > 0);
3030 :
3031 : /*
3032 : * Report that we are now cleaning up indexes and the number of indexes to
3033 : * cleanup.
3034 : */
3035 102466 : progress_start_val[0] = PROGRESS_VACUUM_PHASE_INDEX_CLEANUP;
3036 102466 : progress_start_val[1] = vacrel->nindexes;
3037 102466 : pgstat_progress_update_multi_param(2, progress_start_index, progress_start_val);
3038 :
3039 102466 : if (!ParallelVacuumIsActive(vacrel))
3040 : {
3041 262582 : for (int idx = 0; idx < vacrel->nindexes; idx++)
3042 : {
3043 160150 : Relation indrel = vacrel->indrels[idx];
3044 160150 : IndexBulkDeleteResult *istat = vacrel->indstats[idx];
3045 :
3046 320300 : vacrel->indstats[idx] =
3047 160150 : lazy_cleanup_one_index(indrel, istat, reltuples,
3048 : estimated_count, vacrel);
3049 :
3050 : /* Report the number of indexes cleaned up */
3051 160150 : pgstat_progress_update_param(PROGRESS_VACUUM_INDEXES_PROCESSED,
3052 160150 : idx + 1);
3053 : }
3054 : }
3055 : else
3056 : {
3057 : /* Outsource everything to parallel variant */
3058 34 : parallel_vacuum_cleanup_all_indexes(vacrel->pvs, reltuples,
3059 : vacrel->num_index_scans,
3060 : estimated_count);
3061 : }
3062 :
3063 : /* Reset the progress counters */
3064 102466 : pgstat_progress_update_multi_param(2, progress_end_index, progress_end_val);
3065 102466 : }
3066 :
3067 : /*
3068 : * lazy_vacuum_one_index() -- vacuum index relation.
3069 : *
3070 : * Delete all the index tuples containing a TID collected in
3071 : * vacrel->dead_items. Also update running statistics. Exact
3072 : * details depend on index AM's ambulkdelete routine.
3073 : *
3074 : * reltuples is the number of heap tuples to be passed to the
3075 : * bulkdelete callback. It's always assumed to be estimated.
3076 : * See indexam.sgml for more info.
3077 : *
3078 : * Returns bulk delete stats derived from input stats
3079 : */
3080 : static IndexBulkDeleteResult *
3081 2236 : lazy_vacuum_one_index(Relation indrel, IndexBulkDeleteResult *istat,
3082 : double reltuples, LVRelState *vacrel)
3083 : {
3084 : IndexVacuumInfo ivinfo;
3085 : LVSavedErrInfo saved_err_info;
3086 :
3087 2236 : ivinfo.index = indrel;
3088 2236 : ivinfo.heaprel = vacrel->rel;
3089 2236 : ivinfo.analyze_only = false;
3090 2236 : ivinfo.report_progress = false;
3091 2236 : ivinfo.estimated_count = true;
3092 2236 : ivinfo.message_level = DEBUG2;
3093 2236 : ivinfo.num_heap_tuples = reltuples;
3094 2236 : ivinfo.strategy = vacrel->bstrategy;
3095 :
3096 : /*
3097 : * Update error traceback information.
3098 : *
3099 : * The index name is saved during this phase and restored immediately
3100 : * after this phase. See vacuum_error_callback.
3101 : */
3102 : Assert(vacrel->indname == NULL);
3103 2236 : vacrel->indname = pstrdup(RelationGetRelationName(indrel));
3104 2236 : update_vacuum_error_info(vacrel, &saved_err_info,
3105 : VACUUM_ERRCB_PHASE_VACUUM_INDEX,
3106 : InvalidBlockNumber, InvalidOffsetNumber);
3107 :
3108 : /* Do bulk deletion */
3109 2236 : istat = vac_bulkdel_one_index(&ivinfo, istat, vacrel->dead_items,
3110 : vacrel->dead_items_info);
3111 :
3112 : /* Revert to the previous phase information for error traceback */
3113 2236 : restore_vacuum_error_info(vacrel, &saved_err_info);
3114 2236 : pfree(vacrel->indname);
3115 2236 : vacrel->indname = NULL;
3116 :
3117 2236 : return istat;
3118 : }
3119 :
3120 : /*
3121 : * lazy_cleanup_one_index() -- do post-vacuum cleanup for index relation.
3122 : *
3123 : * Calls index AM's amvacuumcleanup routine. reltuples is the number
3124 : * of heap tuples and estimated_count is true if reltuples is an
3125 : * estimated value. See indexam.sgml for more info.
