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 238900 : 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 238900 : 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 238900 : vacrel->next_eager_scan_region_start = InvalidBlockNumber;
502 238900 : vacrel->eager_scan_max_fails_per_region = 0;
503 238900 : vacrel->eager_scan_remaining_fails = 0;
504 238900 : vacrel->eager_scan_remaining_successes = 0;
505 :
506 : /* If eager scanning is explicitly disabled, just return. */
507 238900 : if (params->max_eager_freeze_failure_rate == 0)
508 238900 : 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 238900 : if (vacrel->aggressive)
518 228230 : 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 10670 : if (vacrel->rel_pages < 2 * EAGER_SCAN_REGION_SIZE)
528 10670 : 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 238900 : 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 238900 : TimestampTz starttime = 0;
630 238900 : PgStat_Counter startreadtime = 0,
631 238900 : startwritetime = 0;
632 238900 : WalUsage startwalusage = pgWalUsage;
633 238900 : BufferUsage startbufferusage = pgBufferUsage;
634 : ErrorContextCallback errcallback;
635 238900 : char **indnames = NULL;
636 :
637 238900 : verbose = (params->options & VACOPT_VERBOSE) != 0;
638 451838 : instrument = (verbose || (AmAutoVacuumWorkerProcess() &&
639 212938 : params->log_min_duration >= 0));
640 238900 : if (instrument)
641 : {
642 212958 : pg_rusage_init(&ru0);
643 212958 : if (track_io_timing)
644 : {
645 0 : startreadtime = pgStatBlockReadTime;
646 0 : startwritetime = pgStatBlockWriteTime;
647 : }
648 : }
649 :
650 : /* Used for instrumentation and stats report */
651 238900 : starttime = GetCurrentTimestamp();
652 :
653 238900 : 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 238900 : vacrel = (LVRelState *) palloc0(sizeof(LVRelState));
669 238900 : vacrel->dbname = get_database_name(MyDatabaseId);
670 238900 : vacrel->relnamespace = get_namespace_name(RelationGetNamespace(rel));
671 238900 : vacrel->relname = pstrdup(RelationGetRelationName(rel));
672 238900 : vacrel->indname = NULL;
673 238900 : vacrel->phase = VACUUM_ERRCB_PHASE_UNKNOWN;
674 238900 : vacrel->verbose = verbose;
675 238900 : errcallback.callback = vacuum_error_callback;
676 238900 : errcallback.arg = vacrel;
677 238900 : errcallback.previous = error_context_stack;
678 238900 : error_context_stack = &errcallback;
679 :
680 : /* Set up high level stuff about rel and its indexes */
681 238900 : vacrel->rel = rel;
682 238900 : vac_open_indexes(vacrel->rel, RowExclusiveLock, &vacrel->nindexes,
683 : &vacrel->indrels);
684 238900 : vacrel->bstrategy = bstrategy;
685 238900 : if (instrument && vacrel->nindexes > 0)
686 : {
687 : /* Copy index names used by instrumentation (not error reporting) */
688 203594 : indnames = palloc(sizeof(char *) * vacrel->nindexes);
689 524690 : for (int i = 0; i < vacrel->nindexes; i++)
690 321096 : 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 238900 : VacuumFailsafeActive = false;
711 238900 : vacrel->consider_bypass_optimization = true;
712 238900 : vacrel->do_index_vacuuming = true;
713 238900 : vacrel->do_index_cleanup = true;
714 238900 : vacrel->do_rel_truncate = (params->truncate != VACOPTVALUE_DISABLED);
715 238900 : if (params->index_cleanup == VACOPTVALUE_DISABLED)
716 : {
717 : /* Force disable index vacuuming up-front */
718 260 : vacrel->do_index_vacuuming = false;
719 260 : vacrel->do_index_cleanup = false;
720 : }
721 238640 : else if (params->index_cleanup == VACOPTVALUE_ENABLED)
722 : {
723 : /* Force index vacuuming. Note that failsafe can still bypass. */
724 30 : 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 238900 : vacrel->scanned_pages = 0;
734 238900 : vacrel->eager_scanned_pages = 0;
735 238900 : vacrel->removed_pages = 0;
736 238900 : vacrel->new_frozen_tuple_pages = 0;
737 238900 : vacrel->lpdead_item_pages = 0;
738 238900 : vacrel->missed_dead_pages = 0;
739 238900 : vacrel->nonempty_pages = 0;
740 : /* dead_items_alloc allocates vacrel->dead_items later on */
741 :
742 : /* Allocate/initialize output statistics state */
743 238900 : vacrel->new_rel_tuples = 0;
744 238900 : vacrel->new_live_tuples = 0;
745 238900 : vacrel->indstats = (IndexBulkDeleteResult **)
746 238900 : palloc0(vacrel->nindexes * sizeof(IndexBulkDeleteResult *));
747 :
748 : /* Initialize remaining counters (be tidy) */
749 238900 : vacrel->num_index_scans = 0;
750 238900 : vacrel->tuples_deleted = 0;
751 238900 : vacrel->tuples_frozen = 0;
752 238900 : vacrel->lpdead_items = 0;
753 238900 : vacrel->live_tuples = 0;
754 238900 : vacrel->recently_dead_tuples = 0;
755 238900 : vacrel->missed_dead_tuples = 0;
756 :
757 238900 : vacrel->vm_new_visible_pages = 0;
758 238900 : vacrel->vm_new_visible_frozen_pages = 0;
759 238900 : vacrel->vm_new_frozen_pages = 0;
760 :
761 : /*
762 : * Get cutoffs that determine which deleted tuples are considered DEAD,
763 : * not just RECENTLY_DEAD, and which XIDs/MXIDs to freeze. Then determine
764 : * the extent of the blocks that we'll scan in lazy_scan_heap. It has to
765 : * happen in this order to ensure that the OldestXmin cutoff field works
766 : * as an upper bound on the XIDs stored in the pages we'll actually scan
767 : * (NewRelfrozenXid tracking must never be allowed to miss unfrozen XIDs).
768 : *
769 : * Next acquire vistest, a related cutoff that's used in pruning. We use
770 : * vistest in combination with OldestXmin to ensure that
771 : * heap_page_prune_and_freeze() always removes any deleted tuple whose
772 : * xmax is < OldestXmin. lazy_scan_prune must never become confused about
773 : * whether a tuple should be frozen or removed. (In the future we might
774 : * want to teach lazy_scan_prune to recompute vistest from time to time,
775 : * to increase the number of dead tuples it can prune away.)
776 : */
777 238900 : vacrel->aggressive = vacuum_get_cutoffs(rel, params, &vacrel->cutoffs);
778 238900 : vacrel->rel_pages = orig_rel_pages = RelationGetNumberOfBlocks(rel);
779 238900 : vacrel->vistest = GlobalVisTestFor(rel);
780 :
781 : /* Initialize state used to track oldest extant XID/MXID */
782 238900 : vacrel->NewRelfrozenXid = vacrel->cutoffs.OldestXmin;
783 238900 : vacrel->NewRelminMxid = vacrel->cutoffs.OldestMxact;
784 :
785 : /*
786 : * Initialize state related to tracking all-visible page skipping. This is
787 : * very important to determine whether or not it is safe to advance the
788 : * relfrozenxid/relminmxid.
789 : */
790 238900 : vacrel->skippedallvis = false;
791 238900 : skipwithvm = true;
792 238900 : if (params->options & VACOPT_DISABLE_PAGE_SKIPPING)
793 : {
794 : /*
795 : * Force aggressive mode, and disable skipping blocks using the
796 : * visibility map (even those set all-frozen)
797 : */
798 344 : vacrel->aggressive = true;
799 344 : skipwithvm = false;
800 : }
801 :
802 238900 : vacrel->skipwithvm = skipwithvm;
803 :
804 : /*
805 : * Set up eager scan tracking state. This must happen after determining
806 : * whether or not the vacuum must be aggressive, because only normal
807 : * vacuums use the eager scan algorithm.