3126 : *
3127 : * Returns bulk delete stats derived from input stats
3128 : */
3129 : static IndexBulkDeleteResult *
3130 160150 : lazy_cleanup_one_index(Relation indrel, IndexBulkDeleteResult *istat,
3131 : double reltuples, bool estimated_count,
3132 : LVRelState *vacrel)
3133 : {
3134 : IndexVacuumInfo ivinfo;
3135 : LVSavedErrInfo saved_err_info;
3136 :
3137 160150 : ivinfo.index = indrel;
3138 160150 : ivinfo.heaprel = vacrel->rel;
3139 160150 : ivinfo.analyze_only = false;
3140 160150 : ivinfo.report_progress = false;
3141 160150 : ivinfo.estimated_count = estimated_count;
3142 160150 : ivinfo.message_level = DEBUG2;
3143 :
3144 160150 : ivinfo.num_heap_tuples = reltuples;
3145 160150 : ivinfo.strategy = vacrel->bstrategy;
3146 :
3147 : /*
3148 : * Update error traceback information.
3149 : *
3150 : * The index name is saved during this phase and restored immediately
3151 : * after this phase. See vacuum_error_callback.
3152 : */
3153 : Assert(vacrel->indname == NULL);
3154 160150 : vacrel->indname = pstrdup(RelationGetRelationName(indrel));
3155 160150 : update_vacuum_error_info(vacrel, &saved_err_info,
3156 : VACUUM_ERRCB_PHASE_INDEX_CLEANUP,
3157 : InvalidBlockNumber, InvalidOffsetNumber);
3158 :
3159 160150 : istat = vac_cleanup_one_index(&ivinfo, istat);
3160 :
3161 : /* Revert to the previous phase information for error traceback */
3162 160150 : restore_vacuum_error_info(vacrel, &saved_err_info);
3163 160150 : pfree(vacrel->indname);
3164 160150 : vacrel->indname = NULL;
3165 :
3166 160150 : return istat;
3167 : }
3168 :
3169 : /*
3170 : * should_attempt_truncation - should we attempt to truncate the heap?
3171 : *
3172 : * Don't even think about it unless we have a shot at releasing a goodly
3173 : * number of pages. Otherwise, the time taken isn't worth it, mainly because
3174 : * an AccessExclusive lock must be replayed on any hot standby, where it can
3175 : * be particularly disruptive.
3176 : *
3177 : * Also don't attempt it if wraparound failsafe is in effect. The entire
3178 : * system might be refusing to allocate new XIDs at this point. The system
3179 : * definitely won't return to normal unless and until VACUUM actually advances
3180 : * the oldest relfrozenxid -- which hasn't happened for target rel just yet.
3181 : * If lazy_truncate_heap attempted to acquire an AccessExclusiveLock to
3182 : * truncate the table under these circumstances, an XID exhaustion error might
3183 : * make it impossible for VACUUM to fix the underlying XID exhaustion problem.
3184 : * There is very little chance of truncation working out when the failsafe is
3185 : * in effect in any case. lazy_scan_prune makes the optimistic assumption
3186 : * that any LP_DEAD items it encounters will always be LP_UNUSED by the time
3187 : * we're called.
3188 : */
3189 : static bool
3190 129846 : should_attempt_truncation(LVRelState *vacrel)
3191 : {
3192 : BlockNumber possibly_freeable;
3193 :
3194 129846 : if (!vacrel->do_rel_truncate || VacuumFailsafeActive)
3195 21974 : return false;
3196 :
3197 107872 : possibly_freeable = vacrel->rel_pages - vacrel->nonempty_pages;
3198 107872 : if (possibly_freeable > 0 &&
3199 328 : (possibly_freeable >= REL_TRUNCATE_MINIMUM ||
3200 328 : possibly_freeable >= vacrel->rel_pages / REL_TRUNCATE_FRACTION))
3201 312 : return true;
3202 :
3203 107560 : return false;
3204 : }
3205 :
3206 : /*
3207 : * lazy_truncate_heap - try to truncate off any empty pages at the end
3208 : */
3209 : static void
3210 312 : lazy_truncate_heap(LVRelState *vacrel)
3211 : {
3212 312 : BlockNumber orig_rel_pages = vacrel->rel_pages;
3213 : BlockNumber new_rel_pages;
3214 : bool lock_waiter_detected;
3215 : int lock_retry;
3216 :
3217 : /* Report that we are now truncating */
3218 312 : pgstat_progress_update_param(PROGRESS_VACUUM_PHASE,
3219 : PROGRESS_VACUUM_PHASE_TRUNCATE);
3220 :
3221 : /* Update error traceback information one last time */
3222 312 : update_vacuum_error_info(vacrel, NULL, VACUUM_ERRCB_PHASE_TRUNCATE,
3223 : vacrel->nonempty_pages, InvalidOffsetNumber);
3224 :
3225 : /*
3226 : * Loop until no more truncating can be done.