808 : */
809 238900 : heap_vacuum_eager_scan_setup(vacrel, params);
810 :
811 238900 : if (verbose)
812 : {
813 20 : if (vacrel->aggressive)
814 2 : ereport(INFO,
815 : (errmsg("aggressively vacuuming \"%s.%s.%s\"",
816 : vacrel->dbname, vacrel->relnamespace,
817 : vacrel->relname)));
818 : else
819 18 : ereport(INFO,
820 : (errmsg("vacuuming \"%s.%s.%s\"",
821 : vacrel->dbname, vacrel->relnamespace,
822 : vacrel->relname)));
823 : }
824 :
825 : /*
826 : * Allocate dead_items memory using dead_items_alloc. This handles
827 : * parallel VACUUM initialization as part of allocating shared memory
828 : * space used for dead_items. (But do a failsafe precheck first, to
829 : * ensure that parallel VACUUM won't be attempted at all when relfrozenxid
830 : * is already dangerously old.)
831 : */
832 238900 : lazy_check_wraparound_failsafe(vacrel);
833 238900 : dead_items_alloc(vacrel, params->nworkers);
834 :
835 : /*
836 : * Call lazy_scan_heap to perform all required heap pruning, index
837 : * vacuuming, and heap vacuuming (plus related processing)
838 : */
839 238900 : lazy_scan_heap(vacrel);
840 :
841 : /*
842 : * Free resources managed by dead_items_alloc. This ends parallel mode in
843 : * passing when necessary.
844 : */
845 238900 : dead_items_cleanup(vacrel);
846 : Assert(!IsInParallelMode());
847 :
848 : /*
849 : * Update pg_class entries for each of rel's indexes where appropriate.
850 : *
851 : * Unlike the later update to rel's pg_class entry, this is not critical.
852 : * Maintains relpages/reltuples statistics used by the planner only.
853 : */
854 238900 : if (vacrel->do_index_cleanup)
855 175776 : update_relstats_all_indexes(vacrel);
856 :
857 : /* Done with rel's indexes */
858 238900 : vac_close_indexes(vacrel->nindexes, vacrel->indrels, NoLock);
859 :
860 : /* Optionally truncate rel */
861 238900 : if (should_attempt_truncation(vacrel))
862 284 : lazy_truncate_heap(vacrel);
863 :
864 : /* Pop the error context stack */
865 238900 : error_context_stack = errcallback.previous;
866 :
867 : /* Report that we are now doing final cleanup */
868 238900 : pgstat_progress_update_param(PROGRESS_VACUUM_PHASE,
869 : PROGRESS_VACUUM_PHASE_FINAL_CLEANUP);
870 :
871 : /*
872 : * Prepare to update rel's pg_class entry.
873 : *
874 : * Aggressive VACUUMs must always be able to advance relfrozenxid to a
875 : * value >= FreezeLimit, and relminmxid to a value >= MultiXactCutoff.
876 : * Non-aggressive VACUUMs may advance them by any amount, or not at all.
877 : */
878 : Assert(vacrel->NewRelfrozenXid == vacrel->cutoffs.OldestXmin ||
879 : TransactionIdPrecedesOrEquals(vacrel->aggressive ? vacrel->cutoffs.FreezeLimit :
880 : vacrel->cutoffs.relfrozenxid,
881 : vacrel->NewRelfrozenXid));
882 : Assert(vacrel->NewRelminMxid == vacrel->cutoffs.OldestMxact ||
883 : MultiXactIdPrecedesOrEquals(vacrel->aggressive ? vacrel->cutoffs.MultiXactCutoff :
884 : vacrel->cutoffs.relminmxid,
885 : vacrel->NewRelminMxid));
886 238900 : if (vacrel->skippedallvis)
887 : {
888 : /*
889 : * Must keep original relfrozenxid in a non-aggressive VACUUM that
890 : * chose to skip an all-visible page range. The state that tracks new
891 : * values will have missed unfrozen XIDs from the pages we skipped.
892 : */
893 : Assert(!vacrel->aggressive);
894 60 : vacrel->NewRelfrozenXid = InvalidTransactionId;
895 60 : vacrel->NewRelminMxid = InvalidMultiXactId;
896 : }
897 :
898 : /*
899 : * For safety, clamp relallvisible to be not more than what we're setting
900 : * pg_class.relpages to
901 : */
902 238900 : new_rel_pages = vacrel->rel_pages; /* After possible rel truncation */
903 238900 : visibilitymap_count(rel, &new_rel_allvisible, &new_rel_allfrozen);
904 238900 : if (new_rel_allvisible > new_rel_pages)
905 0 : new_rel_allvisible = new_rel_pages;
906 :
907 : /*
908 : * An all-frozen block _must_ be all-visible. As such, clamp the count of
909 : * all-frozen blocks to the count of all-visible blocks. This matches the
910 : * clamping of relallvisible above.
911 : */
912 238900 : if (new_rel_allfrozen > new_rel_allvisible)
913 0 : new_rel_allfrozen = new_rel_allvisible;
914 :
915 : /*
916 : * Now actually update rel's pg_class entry.
917 : *
918 : * In principle new_live_tuples could be -1 indicating that we (still)
919 : * don't know the tuple count. In practice that can't happen, since we
920 : * scan every page that isn't skipped using the visibility map.
921 : */
922 238900 : vac_update_relstats(rel, new_rel_pages, vacrel->new_live_tuples,
923 : new_rel_allvisible, new_rel_allfrozen,
924 238900 : vacrel->nindexes > 0,
925 : vacrel->NewRelfrozenXid, vacrel->NewRelminMxid,
926 : &frozenxid_updated, &minmulti_updated, false);
927 :
928 : /*
929 : * Report results to the cumulative stats system, too.
930 : *
931 : * Deliberately avoid telling the stats system about LP_DEAD items that
932 : * remain in the table due to VACUUM bypassing index and heap vacuuming.
933 : * ANALYZE will consider the remaining LP_DEAD items to be dead "tuples".
934 : * It seems like a good idea to err on the side of not vacuuming again too
935 : * soon in cases where the failsafe prevented significant amounts of heap
936 : * vacuuming.
937 : */
938 143310 : pgstat_report_vacuum(RelationGetRelid(rel),
939 238900 : rel->rd_rel->relisshared,
940 95590 : Max(vacrel->new_live_tuples, 0),
941 238900 : vacrel->recently_dead_tuples +
942 238900 : vacrel->missed_dead_tuples,
943 : starttime);
944 238900 : pgstat_progress_end_command();
945 :
946 238900 : if (instrument)
947 : {
948 212958 : TimestampTz endtime = GetCurrentTimestamp();
949 :
950 213150 : if (verbose || params->log_min_duration == 0 ||
951 192 : TimestampDifferenceExceeds(starttime, endtime,
952 : params->log_min_duration))
953 : {
954 : long secs_dur;
955 : int usecs_dur;
956 : WalUsage walusage;
957 : BufferUsage bufferusage;
958 : StringInfoData buf;
959 : char *msgfmt;
960 : int32 diff;
961 212766 : double read_rate = 0,
962 212766 : write_rate = 0;
963 : int64 total_blks_hit;
964 : int64 total_blks_read;
965 : int64 total_blks_dirtied;
966 :
967 212766 : TimestampDifference(starttime, endtime, &secs_dur, &usecs_dur);
968 212766 : memset(&walusage, 0, sizeof(WalUsage));
969 212766 : WalUsageAccumDiff(&walusage, &pgWalUsage, &startwalusage);
970 212766 : memset(&bufferusage, 0, sizeof(BufferUsage));
971 212766 : BufferUsageAccumDiff(&bufferusage, &pgBufferUsage, &startbufferusage);
972 :
973 212766 : total_blks_hit = bufferusage.shared_blks_hit +
974 212766 : bufferusage.local_blks_hit;
975 212766 : total_blks_read = bufferusage.shared_blks_read +
976 212766 : bufferusage.local_blks_read;
977 212766 : total_blks_dirtied = bufferusage.shared_blks_dirtied +
978 212766 : bufferusage.local_blks_dirtied;
979 :
980 212766 : initStringInfo(&buf);
981 212766 : if (verbose)
982 : {
983 : /*
984 : * Aggressiveness already reported earlier, in dedicated
985 : * VACUUM VERBOSE ereport
986 : */
987 : Assert(!params->is_wraparound);
988 20 : msgfmt = _("finished vacuuming \"%s.%s.%s\": index scans: %d\n");
989 : }
990 212746 : else if (params->is_wraparound)
991 : {
992 : /*
993 : * While it's possible for a VACUUM to be both is_wraparound
994 : * and !aggressive, that's just a corner-case -- is_wraparound
995 : * implies aggressive. Produce distinct output for the corner
996 : * case all the same, just in case.