3227 : */
3228 : do
3229 : {
3230 : /*
3231 : * We need full exclusive lock on the relation in order to do
3232 : * truncation. If we can't get it, give up rather than waiting --- we
3233 : * don't want to block other backends, and we don't want to deadlock
3234 : * (which is quite possible considering we already hold a lower-grade
3235 : * lock).
3236 : */
3237 312 : lock_waiter_detected = false;
3238 312 : lock_retry = 0;
3239 : while (true)
3240 : {
3241 712 : if (ConditionalLockRelation(vacrel->rel, AccessExclusiveLock))
3242 308 : break;
3243 :
3244 : /*
3245 : * Check for interrupts while trying to (re-)acquire the exclusive
3246 : * lock.
3247 : */
3248 404 : CHECK_FOR_INTERRUPTS();
3249 :
3250 404 : if (++lock_retry > (VACUUM_TRUNCATE_LOCK_TIMEOUT /
3251 : VACUUM_TRUNCATE_LOCK_WAIT_INTERVAL))
3252 : {
3253 : /*
3254 : * We failed to establish the lock in the specified number of
3255 : * retries. This means we give up truncating.
3256 : */
3257 4 : ereport(vacrel->verbose ? INFO : DEBUG2,
3258 : (errmsg("\"%s\": stopping truncate due to conflicting lock request",
3259 : vacrel->relname)));
3260 6 : return;
3261 : }
3262 :
3263 400 : (void) WaitLatch(MyLatch,
3264 : WL_LATCH_SET | WL_TIMEOUT | WL_EXIT_ON_PM_DEATH,
3265 : VACUUM_TRUNCATE_LOCK_WAIT_INTERVAL,
3266 : WAIT_EVENT_VACUUM_TRUNCATE);
3267 400 : ResetLatch(MyLatch);
3268 : }
3269 :
3270 : /*
3271 : * Now that we have exclusive lock, look to see if the rel has grown
3272 : * whilst we were vacuuming with non-exclusive lock. If so, give up;
3273 : * the newly added pages presumably contain non-deletable tuples.
3274 : */
3275 308 : new_rel_pages = RelationGetNumberOfBlocks(vacrel->rel);
3276 308 : if (new_rel_pages != orig_rel_pages)
3277 : {
3278 : /*
3279 : * Note: we intentionally don't update vacrel->rel_pages with the
3280 : * new rel size here. If we did, it would amount to assuming that
3281 : * the new pages are empty, which is unlikely. Leaving the numbers
3282 : * alone amounts to assuming that the new pages have the same
3283 : * tuple density as existing ones, which is less unlikely.
3284 : */
3285 0 : UnlockRelation(vacrel->rel, AccessExclusiveLock);
3286 0 : return;
3287 : }
3288 :
3289 : /*
3290 : * Scan backwards from the end to verify that the end pages actually
3291 : * contain no tuples. This is *necessary*, not optional, because
3292 : * other backends could have added tuples to these pages whilst we
3293 : * were vacuuming.
3294 : */
3295 308 : new_rel_pages = count_nondeletable_pages(vacrel, &lock_waiter_detected);
3296 308 : vacrel->blkno = new_rel_pages;
3297 :
3298 308 : if (new_rel_pages >= orig_rel_pages)
3299 : {
3300 : /* can't do anything after all */
3301 2 : UnlockRelation(vacrel->rel, AccessExclusiveLock);
3302 2 : return;
3303 : }
3304 :
3305 : /*
3306 : * Okay to truncate.
3307 : */
3308 306 : RelationTruncate(vacrel->rel, new_rel_pages);
3309 :
3310 : /*
3311 : * We can release the exclusive lock as soon as we have truncated.
3312 : * Other backends can't safely access the relation until they have
3313 : * processed the smgr invalidation that smgrtruncate sent out ... but
3314 : * that should happen as part of standard invalidation processing once
3315 : * they acquire lock on the relation.
3316 : */
3317 306 : UnlockRelation(vacrel->rel, AccessExclusiveLock);
3318 :
3319 : /*
3320 : * Update statistics. Here, it *is* correct to adjust rel_pages
3321 : * without also touching reltuples, since the tuple count wasn't
3322 : * changed by the truncation.
3323 : */
3324 306 : vacrel->removed_pages += orig_rel_pages - new_rel_pages;
3325 306 : vacrel->rel_pages = new_rel_pages;
3326 :
3327 306 : ereport(vacrel->verbose ? INFO : DEBUG2,
3328 : (errmsg("table \"%s\": truncated %u to %u pages",
3329 : vacrel->relname,
3330 : orig_rel_pages, new_rel_pages)));
3331 306 : orig_rel_pages = new_rel_pages;
3332 306 : } while (new_rel_pages > vacrel->nonempty_pages && lock_waiter_detected);
3333 : }
3334 :
3335 : /*
3336 : * Rescan end pages to verify that they are (still) empty of tuples.
3337 : *
3338 : * Returns number of nondeletable pages (last nonempty page + 1).
3339 : */
3340 : static BlockNumber
3341 308 : count_nondeletable_pages(LVRelState *vacrel, bool *lock_waiter_detected)
3342 : {
3343 : BlockNumber blkno;
3344 : BlockNumber prefetchedUntil;
3345 : instr_time starttime;
3346 :
3347 : /* Initialize the starttime if we check for conflicting lock requests */
3348 308 : INSTR_TIME_SET_CURRENT(starttime);
3349 :
3350 : /*
3351 : * Start checking blocks at what we believe relation end to be and move
3352 : * backwards. (Strange coding of loop control is needed because blkno is
3353 : * unsigned.) To make the scan faster, we prefetch a few blocks at a time
3354 : * in forward direction, so that OS-level readahead can kick in.
3355 : */
3356 308 : blkno = vacrel->rel_pages;
3357 : StaticAssertStmt((PREFETCH_SIZE & (PREFETCH_SIZE - 1)) == 0,
3358 : "prefetch size must be power of 2");
3359 308 : prefetchedUntil = InvalidBlockNumber;
3360 5080 : while (blkno > vacrel->nonempty_pages)
3361 : {
3362 : Buffer buf;
3363 : Page page;
3364 : OffsetNumber offnum,
3365 : maxoff;
3366 : bool hastup;
3367 :
3368 : /*
3369 : * Check if another process requests a lock on our relation. We are
3370 : * holding an AccessExclusiveLock here, so they will be waiting. We
3371 : * only do this once per VACUUM_TRUNCATE_LOCK_CHECK_INTERVAL, and we
3372 : * only check if that interval has elapsed once every 32 blocks to
3373 : * keep the number of system calls and actual shared lock table
3374 : * lookups to a minimum.
3375 : */
3376 4776 : if ((blkno % 32) == 0)
3377 : {
3378 : instr_time currenttime;
3379 : instr_time elapsed;
3380 :
3381 156 : INSTR_TIME_SET_CURRENT(currenttime);
3382 156 : elapsed = currenttime;
3383 156 : INSTR_TIME_SUBTRACT(elapsed, starttime);
3384 156 : if ((INSTR_TIME_GET_MICROSEC(elapsed) / 1000)
3385 : >= VACUUM_TRUNCATE_LOCK_CHECK_INTERVAL)
3386 : {
3387 0 : if (LockHasWaitersRelation(vacrel->rel, AccessExclusiveLock))
3388 : {
3389 0 : ereport(vacrel->verbose ? INFO : DEBUG2,
3390 : (errmsg("table \"%s\": suspending truncate due to conflicting lock request",
3391 : vacrel->relname)));
3392 :
3393 0 : *lock_waiter_detected = true;
3394 0 : return blkno;
3395 : }
3396 0 : starttime = currenttime;
3397 : }
3398 : }
3399 :
3400 : /*
3401 : * We don't insert a vacuum delay point here, because we have an
3402 : * exclusive lock on the table which we want to hold for as short a
3403 : * time as possible. We still need to check for interrupts however.
3404 : */
3405 4776 : CHECK_FOR_INTERRUPTS();
3406 :
3407 4776 : blkno--;
3408 :
3409 : /* If we haven't prefetched this lot yet, do so now. */
3410 4776 : if (prefetchedUntil > blkno)
3411 : {
3412 : BlockNumber prefetchStart;
3413 : BlockNumber pblkno;
3414 :
3415 418 : prefetchStart = blkno & ~(PREFETCH_SIZE - 1);
3416 7086 : for (pblkno = prefetchStart; pblkno <= blkno; pblkno++)
3417 : {
3418 6668 : PrefetchBuffer(vacrel->rel, MAIN_FORKNUM, pblkno);
3419 6668 : CHECK_FOR_INTERRUPTS();
3420 : }
3421 418 : prefetchedUntil = prefetchStart;
3422 : }
3423 :
3424 4776 : buf = ReadBufferExtended(vacrel->rel, MAIN_FORKNUM, blkno, RBM_NORMAL,
3425 : vacrel->bstrategy);
3426 :
3427 : /* In this phase we only need shared access to the buffer */
3428 4776 : LockBuffer(buf, BUFFER_LOCK_SHARE);
3429 :
3430 4776 : page = BufferGetPage(buf);
3431 :
3432 4776 : if (PageIsNew(page) || PageIsEmpty(page))
3433 : {
3434 2384 : UnlockReleaseBuffer(buf);
3435 2384 : continue;
3436 : }
3437 :
3438 2392 : hastup = false;
3439 2392 : maxoff = PageGetMaxOffsetNumber(page);
3440 4780 : for (offnum = FirstOffsetNumber;
3441 : offnum <= maxoff;
3442 2388 : offnum = OffsetNumberNext(offnum))
3443 : {
3444 : ItemId itemid;
3445 :
3446 2392 : itemid = PageGetItemId(page, offnum);
3447 :
3448 : /*
3449 : * Note: any non-unused item should be taken as a reason to keep
3450 : * this page. Even an LP_DEAD item makes truncation unsafe, since
3451 : * we must not have cleaned out its index entries.