997 : */
998 212706 : if (vacrel->aggressive)
999 212690 : msgfmt = _("automatic aggressive vacuum to prevent wraparound of table \"%s.%s.%s\": index scans: %d\n");
1000 : else
1001 16 : msgfmt = _("automatic vacuum to prevent wraparound of table \"%s.%s.%s\": index scans: %d\n");
1002 : }
1003 : else
1004 : {
1005 40 : if (vacrel->aggressive)
1006 32 : msgfmt = _("automatic aggressive vacuum of table \"%s.%s.%s\": index scans: %d\n");
1007 : else
1008 8 : msgfmt = _("automatic vacuum of table \"%s.%s.%s\": index scans: %d\n");
1009 : }
1010 212766 : appendStringInfo(&buf, msgfmt,
1011 : vacrel->dbname,
1012 : vacrel->relnamespace,
1013 : vacrel->relname,
1014 : vacrel->num_index_scans);
1015 297486 : appendStringInfo(&buf, _("pages: %u removed, %u remain, %u scanned (%.2f%% of total), %u eagerly scanned\n"),
1016 : vacrel->removed_pages,
1017 : new_rel_pages,
1018 : vacrel->scanned_pages,
1019 : orig_rel_pages == 0 ? 100.0 :
1020 84720 : 100.0 * vacrel->scanned_pages /
1021 : orig_rel_pages,
1022 : vacrel->eager_scanned_pages);
1023 212766 : appendStringInfo(&buf,
1024 212766 : _("tuples: %" PRId64 " removed, %" PRId64 " remain, %" PRId64 " are dead but not yet removable\n"),
1025 : vacrel->tuples_deleted,
1026 212766 : (int64) vacrel->new_rel_tuples,
1027 : vacrel->recently_dead_tuples);
1028 212766 : if (vacrel->missed_dead_tuples > 0)
1029 0 : appendStringInfo(&buf,
1030 0 : _("tuples missed: %" PRId64 " dead from %u pages not removed due to cleanup lock contention\n"),
1031 : vacrel->missed_dead_tuples,
1032 : vacrel->missed_dead_pages);
1033 212766 : diff = (int32) (ReadNextTransactionId() -
1034 212766 : vacrel->cutoffs.OldestXmin);
1035 212766 : appendStringInfo(&buf,
1036 212766 : _("removable cutoff: %u, which was %d XIDs old when operation ended\n"),
1037 : vacrel->cutoffs.OldestXmin, diff);
1038 212766 : if (frozenxid_updated)
1039 : {
1040 39490 : diff = (int32) (vacrel->NewRelfrozenXid -
1041 39490 : vacrel->cutoffs.relfrozenxid);
1042 39490 : appendStringInfo(&buf,
1043 39490 : _("new relfrozenxid: %u, which is %d XIDs ahead of previous value\n"),
1044 : vacrel->NewRelfrozenXid, diff);
1045 : }
1046 212766 : if (minmulti_updated)
1047 : {
1048 10 : diff = (int32) (vacrel->NewRelminMxid -
1049 10 : vacrel->cutoffs.relminmxid);
1050 10 : appendStringInfo(&buf,
1051 10 : _("new relminmxid: %u, which is %d MXIDs ahead of previous value\n"),
1052 : vacrel->NewRelminMxid, diff);
1053 : }
1054 297486 : appendStringInfo(&buf, _("frozen: %u pages from table (%.2f%% of total) had %" PRId64 " tuples frozen\n"),
1055 : vacrel->new_frozen_tuple_pages,
1056 : orig_rel_pages == 0 ? 100.0 :
1057 84720 : 100.0 * vacrel->new_frozen_tuple_pages /
1058 : orig_rel_pages,
1059 : vacrel->tuples_frozen);
1060 :
1061 212766 : appendStringInfo(&buf,
1062 212766 : _("visibility map: %u pages set all-visible, %u pages set all-frozen (%u were all-visible)\n"),
1063 : vacrel->vm_new_visible_pages,
1064 212766 : vacrel->vm_new_visible_frozen_pages +
1065 212766 : vacrel->vm_new_frozen_pages,
1066 : vacrel->vm_new_frozen_pages);
1067 212766 : if (vacrel->do_index_vacuuming)
1068 : {
1069 150120 : if (vacrel->nindexes == 0 || vacrel->num_index_scans == 0)
1070 150088 : appendStringInfoString(&buf, _("index scan not needed: "));
1071 : else
1072 32 : appendStringInfoString(&buf, _("index scan needed: "));
1073 :
1074 150120 : msgfmt = _("%u pages from table (%.2f%% of total) had %" PRId64 " dead item identifiers removed\n");
1075 : }
1076 : else
1077 : {
1078 62646 : if (!VacuumFailsafeActive)
1079 0 : appendStringInfoString(&buf, _("index scan bypassed: "));
1080 : else
1081 62646 : appendStringInfoString(&buf, _("index scan bypassed by failsafe: "));
1082 :
1083 62646 : msgfmt = _("%u pages from table (%.2f%% of total) have %" PRId64 " dead item identifiers\n");
1084 : }
1085 297486 : appendStringInfo(&buf, msgfmt,
1086 : vacrel->lpdead_item_pages,
1087 : orig_rel_pages == 0 ? 100.0 :
1088 84720 : 100.0 * vacrel->lpdead_item_pages / orig_rel_pages,
1089 : vacrel->lpdead_items);
1090 533486 : for (int i = 0; i < vacrel->nindexes; i++)
1091 : {
1092 320720 : IndexBulkDeleteResult *istat = vacrel->indstats[i];
1093 :
1094 320720 : if (!istat)
1095 320666 : continue;
1096 :
1097 54 : appendStringInfo(&buf,
1098 54 : _("index \"%s\": pages: %u in total, %u newly deleted, %u currently deleted, %u reusable\n"),
1099 54 : indnames[i],
1100 : istat->num_pages,
1101 : istat->pages_newly_deleted,
1102 : istat->pages_deleted,
1103 : istat->pages_free);
1104 : }
1105 212766 : if (track_cost_delay_timing)
1106 : {
1107 : /*
1108 : * We bypass the changecount mechanism because this value is
1109 : * only updated by the calling process. We also rely on the
1110 : * above call to pgstat_progress_end_command() to not clear
1111 : * the st_progress_param array.
1112 : */
1113 0 : appendStringInfo(&buf, _("delay time: %.3f ms\n"),
1114 0 : (double) MyBEEntry->st_progress_param[PROGRESS_VACUUM_DELAY_TIME] / 1000000.0);
1115 : }
1116 212766 : if (track_io_timing)
1117 : {
1118 0 : double read_ms = (double) (pgStatBlockReadTime - startreadtime) / 1000;
1119 0 : double write_ms = (double) (pgStatBlockWriteTime - startwritetime) / 1000;
1120 :
1121 0 : appendStringInfo(&buf, _("I/O timings: read: %.3f ms, write: %.3f ms\n"),
1122 : read_ms, write_ms);
1123 : }
1124 212766 : if (secs_dur > 0 || usecs_dur > 0)
1125 : {
1126 212766 : read_rate = (double) BLCKSZ * total_blks_read /
1127 212766 : (1024 * 1024) / (secs_dur + usecs_dur / 1000000.0);
1128 212766 : write_rate = (double) BLCKSZ * total_blks_dirtied /
1129 212766 : (1024 * 1024) / (secs_dur + usecs_dur / 1000000.0);
1130 : }
1131 212766 : appendStringInfo(&buf, _("avg read rate: %.3f MB/s, avg write rate: %.3f MB/s\n"),
1132 : read_rate, write_rate);
1133 212766 : appendStringInfo(&buf,
1134 212766 : _("buffer usage: %" PRId64 " hits, %" PRId64 " reads, %" PRId64 " dirtied\n"),
1135 : total_blks_hit,
1136 : total_blks_read,
1137 : total_blks_dirtied);
1138 212766 : appendStringInfo(&buf,
1139 212766 : _("WAL usage: %" PRId64 " records, %" PRId64 " full page images, %" PRIu64 " bytes, %" PRId64 " buffers full\n"),
1140 : walusage.wal_records,
1141 : walusage.wal_fpi,
1142 : walusage.wal_bytes,
1143 : walusage.wal_buffers_full);
1144 212766 : appendStringInfo(&buf, _("system usage: %s"), pg_rusage_show(&ru0));
1145 :
1146 212766 : ereport(verbose ? INFO : LOG,
1147 : (errmsg_internal("%s", buf.data)));
1148 212764 : pfree(buf.data);
1149 : }
1150 : }
1151 :
1152 : /* Cleanup index statistics and index names */
1153 597188 : for (int i = 0; i < vacrel->nindexes; i++)
1154 : {
1155 358290 : if (vacrel->indstats[i])
1156 2702 : pfree(vacrel->indstats[i]);
1157 :
1158 358290 : if (instrument)
1159 321090 : pfree(indnames[i]);
1160 : }
1161 238898 : }
1162 :
1163 : /*
1164 : * lazy_scan_heap() -- workhorse function for VACUUM
1165 : *
1166 : * This routine prunes each page in the heap, and considers the need to
1167 : * freeze remaining tuples with storage (not including pages that can be
1168 : * skipped using the visibility map). Also performs related maintenance
1169 : * of the FSM and visibility map. These steps all take place during an
1170 : * initial pass over the target heap relation.
1171 : *
1172 : * Also invokes lazy_vacuum_all_indexes to vacuum indexes, which largely
1173 : * consists of deleting index tuples that point to LP_DEAD items left in
1174 : * heap pages following pruning. Earlier initial pass over the heap will
1175 : * have collected the TIDs whose index tuples need to be removed.
1176 : *
1177 : * Finally, invokes lazy_vacuum_heap_rel to vacuum heap pages, which
1178 : * largely consists of marking LP_DEAD items (from vacrel->dead_items)
1179 : * as LP_UNUSED. This has to happen in a second, final pass over the
1180 : * heap, to preserve a basic invariant that all index AMs rely on: no
1181 : * extant index tuple can ever be allowed to contain a TID that points to
1182 : * an LP_UNUSED line pointer in the heap. We must disallow premature
1183 : * recycling of line pointers to avoid index scans that get confused
1184 : * about which TID points to which tuple immediately after recycling.
1185 : * (Actually, this isn't a concern when target heap relation happens to
1186 : * have no indexes, which allows us to safely apply the one-pass strategy
1187 : * as an optimization).
1188 : *
1189 : * In practice we often have enough space to fit all TIDs, and so won't
1190 : * need to call lazy_vacuum more than once, after our initial pass over
1191 : * the heap has totally finished. Otherwise things are slightly more
1192 : * complicated: our "initial pass" over the heap applies only to those
1193 : * pages that were pruned before we needed to call lazy_vacuum, and our
1194 : * "final pass" over the heap only vacuums these same heap pages.
1195 : * However, we process indexes in full every time lazy_vacuum is called,
1196 : * which makes index processing very inefficient when memory is in short
1197 : * supply.
1198 : */
1199 : static void
1200 238900 : lazy_scan_heap(LVRelState *vacrel)
1201 : {
1202 : ReadStream *stream;
1203 238900 : BlockNumber rel_pages = vacrel->rel_pages,
1204 238900 : blkno = 0,
1205 238900 : next_fsm_block_to_vacuum = 0;
1206 238900 : BlockNumber orig_eager_scan_success_limit =
1207 : vacrel->eager_scan_remaining_successes; /* for logging */
1208 238900 : Buffer vmbuffer = InvalidBuffer;
1209 238900 : const int initprog_index[] = {
1210 : PROGRESS_VACUUM_PHASE,
1211 : PROGRESS_VACUUM_TOTAL_HEAP_BLKS,
1212 : PROGRESS_VACUUM_MAX_DEAD_TUPLE_BYTES
1213 : };
1214 : int64 initprog_val[3];
1215 :
1216 : /* Report that we're scanning the heap, advertising total # of blocks */
1217 238900 : initprog_val[0] = PROGRESS_VACUUM_PHASE_SCAN_HEAP;
1218 238900 : initprog_val[1] = rel_pages;
1219 238900 : initprog_val[2] = vacrel->dead_items_info->max_bytes;
1220 238900 : pgstat_progress_update_multi_param(3, initprog_index, initprog_val);
1221 :
1222 : /* Initialize for the first heap_vac_scan_next_block() call */
1223 238900 : vacrel->current_block = InvalidBlockNumber;
1224 238900 : vacrel->next_unskippable_block = InvalidBlockNumber;
1225 238900 : vacrel->next_unskippable_allvis = false;
1226 238900 : vacrel->next_unskippable_eager_scanned = false;
1227 238900 : vacrel->next_unskippable_vmbuffer = InvalidBuffer;
1228 :
1229 : /*
1230 : * Set up the read stream for vacuum's first pass through the heap.
1231 : *
1232 : * This could be made safe for READ_STREAM_USE_BATCHING, but only with
1233 : * explicit work in heap_vac_scan_next_block.
1234 : */
1235 238900 : stream = read_stream_begin_relation(READ_STREAM_MAINTENANCE,
1236 : vacrel->bstrategy,
1237 : vacrel->rel,
1238 : MAIN_FORKNUM,
1239 : heap_vac_scan_next_block,
1240 : vacrel,
1241 : sizeof(uint8));
1242 :
1243 : while (true)
1244 877746 : {
1245 : Buffer buf;
1246 : Page page;
1247 1116646 : uint8 blk_info = 0;
1248 : bool has_lpdead_items;
1249 1116646 : void *per_buffer_data = NULL;
1250 1116646 : bool vm_page_frozen = false;
1251 1116646 : bool got_cleanup_lock = false;
1252 :
1253 1116646 : vacuum_delay_point(false);
1254 :
1255 : /*
1256 : * Regularly check if wraparound failsafe should trigger.
1257 : *
1258 : * There is a similar check inside lazy_vacuum_all_indexes(), but
1259 : * relfrozenxid might start to look dangerously old before we reach
1260 : * that point. This check also provides failsafe coverage for the
1261 : * one-pass strategy, and the two-pass strategy with the index_cleanup
1262 : * param set to 'off'.
1263 : */
1264 1116646 : if (vacrel->scanned_pages > 0 &&
1265 877746 : vacrel->scanned_pages % FAILSAFE_EVERY_PAGES == 0)
1266 0 : lazy_check_wraparound_failsafe(vacrel);
1267 :
1268 : /*
1269 : * Consider if we definitely have enough space to process TIDs on page
1270 : * already. If we are close to overrunning the available space for
1271 : * dead_items TIDs, pause and do a cycle of vacuuming before we tackle
1272 : * this page. However, let's force at least one page-worth of tuples
1273 : * to be stored as to ensure we do at least some work when the memory
1274 : * configured is so low that we run out before storing anything.