3452 : */
3453 2392 : if (ItemIdIsUsed(itemid))
3454 : {
3455 4 : hastup = true;
3456 4 : break; /* can stop scanning */
3457 : }
3458 : } /* scan along page */
3459 :
3460 2392 : UnlockReleaseBuffer(buf);
3461 :
3462 : /* Done scanning if we found a tuple here */
3463 2392 : if (hastup)
3464 4 : return blkno + 1;
3465 : }
3466 :
3467 : /*
3468 : * If we fall out of the loop, all the previously-thought-to-be-empty
3469 : * pages still are; we need not bother to look at the last known-nonempty
3470 : * page.
3471 : */
3472 304 : return vacrel->nonempty_pages;
3473 : }
3474 :
3475 : /*
3476 : * Allocate dead_items and dead_items_info (either using palloc, or in dynamic
3477 : * shared memory). Sets both in vacrel for caller.
3478 : *
3479 : * Also handles parallel initialization as part of allocating dead_items in
3480 : * DSM when required.
3481 : */
3482 : static void
3483 129846 : dead_items_alloc(LVRelState *vacrel, int nworkers)
3484 : {
3485 : VacDeadItemsInfo *dead_items_info;
3486 363416 : int vac_work_mem = AmAutoVacuumWorkerProcess() &&
3487 103724 : autovacuum_work_mem != -1 ?
3488 233570 : autovacuum_work_mem : maintenance_work_mem;
3489 :
3490 : /*
3491 : * Initialize state for a parallel vacuum. As of now, only one worker can
3492 : * be used for an index, so we invoke parallelism only if there are at
3493 : * least two indexes on a table.
3494 : */
3495 129846 : if (nworkers >= 0 && vacrel->nindexes > 1 && vacrel->do_index_vacuuming)
3496 : {
3497 : /*
3498 : * Since parallel workers cannot access data in temporary tables, we
3499 : * can't perform parallel vacuum on them.
3500 : */
3501 10286 : if (RelationUsesLocalBuffers(vacrel->rel))
3502 : {
3503 : /*
3504 : * Give warning only if the user explicitly tries to perform a
3505 : * parallel vacuum on the temporary table.
3506 : */
3507 6 : if (nworkers > 0)
3508 6 : ereport(WARNING,
3509 : (errmsg("disabling parallel option of vacuum on \"%s\" --- cannot vacuum temporary tables in parallel",
3510 : vacrel->relname)));
3511 : }
3512 : else
3513 10280 : vacrel->pvs = parallel_vacuum_init(vacrel->rel, vacrel->indrels,
3514 : vacrel->nindexes, nworkers,
3515 : vac_work_mem,
3516 10280 : vacrel->verbose ? INFO : DEBUG2,
3517 : vacrel->bstrategy);
3518 :
3519 : /*
3520 : * If parallel mode started, dead_items and dead_items_info spaces are
3521 : * allocated in DSM.
3522 : */
3523 10286 : if (ParallelVacuumIsActive(vacrel))
3524 : {
3525 34 : vacrel->dead_items = parallel_vacuum_get_dead_items(vacrel->pvs,
3526 : &vacrel->dead_items_info);
3527 34 : return;
3528 : }
3529 : }
3530 :
3531 : /*
3532 : * Serial VACUUM case. Allocate both dead_items and dead_items_info
3533 : * locally.
3534 : */
3535 :
3536 129812 : dead_items_info = (VacDeadItemsInfo *) palloc(sizeof(VacDeadItemsInfo));
3537 129812 : dead_items_info->max_bytes = vac_work_mem * (Size) 1024;
3538 129812 : dead_items_info->num_items = 0;
3539 129812 : vacrel->dead_items_info = dead_items_info;
3540 :
3541 129812 : vacrel->dead_items = TidStoreCreateLocal(dead_items_info->max_bytes, true);
3542 : }
3543 :
3544 : /*
3545 : * Add the given block number and offset numbers to dead_items.