1275 : */
1276 1116646 : if (vacrel->dead_items_info->num_items > 0 &&
1277 47506 : TidStoreMemoryUsage(vacrel->dead_items) > vacrel->dead_items_info->max_bytes)
1278 : {
1279 : /*
1280 : * Before beginning index vacuuming, we release any pin we may
1281 : * hold on the visibility map page. This isn't necessary for
1282 : * correctness, but we do it anyway to avoid holding the pin
1283 : * across a lengthy, unrelated operation.
1284 : */
1285 4 : if (BufferIsValid(vmbuffer))
1286 : {
1287 4 : ReleaseBuffer(vmbuffer);
1288 4 : vmbuffer = InvalidBuffer;
1289 : }
1290 :
1291 : /* Perform a round of index and heap vacuuming */
1292 4 : vacrel->consider_bypass_optimization = false;
1293 4 : lazy_vacuum(vacrel);
1294 :
1295 : /*
1296 : * Vacuum the Free Space Map to make newly-freed space visible on
1297 : * upper-level FSM pages. Note that blkno is the previously
1298 : * processed block.
1299 : */
1300 4 : FreeSpaceMapVacuumRange(vacrel->rel, next_fsm_block_to_vacuum,
1301 : blkno + 1);
1302 4 : next_fsm_block_to_vacuum = blkno;
1303 :
1304 : /* Report that we are once again scanning the heap */
1305 4 : pgstat_progress_update_param(PROGRESS_VACUUM_PHASE,
1306 : PROGRESS_VACUUM_PHASE_SCAN_HEAP);
1307 : }
1308 :
1309 1116646 : buf = read_stream_next_buffer(stream, &per_buffer_data);
1310 :
1311 : /* The relation is exhausted. */
1312 1116646 : if (!BufferIsValid(buf))
1313 238900 : break;
1314 :
1315 877746 : blk_info = *((uint8 *) per_buffer_data);
1316 877746 : CheckBufferIsPinnedOnce(buf);
1317 877746 : page = BufferGetPage(buf);
1318 877746 : blkno = BufferGetBlockNumber(buf);
1319 :
1320 877746 : vacrel->scanned_pages++;
1321 877746 : if (blk_info & VAC_BLK_WAS_EAGER_SCANNED)
1322 0 : vacrel->eager_scanned_pages++;
1323 :
1324 : /* Report as block scanned, update error traceback information */
1325 877746 : pgstat_progress_update_param(PROGRESS_VACUUM_HEAP_BLKS_SCANNED, blkno);
1326 877746 : update_vacuum_error_info(vacrel, NULL, VACUUM_ERRCB_PHASE_SCAN_HEAP,
1327 : blkno, InvalidOffsetNumber);
1328 :
1329 : /*
1330 : * Pin the visibility map page in case we need to mark the page
1331 : * all-visible. In most cases this will be very cheap, because we'll
1332 : * already have the correct page pinned anyway.
1333 : */
1334 877746 : visibilitymap_pin(vacrel->rel, blkno, &vmbuffer);
1335 :
1336 : /*
1337 : * We need a buffer cleanup lock to prune HOT chains and defragment
1338 : * the page in lazy_scan_prune. But when it's not possible to acquire
1339 : * a cleanup lock right away, we may be able to settle for reduced
1340 : * processing using lazy_scan_noprune.
1341 : */
1342 877746 : got_cleanup_lock = ConditionalLockBufferForCleanup(buf);
1343 :
1344 877746 : if (!got_cleanup_lock)
1345 408 : LockBuffer(buf, BUFFER_LOCK_SHARE);
1346 :
1347 : /* Check for new or empty pages before lazy_scan_[no]prune call */
1348 877746 : if (lazy_scan_new_or_empty(vacrel, buf, blkno, page, !got_cleanup_lock,
1349 877746 : vmbuffer))
1350 : {
1351 : /* Processed as new/empty page (lock and pin released) */
1352 1224 : continue;
1353 : }
1354 :
1355 : /*
1356 : * If we didn't get the cleanup lock, we can still collect LP_DEAD
1357 : * items in the dead_items area for later vacuuming, count live and
1358 : * recently dead tuples for vacuum logging, and determine if this
1359 : * block could later be truncated. If we encounter any xid/mxids that
1360 : * require advancing the relfrozenxid/relminxid, we'll have to wait
1361 : * for a cleanup lock and call lazy_scan_prune().
1362 : */
1363 876522 : if (!got_cleanup_lock &&
1364 408 : !lazy_scan_noprune(vacrel, buf, blkno, page, &has_lpdead_items))
1365 : {
1366 : /*
1367 : * lazy_scan_noprune could not do all required processing. Wait
1368 : * for a cleanup lock, and call lazy_scan_prune in the usual way.
1369 : */
1370 : Assert(vacrel->aggressive);
1371 166 : LockBuffer(buf, BUFFER_LOCK_UNLOCK);
1372 166 : LockBufferForCleanup(buf);
1373 166 : got_cleanup_lock = true;
1374 : }
1375 :
1376 : /*
1377 : * If we have a cleanup lock, we must now prune, freeze, and count
1378 : * tuples. We may have acquired the cleanup lock originally, or we may
1379 : * have gone back and acquired it after lazy_scan_noprune() returned
1380 : * false. Either way, the page hasn't been processed yet.
1381 : *
1382 : * Like lazy_scan_noprune(), lazy_scan_prune() will count
1383 : * recently_dead_tuples and live tuples for vacuum logging, determine
1384 : * if the block can later be truncated, and accumulate the details of
1385 : * remaining LP_DEAD line pointers on the page into dead_items. These
1386 : * dead items include those pruned by lazy_scan_prune() as well as
1387 : * line pointers previously marked LP_DEAD.
1388 : */
1389 876522 : if (got_cleanup_lock)
1390 876280 : lazy_scan_prune(vacrel, buf, blkno, page,
1391 : vmbuffer,
1392 876280 : blk_info & VAC_BLK_ALL_VISIBLE_ACCORDING_TO_VM,
1393 : &has_lpdead_items, &vm_page_frozen);
1394 :
1395 : /*
1396 : * Count an eagerly scanned page as a failure or a success.
1397 : *
1398 : * Only lazy_scan_prune() freezes pages, so if we didn't get the
1399 : * cleanup lock, we won't have frozen the page. However, we only count
1400 : * pages that were too new to require freezing as eager freeze
1401 : * failures.
1402 : *
1403 : * We could gather more information from lazy_scan_noprune() about
1404 : * whether or not there were tuples with XIDs or MXIDs older than the
1405 : * FreezeLimit or MultiXactCutoff. However, for simplicity, we simply
1406 : * exclude pages skipped due to cleanup lock contention from eager
1407 : * freeze algorithm caps.
1408 : */
1409 876522 : if (got_cleanup_lock &&
1410 876280 : (blk_info & VAC_BLK_WAS_EAGER_SCANNED))
1411 : {
1412 : /* Aggressive vacuums do not eager scan. */
1413 : Assert(!vacrel->aggressive);
1414 :
1415 0 : if (vm_page_frozen)
1416 : {
1417 0 : if (vacrel->eager_scan_remaining_successes > 0)
1418 0 : vacrel->eager_scan_remaining_successes--;
1419 :
1420 0 : if (vacrel->eager_scan_remaining_successes == 0)
1421 : {
1422 : /*
1423 : * Report only once that we disabled eager scanning. We
1424 : * may eagerly read ahead blocks in excess of the success
1425 : * or failure caps before attempting to freeze them, so we
1426 : * could reach here even after disabling additional eager
1427 : * scanning.
1428 : */
1429 0 : if (vacrel->eager_scan_max_fails_per_region > 0)
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 : * If we hit our success cap, permanently disable eager
1438 : * scanning by setting the other eager scan management
1439 : * fields to their disabled values.
1440 : */
1441 0 : vacrel->eager_scan_remaining_fails = 0;
1442 0 : vacrel->next_eager_scan_region_start = InvalidBlockNumber;
1443 0 : vacrel->eager_scan_max_fails_per_region = 0;
1444 : }
1445 : }
1446 0 : else if (vacrel->eager_scan_remaining_fails > 0)
1447 0 : vacrel->eager_scan_remaining_fails--;
1448 : }
1449 :
1450 : /*
1451 : * Now drop the buffer lock and, potentially, update the FSM.
1452 : *
1453 : * Our goal is to update the freespace map the last time we touch the
1454 : * page. If we'll process a block in the second pass, we may free up
1455 : * additional space on the page, so it is better to update the FSM
1456 : * after the second pass. If the relation has no indexes, or if index
1457 : * vacuuming is disabled, there will be no second heap pass; if this
1458 : * particular page has no dead items, the second heap pass will not
1459 : * touch this page. So, in those cases, update the FSM now.
1460 : *
1461 : * Note: In corner cases, it's possible to miss updating the FSM
1462 : * entirely. If index vacuuming is currently enabled, we'll skip the
1463 : * FSM update now. But if failsafe mode is later activated, or there
1464 : * are so few dead tuples that index vacuuming is bypassed, there will
1465 : * also be no opportunity to update the FSM later, because we'll never
1466 : * revisit this page. Since updating the FSM is desirable but not
1467 : * absolutely required, that's OK.
1468 : */
1469 876522 : if (vacrel->nindexes == 0
1470 842356 : || !vacrel->do_index_vacuuming
1471 638188 : || !has_lpdead_items)
1472 851292 : {
1473 851292 : Size freespace = PageGetHeapFreeSpace(page);
1474 :
1475 851292 : UnlockReleaseBuffer(buf);
1476 851292 : RecordPageWithFreeSpace(vacrel->rel, blkno, freespace);
1477 :
1478 : /*
1479 : * Periodically perform FSM vacuuming to make newly-freed space
1480 : * visible on upper FSM pages. This is done after vacuuming if the
1481 : * table has indexes. There will only be newly-freed space if we
1482 : * held the cleanup lock and lazy_scan_prune() was called.
1483 : */
1484 851292 : if (got_cleanup_lock && vacrel->nindexes == 0 && has_lpdead_items &&
1485 0 : blkno - next_fsm_block_to_vacuum >= VACUUM_FSM_EVERY_PAGES)
1486 : {
1487 0 : FreeSpaceMapVacuumRange(vacrel->rel, next_fsm_block_to_vacuum,
1488 : blkno);
1489 0 : next_fsm_block_to_vacuum = blkno;
1490 : }
1491 : }
1492 : else
1493 25230 : UnlockReleaseBuffer(buf);
1494 : }
1495 :
1496 238900 : vacrel->blkno = InvalidBlockNumber;
1497 238900 : if (BufferIsValid(vmbuffer))
1498 95760 : ReleaseBuffer(vmbuffer);
1499 :
1500 : /*
1501 : * Report that everything is now scanned. We never skip scanning the last
1502 : * block in the relation, so we can pass rel_pages here.
1503 : */
1504 238900 : pgstat_progress_update_param(PROGRESS_VACUUM_HEAP_BLKS_SCANNED,
1505 : rel_pages);
1506 :
1507 : /* now we can compute the new value for pg_class.reltuples */
1508 477800 : vacrel->new_live_tuples = vac_estimate_reltuples(vacrel->rel, rel_pages,
1509 : vacrel->scanned_pages,
1510 238900 : vacrel->live_tuples);
1511 :
1512 : /*
1513 : * Also compute the total number of surviving heap entries. In the
1514 : * (unlikely) scenario that new_live_tuples is -1, take it as zero.
1515 : */
1516 238900 : vacrel->new_rel_tuples =
1517 238900 : Max(vacrel->new_live_tuples, 0) + vacrel->recently_dead_tuples +
1518 238900 : vacrel->missed_dead_tuples;
1519 :
1520 238900 : read_stream_end(stream);
1521 :
1522 : /*
1523 : * Do index vacuuming (call each index's ambulkdelete routine), then do
1524 : * related heap vacuuming
1525 : */
1526 238900 : if (vacrel->dead_items_info->num_items > 0)
1527 1290 : lazy_vacuum(vacrel);
1528 :
1529 : /*
1530 : * Vacuum the remainder of the Free Space Map. We must do this whether or
1531 : * not there were indexes, and whether or not we bypassed index vacuuming.
1532 : * We can pass rel_pages here because we never skip scanning the last
1533 : * block of the relation.
1534 : */
1535 238900 : if (rel_pages > next_fsm_block_to_vacuum)
1536 95760 : FreeSpaceMapVacuumRange(vacrel->rel, next_fsm_block_to_vacuum, rel_pages);
1537 :
1538 : /* report all blocks vacuumed */
1539 238900 : pgstat_progress_update_param(PROGRESS_VACUUM_HEAP_BLKS_VACUUMED, rel_pages);
1540 :
1541 : /* Do final index cleanup (call each index's amvacuumcleanup routine) */
1542 238900 : if (vacrel->nindexes > 0 && vacrel->do_index_cleanup)
1543 167132 : lazy_cleanup_all_indexes(vacrel);
1544 238900 : }
1545 :
1546 : /*
1547 : * heap_vac_scan_next_block() -- read stream callback to get the next block
1548 : * for vacuum to process
1549 : *
1550 : * Every time lazy_scan_heap() needs a new block to process during its first
1551 : * phase, it invokes read_stream_next_buffer() with a stream set up to call
1552 : * heap_vac_scan_next_block() to get the next block.
1553 : *
1554 : * heap_vac_scan_next_block() uses the visibility map, vacuum options, and
1555 : * various thresholds to skip blocks which do not need to be processed and
1556 : * returns the next block to process or InvalidBlockNumber if there are no
1557 : * remaining blocks.
1558 : *
1559 : * The visibility status of the next block to process and whether or not it
1560 : * was eager scanned is set in the per_buffer_data.
1561 : *
1562 : * callback_private_data contains a reference to the LVRelState, passed to the
1563 : * read stream API during stream setup. The LVRelState is an in/out parameter
1564 : * here (locally named `vacrel`). Vacuum options and information about the
1565 : * relation are read from it. vacrel->skippedallvis is set if we skip a block
1566 : * that's all-visible but not all-frozen (to ensure that we don't update
1567 : * relfrozenxid in that case). vacrel also holds information about the next
1568 : * unskippable block -- as bookkeeping for this function.
1569 : */
1570 : static BlockNumber
1571 1116646 : heap_vac_scan_next_block(ReadStream *stream,
1572 : void *callback_private_data,
1573 : void *per_buffer_data)
1574 : {
1575 : BlockNumber next_block;
1576 1116646 : LVRelState *vacrel = callback_private_data;
1577 1116646 : uint8 blk_info = 0;
1578 :
1579 : /* relies on InvalidBlockNumber + 1 overflowing to 0 on first call */
1580 1116646 : next_block = vacrel->current_block + 1;
1581 :
1582 : /* Have we reached the end of the relation? */
1583 1116646 : if (next_block >= vacrel->rel_pages)
1584 : {
1585 238900 : if (BufferIsValid(vacrel->next_unskippable_vmbuffer))
1586 : {
1587 92858 : ReleaseBuffer(vacrel->next_unskippable_vmbuffer);
1588 92858 : vacrel->next_unskippable_vmbuffer = InvalidBuffer;
1589 : }
1590 238900 : return InvalidBlockNumber;
1591 : }
1592 :
1593 : /*
1594 : * We must be in one of the three following states:
1595 : */
1596 877746 : if (next_block > vacrel->next_unskippable_block ||
1597 387182 : vacrel->next_unskippable_block == InvalidBlockNumber)
1598 : {
1599 : /*
1600 : * 1. We have just processed an unskippable block (or we're at the
1601 : * beginning of the scan). Find the next unskippable block using the
1602 : * visibility map.