3546 : */
3547 : static void
3548 27578 : dead_items_add(LVRelState *vacrel, BlockNumber blkno, OffsetNumber *offsets,
3549 : int num_offsets)
3550 : {
3551 27578 : const int prog_index[2] = {
3552 : PROGRESS_VACUUM_NUM_DEAD_ITEM_IDS,
3553 : PROGRESS_VACUUM_DEAD_TUPLE_BYTES
3554 : };
3555 : int64 prog_val[2];
3556 :
3557 27578 : TidStoreSetBlockOffsets(vacrel->dead_items, blkno, offsets, num_offsets);
3558 27578 : vacrel->dead_items_info->num_items += num_offsets;
3559 :
3560 : /* update the progress information */
3561 27578 : prog_val[0] = vacrel->dead_items_info->num_items;
3562 27578 : prog_val[1] = TidStoreMemoryUsage(vacrel->dead_items);
3563 27578 : pgstat_progress_update_multi_param(2, prog_index, prog_val);
3564 27578 : }
3565 :
3566 : /*
3567 : * Forget all collected dead items.
3568 : */
3569 : static void
3570 1200 : dead_items_reset(LVRelState *vacrel)
3571 : {
3572 1200 : if (ParallelVacuumIsActive(vacrel))
3573 : {
3574 10 : parallel_vacuum_reset_dead_items(vacrel->pvs);
3575 10 : return;
3576 : }
3577 :
3578 : /* Recreate the tidstore with the same max_bytes limitation */
3579 1190 : TidStoreDestroy(vacrel->dead_items);
3580 1190 : vacrel->dead_items = TidStoreCreateLocal(vacrel->dead_items_info->max_bytes, true);
3581 :
3582 : /* Reset the counter */
3583 1190 : vacrel->dead_items_info->num_items = 0;
3584 : }
3585 :
3586 : /*
3587 : * Perform cleanup for resources allocated in dead_items_alloc
3588 : */
3589 : static void
3590 129846 : dead_items_cleanup(LVRelState *vacrel)
3591 : {
3592 129846 : if (!ParallelVacuumIsActive(vacrel))
3593 : {
3594 : /* Don't bother with pfree here */
3595 129812 : return;
3596 : }
3597 :
3598 : /* End parallel mode */
3599 34 : parallel_vacuum_end(vacrel->pvs, vacrel->indstats);
3600 34 : vacrel->pvs = NULL;
3601 : }
3602 :
3603 : /*
3604 : * Check if every tuple in the given page is visible to all current and future
3605 : * transactions. Also return the visibility_cutoff_xid which is the highest
3606 : * xmin amongst the visible tuples. Set *all_frozen to true if every tuple
3607 : * on this page is frozen.
3608 : *
3609 : * This is a stripped down version of lazy_scan_prune(). If you change
3610 : * anything here, make sure that everything stays in sync. Note that an
3611 : * assertion calls us to verify that everybody still agrees. Be sure to avoid
3612 : * introducing new side-effects here.
3613 : */
3614 : static bool
3615 22962 : heap_page_is_all_visible(LVRelState *vacrel, Buffer buf,
3616 : TransactionId *visibility_cutoff_xid,
3617 : bool *all_frozen)
3618 : {
3619 22962 : Page page = BufferGetPage(buf);
3620 22962 : BlockNumber blockno = BufferGetBlockNumber(buf);
3621 : OffsetNumber offnum,
3622 : maxoff;
3623 22962 : bool all_visible = true;
3624 :
3625 22962 : *visibility_cutoff_xid = InvalidTransactionId;
3626 22962 : *all_frozen = true;
3627 :
3628 22962 : maxoff = PageGetMaxOffsetNumber(page);
3629 1169586 : for (offnum = FirstOffsetNumber;
3630 1146682 : offnum <= maxoff && all_visible;
3631 1146624 : offnum = OffsetNumberNext(offnum))
3632 : {
3633 : ItemId itemid;
3634 : HeapTupleData tuple;
3635 :
3636 : /*
3637 : * Set the offset number so that we can display it along with any
3638 : * error that occurred while processing this tuple.