1603 : */
1604 : bool skipsallvis;
1605 :
1606 586324 : find_next_unskippable_block(vacrel, &skipsallvis);
1607 :
1608 : /*
1609 : * We now know the next block that we must process. It can be the
1610 : * next block after the one we just processed, or something further
1611 : * ahead. If it's further ahead, we can jump to it, but we choose to
1612 : * do so only if we can skip at least SKIP_PAGES_THRESHOLD consecutive
1613 : * pages. Since we're reading sequentially, the OS should be doing
1614 : * readahead for us, so there's no gain in skipping a page now and
1615 : * then. Skipping such a range might even discourage sequential
1616 : * detection.
1617 : *
1618 : * This test also enables more frequent relfrozenxid advancement
1619 : * during non-aggressive VACUUMs. If the range has any all-visible
1620 : * pages then skipping makes updating relfrozenxid unsafe, which is a
1621 : * real downside.
1622 : */
1623 586324 : if (vacrel->next_unskippable_block - next_block >= SKIP_PAGES_THRESHOLD)
1624 : {
1625 8530 : next_block = vacrel->next_unskippable_block;
1626 8530 : if (skipsallvis)
1627 64 : vacrel->skippedallvis = true;
1628 : }
1629 : }
1630 :
1631 : /* Now we must be in one of the two remaining states: */
1632 877746 : if (next_block < vacrel->next_unskippable_block)
1633 : {
1634 : /*
1635 : * 2. We are processing a range of blocks that we could have skipped
1636 : * but chose not to. We know that they are all-visible in the VM,
1637 : * otherwise they would've been unskippable.
1638 : */
1639 291422 : vacrel->current_block = next_block;
1640 291422 : blk_info |= VAC_BLK_ALL_VISIBLE_ACCORDING_TO_VM;
1641 291422 : *((uint8 *) per_buffer_data) = blk_info;
1642 291422 : return vacrel->current_block;
1643 : }
1644 : else
1645 : {
1646 : /*
1647 : * 3. We reached the next unskippable block. Process it. On next
1648 : * iteration, we will be back in state 1.
1649 : */
1650 : Assert(next_block == vacrel->next_unskippable_block);
1651 :
1652 586324 : vacrel->current_block = next_block;
1653 586324 : if (vacrel->next_unskippable_allvis)
1654 87646 : blk_info |= VAC_BLK_ALL_VISIBLE_ACCORDING_TO_VM;
1655 586324 : if (vacrel->next_unskippable_eager_scanned)
1656 0 : blk_info |= VAC_BLK_WAS_EAGER_SCANNED;
1657 586324 : *((uint8 *) per_buffer_data) = blk_info;
1658 586324 : return vacrel->current_block;
1659 : }
1660 : }
1661 :
1662 : /*
1663 : * Find the next unskippable block in a vacuum scan using the visibility map.
1664 : * The next unskippable block and its visibility information is updated in
1665 : * vacrel.
1666 : *
1667 : * Note: our opinion of which blocks can be skipped can go stale immediately.
1668 : * It's okay if caller "misses" a page whose all-visible or all-frozen marking
1669 : * was concurrently cleared, though. All that matters is that caller scan all
1670 : * pages whose tuples might contain XIDs < OldestXmin, or MXIDs < OldestMxact.
1671 : * (Actually, non-aggressive VACUUMs can choose to skip all-visible pages with
1672 : * older XIDs/MXIDs. The *skippedallvis flag will be set here when the choice
1673 : * to skip such a range is actually made, making everything safe.)
1674 : */
1675 : static void
1676 586324 : find_next_unskippable_block(LVRelState *vacrel, bool *skipsallvis)
1677 : {
1678 586324 : BlockNumber rel_pages = vacrel->rel_pages;
1679 586324 : BlockNumber next_unskippable_block = vacrel->next_unskippable_block + 1;
1680 586324 : Buffer next_unskippable_vmbuffer = vacrel->next_unskippable_vmbuffer;
1681 586324 : bool next_unskippable_eager_scanned = false;
1682 : bool next_unskippable_allvis;
1683 :
1684 586324 : *skipsallvis = false;
1685 :
1686 889920 : for (;; next_unskippable_block++)
1687 889920 : {
1688 1476244 : uint8 mapbits = visibilitymap_get_status(vacrel->rel,
1689 : next_unskippable_block,
1690 : &next_unskippable_vmbuffer);
1691 :
1692 1476244 : next_unskippable_allvis = (mapbits & VISIBILITYMAP_ALL_VISIBLE) != 0;
1693 :
1694 : /*
1695 : * At the start of each eager scan region, normal vacuums with eager
1696 : * scanning enabled reset the failure counter, allowing vacuum to
1697 : * resume eager scanning if it had been suspended in the previous
1698 : * region.
1699 : */
1700 1476244 : if (next_unskippable_block >= vacrel->next_eager_scan_region_start)
1701 : {
1702 0 : vacrel->eager_scan_remaining_fails =
1703 0 : vacrel->eager_scan_max_fails_per_region;
1704 0 : vacrel->next_eager_scan_region_start += EAGER_SCAN_REGION_SIZE;
1705 : }
1706 :
1707 : /*
1708 : * A block is unskippable if it is not all visible according to the
1709 : * visibility map.
1710 : */
1711 1476244 : if (!next_unskippable_allvis)
1712 : {
1713 : Assert((mapbits & VISIBILITYMAP_ALL_FROZEN) == 0);
1714 498678 : break;
1715 : }
1716 :
1717 : /*
1718 : * Caller must scan the last page to determine whether it has tuples
1719 : * (caller must have the opportunity to set vacrel->nonempty_pages).
1720 : * This rule avoids having lazy_truncate_heap() take access-exclusive
1721 : * lock on rel to attempt a truncation that fails anyway, just because
1722 : * there are tuples on the last page (it is likely that there will be
1723 : * tuples on other nearby pages as well, but those can be skipped).
1724 : *
1725 : * Implement this by always treating the last block as unsafe to skip.
1726 : */
1727 977566 : if (next_unskippable_block == rel_pages - 1)
1728 86838 : break;
1729 :
1730 : /* DISABLE_PAGE_SKIPPING makes all skipping unsafe */
1731 890728 : if (!vacrel->skipwithvm)
1732 808 : break;
1733 :
1734 : /*
1735 : * All-frozen pages cannot contain XIDs < OldestXmin (XIDs that aren't
1736 : * already frozen by now), so this page can be skipped.
1737 : */
1738 889920 : if ((mapbits & VISIBILITYMAP_ALL_FROZEN) != 0)
1739 884562 : continue;
1740 :
1741 : /*
1742 : * Aggressive vacuums cannot skip any all-visible pages that are not
1743 : * also all-frozen.
1744 : */
1745 5358 : if (vacrel->aggressive)
1746 0 : break;
1747 :
1748 : /*
1749 : * Normal vacuums with eager scanning enabled only skip all-visible
1750 : * but not all-frozen pages if they have hit the failure limit for the
1751 : * current eager scan region.
1752 : */
1753 5358 : if (vacrel->eager_scan_remaining_fails > 0)
1754 : {
1755 0 : next_unskippable_eager_scanned = true;
1756 0 : break;
1757 : }
1758 :
1759 : /*
1760 : * All-visible blocks are safe to skip in a normal vacuum. But
1761 : * remember that the final range contains such a block for later.