3639 : */
3640 1146624 : vacrel->offnum = offnum;
3641 1146624 : itemid = PageGetItemId(page, offnum);
3642 :
3643 : /* Unused or redirect line pointers are of no interest */
3644 1146624 : if (!ItemIdIsUsed(itemid) || ItemIdIsRedirected(itemid))
3645 289554 : continue;
3646 :
3647 857070 : ItemPointerSet(&(tuple.t_self), blockno, offnum);
3648 :
3649 : /*
3650 : * Dead line pointers can have index pointers pointing to them. So
3651 : * they can't be treated as visible
3652 : */
3653 857070 : if (ItemIdIsDead(itemid))
3654 : {
3655 0 : all_visible = false;
3656 0 : *all_frozen = false;
3657 0 : break;
3658 : }
3659 :
3660 : Assert(ItemIdIsNormal(itemid));
3661 :
3662 857070 : tuple.t_data = (HeapTupleHeader) PageGetItem(page, itemid);
3663 857070 : tuple.t_len = ItemIdGetLength(itemid);
3664 857070 : tuple.t_tableOid = RelationGetRelid(vacrel->rel);
3665 :
3666 857070 : switch (HeapTupleSatisfiesVacuum(&tuple, vacrel->cutoffs.OldestXmin,
3667 : buf))
3668 : {
3669 857046 : case HEAPTUPLE_LIVE:
3670 : {
3671 : TransactionId xmin;
3672 :
3673 : /* Check comments in lazy_scan_prune. */
3674 857046 : if (!HeapTupleHeaderXminCommitted(tuple.t_data))
3675 : {
3676 0 : all_visible = false;
3677 0 : *all_frozen = false;
3678 0 : break;
3679 : }
3680 :
3681 : /*
3682 : * The inserter definitely committed. But is it old enough
3683 : * that everyone sees it as committed?
3684 : */
3685 857046 : xmin = HeapTupleHeaderGetXmin(tuple.t_data);
3686 857046 : if (!TransactionIdPrecedes(xmin,
3687 : vacrel->cutoffs.OldestXmin))
3688 : {
3689 36 : all_visible = false;
3690 36 : *all_frozen = false;
3691 36 : break;
3692 : }
3693 :
3694 : /* Track newest xmin on page. */
3695 857010 : if (TransactionIdFollows(xmin, *visibility_cutoff_xid) &&
3696 : TransactionIdIsNormal(xmin))
3697 18906 : *visibility_cutoff_xid = xmin;
3698 :
3699 : /* Check whether this tuple is already frozen or not */
3700 1074850 : if (all_visible && *all_frozen &&
3701 217840 : heap_tuple_needs_eventual_freeze(tuple.t_data))
3702 5408 : *all_frozen = false;
3703 : }
3704 857010 : break;
3705 :
3706 24 : case HEAPTUPLE_DEAD:
3707 : case HEAPTUPLE_RECENTLY_DEAD:
3708 : case HEAPTUPLE_INSERT_IN_PROGRESS:
3709 : case HEAPTUPLE_DELETE_IN_PROGRESS:
3710 : {
3711 24 : all_visible = false;
3712 24 : *all_frozen = false;
3713 24 : break;
3714 : }
3715 0 : default:
3716 0 : elog(ERROR, "unexpected HeapTupleSatisfiesVacuum result");
3717 : break;
3718 : }
3719 : } /* scan along page */
3720 :
3721 : /* Clear the offset information once we have processed the given page. */
3722 22962 : vacrel->offnum = InvalidOffsetNumber;
3723 :
3724 22962 : return all_visible;
3725 : }
3726 :
3727 : /*
3728 : * Update index statistics in pg_class if the statistics are accurate.
3729 : */
3730 : static void
3731 107880 : update_relstats_all_indexes(LVRelState *vacrel)
3732 : {
3733 107880 : Relation *indrels = vacrel->indrels;
3734 107880 : int nindexes = vacrel->nindexes;
3735 107880 : IndexBulkDeleteResult **indstats = vacrel->indstats;
3736 :
3737 : Assert(vacrel->do_index_cleanup);
3738 :
3739 268152 : for (int idx = 0; idx < nindexes; idx++)
3740 : {
3741 160272 : Relation indrel = indrels[idx];
3742 160272 : IndexBulkDeleteResult *istat = indstats[idx];
3743 :
3744 160272 : if (istat == NULL || istat->estimated_count)
3745 157768 : continue;
3746 :
3747 : /* Update index statistics */
3748 2504 : vac_update_relstats(indrel,
3749 : istat->num_pages,
3750 : istat->num_index_tuples,
3751 : 0, 0,
3752 : false,
3753 : InvalidTransactionId,
3754 : InvalidMultiXactId,
3755 : NULL, NULL, false);
3756 : }
3757 107880 : }
3758 :
3759 : /*
3760 : * Error context callback for errors occurring during vacuum. The error
3761 : * context messages for index phases should match the messages set in parallel
3762 : * vacuum. If you change this function for those phases, change
3763 : * parallel_vacuum_error_callback() as well.
3764 : */
3765 : static void
3766 89954 : vacuum_error_callback(void *arg)
3767 : {
3768 89954 : LVRelState *errinfo = arg;
3769 :
3770 89954 : switch (errinfo->phase)
3771 : {
3772 0 : case VACUUM_ERRCB_PHASE_SCAN_HEAP:
3773 0 : if (BlockNumberIsValid(errinfo->blkno))
3774 : {
3775 0 : if (OffsetNumberIsValid(errinfo->offnum))
3776 0 : errcontext("while scanning block %u offset %u of relation \"%s.%s\"",
3777 0 : errinfo->blkno, errinfo->offnum, errinfo->relnamespace, errinfo->relname);
3778 : else
3779 0 : errcontext("while scanning block %u of relation \"%s.%s\"",
3780 : errinfo->blkno, errinfo->relnamespace, errinfo->relname);
3781 : }
3782 : else
3783 0 : errcontext("while scanning relation \"%s.%s\"",
3784 : errinfo->relnamespace, errinfo->relname);
3785 0 : break;
3786 :
3787 0 : case VACUUM_ERRCB_PHASE_VACUUM_HEAP:
3788 0 : if (BlockNumberIsValid(errinfo->blkno))
3789 : {
3790 0 : if (OffsetNumberIsValid(errinfo->offnum))
3791 0 : errcontext("while vacuuming block %u offset %u of relation \"%s.%s\"",
3792 0 : errinfo->blkno, errinfo->offnum, errinfo->relnamespace, errinfo->relname);
3793 : else
3794 0 : errcontext("while vacuuming block %u of relation \"%s.%s\"",
3795 : errinfo->blkno, errinfo->relnamespace, errinfo->relname);
3796 : }
3797 : else
3798 0 : errcontext("while vacuuming relation \"%s.%s\"",
3799 : errinfo->relnamespace, errinfo->relname);
3800 0 : break;
3801 :
3802 0 : case VACUUM_ERRCB_PHASE_VACUUM_INDEX:
3803 0 : errcontext("while vacuuming index \"%s\" of relation \"%s.%s\"",
3804 : errinfo->indname, errinfo->relnamespace, errinfo->relname);
3805 0 : break;
3806 :
3807 0 : case VACUUM_ERRCB_PHASE_INDEX_CLEANUP:
3808 0 : errcontext("while cleaning up index \"%s\" of relation \"%s.%s\"",
3809 : errinfo->indname, errinfo->relnamespace, errinfo->relname);
3810 0 : break;
3811 :
3812 6 : case VACUUM_ERRCB_PHASE_TRUNCATE:
3813 6 : if (BlockNumberIsValid(errinfo->blkno))
3814 6 : errcontext("while truncating relation \"%s.%s\" to %u blocks",
3815 : errinfo->relnamespace, errinfo->relname, errinfo->blkno);
3816 6 : break;
3817 :
3818 89948 : case VACUUM_ERRCB_PHASE_UNKNOWN:
3819 : default:
3820 89948 : return; /* do nothing; the errinfo may not be
3821 : * initialized */
3822 : }
3823 : }
3824 :
3825 : /*
3826 : * Updates the information required for vacuum error callback. This also saves
3827 : * the current information which can be later restored via restore_vacuum_error_info.
3828 : */
3829 : static void
3830 954014 : update_vacuum_error_info(LVRelState *vacrel, LVSavedErrInfo *saved_vacrel,
3831 : int phase, BlockNumber blkno, OffsetNumber offnum)
3832 : {
3833 954014 : if (saved_vacrel)
3834 : {
3835 186520 : saved_vacrel->offnum = vacrel->offnum;
3836 186520 : saved_vacrel->blkno = vacrel->blkno;
3837 186520 : saved_vacrel->phase = vacrel->phase;
3838 : }
3839 :
3840 954014 : vacrel->blkno = blkno;
3841 954014 : vacrel->offnum = offnum;
3842 954014 : vacrel->phase = phase;
3843 954014 : }
3844 :
3845 : /*
3846 : * Restores the vacuum information saved via a prior call to update_vacuum_error_info.
3847 : */
3848 : static void
3849 186520 : restore_vacuum_error_info(LVRelState *vacrel,
3850 : const LVSavedErrInfo *saved_vacrel)
3851 : {
3852 186520 : vacrel->blkno = saved_vacrel->blkno;
3853 186520 : vacrel->offnum = saved_vacrel->offnum;
3854 186520 : vacrel->phase = saved_vacrel->phase;
3855 186520 : }
|