1762 : */
1763 5358 : *skipsallvis = true;
1764 : }
1765 :
1766 : /* write the local variables back to vacrel */
1767 586324 : vacrel->next_unskippable_block = next_unskippable_block;
1768 586324 : vacrel->next_unskippable_allvis = next_unskippable_allvis;
1769 586324 : vacrel->next_unskippable_eager_scanned = next_unskippable_eager_scanned;
1770 586324 : vacrel->next_unskippable_vmbuffer = next_unskippable_vmbuffer;
1771 586324 : }
1772 :
1773 : /*
1774 : * lazy_scan_new_or_empty() -- lazy_scan_heap() new/empty page handling.
1775 : *
1776 : * Must call here to handle both new and empty pages before calling
1777 : * lazy_scan_prune or lazy_scan_noprune, since they're not prepared to deal
1778 : * with new or empty pages.
1779 : *
1780 : * It's necessary to consider new pages as a special case, since the rules for
1781 : * maintaining the visibility map and FSM with empty pages are a little
1782 : * different (though new pages can be truncated away during rel truncation).
1783 : *
1784 : * Empty pages are not really a special case -- they're just heap pages that
1785 : * have no allocated tuples (including even LP_UNUSED items). You might
1786 : * wonder why we need to handle them here all the same. It's only necessary
1787 : * because of a corner-case involving a hard crash during heap relation
1788 : * extension. If we ever make relation-extension crash safe, then it should
1789 : * no longer be necessary to deal with empty pages here (or new pages, for
1790 : * that matter).
1791 : *
1792 : * Caller must hold at least a shared lock. We might need to escalate the
1793 : * lock in that case, so the type of lock caller holds needs to be specified
1794 : * using 'sharelock' argument.
1795 : *
1796 : * Returns false in common case where caller should go on to call
1797 : * lazy_scan_prune (or lazy_scan_noprune). Otherwise returns true, indicating
1798 : * that lazy_scan_heap is done processing the page, releasing lock on caller's
1799 : * behalf.
1800 : *
1801 : * No vm_page_frozen output parameter (like that passed to lazy_scan_prune())
1802 : * is passed here because neither empty nor new pages can be eagerly frozen.
1803 : * New pages are never frozen. Empty pages are always set frozen in the VM at
1804 : * the same time that they are set all-visible, and we don't eagerly scan
1805 : * frozen pages.
1806 : */
1807 : static bool
1808 877746 : lazy_scan_new_or_empty(LVRelState *vacrel, Buffer buf, BlockNumber blkno,
1809 : Page page, bool sharelock, Buffer vmbuffer)
1810 : {
1811 : Size freespace;
1812 :
1813 877746 : if (PageIsNew(page))
1814 : {
1815 : /*
1816 : * All-zeroes pages can be left over if either a backend extends the
1817 : * relation by a single page, but crashes before the newly initialized
1818 : * page has been written out, or when bulk-extending the relation
1819 : * (which creates a number of empty pages at the tail end of the
1820 : * relation), and then enters them into the FSM.
1821 : *
1822 : * Note we do not enter the page into the visibilitymap. That has the
1823 : * downside that we repeatedly visit this page in subsequent vacuums,
1824 : * but otherwise we'll never discover the space on a promoted standby.
1825 : * The harm of repeated checking ought to normally not be too bad. The
1826 : * space usually should be used at some point, otherwise there
1827 : * wouldn't be any regular vacuums.
1828 : *
1829 : * Make sure these pages are in the FSM, to ensure they can be reused.
1830 : * Do that by testing if there's any space recorded for the page. If
1831 : * not, enter it. We do so after releasing the lock on the heap page,
1832 : * the FSM is approximate, after all.
1833 : */
1834 1170 : UnlockReleaseBuffer(buf);
1835 :
1836 1170 : if (GetRecordedFreeSpace(vacrel->rel, blkno) == 0)
1837 : {
1838 858 : freespace = BLCKSZ - SizeOfPageHeaderData;
1839 :
1840 858 : RecordPageWithFreeSpace(vacrel->rel, blkno, freespace);
1841 : }
1842 :
1843 1170 : return true;
1844 : }
1845 :
1846 876576 : if (PageIsEmpty(page))
1847 : {
1848 : /*
1849 : * It seems likely that caller will always be able to get a cleanup
1850 : * lock on an empty page. But don't take any chances -- escalate to
1851 : * an exclusive lock (still don't need a cleanup lock, though).
1852 : */
1853 54 : if (sharelock)
1854 : {
1855 0 : LockBuffer(buf, BUFFER_LOCK_UNLOCK);
1856 0 : LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
1857 :
1858 0 : if (!PageIsEmpty(page))
1859 : {
1860 : /* page isn't new or empty -- keep lock and pin for now */
1861 0 : return false;
1862 : }
1863 : }
1864 : else
1865 : {
1866 : /* Already have a full cleanup lock (which is more than enough) */
1867 : }
1868 :
1869 : /*
1870 : * Unlike new pages, empty pages are always set all-visible and
1871 : * all-frozen.
1872 : */
1873 54 : if (!PageIsAllVisible(page))
1874 : {
1875 0 : START_CRIT_SECTION();
1876 :
1877 : /* mark buffer dirty before writing a WAL record */
1878 0 : MarkBufferDirty(buf);
1879 :
1880 : /*
1881 : * It's possible that another backend has extended the heap,
1882 : * initialized the page, and then failed to WAL-log the page due
1883 : * to an ERROR. Since heap extension is not WAL-logged, recovery
1884 : * might try to replay our record setting the page all-visible and
1885 : * find that the page isn't initialized, which will cause a PANIC.
1886 : * To prevent that, check whether the page has been previously
1887 : * WAL-logged, and if not, do that now.
1888 : */
1889 0 : if (RelationNeedsWAL(vacrel->rel) &&
1890 0 : PageGetLSN(page) == InvalidXLogRecPtr)
1891 0 : log_newpage_buffer(buf, true);
1892 :
1893 0 : PageSetAllVisible(page);
1894 0 : visibilitymap_set(vacrel->rel, blkno, buf,
1895 : InvalidXLogRecPtr,
1896 : vmbuffer, InvalidTransactionId,
1897 : VISIBILITYMAP_ALL_VISIBLE |
1898 : VISIBILITYMAP_ALL_FROZEN);
1899 0 : END_CRIT_SECTION();
1900 :
1901 : /* Count the newly all-frozen pages for logging */
1902 0 : vacrel->vm_new_visible_pages++;
1903 0 : vacrel->vm_new_visible_frozen_pages++;
1904 : }
1905 :
1906 54 : freespace = PageGetHeapFreeSpace(page);
1907 54 : UnlockReleaseBuffer(buf);
1908 54 : RecordPageWithFreeSpace(vacrel->rel, blkno, freespace);
1909 54 : return true;
1910 : }
1911 :
1912 : /* page isn't new or empty -- keep lock and pin */
1913 876522 : return false;
1914 : }
1915 :
1916 : /* qsort comparator for sorting OffsetNumbers */
1917 : static int
1918 5282672 : cmpOffsetNumbers(const void *a, const void *b)
1919 : {
1920 5282672 : return pg_cmp_u16(*(const OffsetNumber *) a, *(const OffsetNumber *) b);
1921 : }
1922 :
1923 : /*
1924 : * lazy_scan_prune() -- lazy_scan_heap() pruning and freezing.
1925 : *
1926 : * Caller must hold pin and buffer cleanup lock on the buffer.
1927 : *
1928 : * vmbuffer is the buffer containing the VM block with visibility information
1929 : * for the heap block, blkno. all_visible_according_to_vm is the saved
1930 : * visibility status of the heap block looked up earlier by the caller. We
1931 : * won't rely entirely on this status, as it may be out of date.
1932 : *
1933 : * *has_lpdead_items is set to true or false depending on whether, upon return
1934 : * from this function, any LP_DEAD items are still present on the page.
1935 : *
1936 : * *vm_page_frozen is set to true if the page is newly set all-frozen in the
1937 : * VM. The caller currently only uses this for determining whether an eagerly
1938 : * scanned page was successfully set all-frozen.
1939 : */
1940 : static void
1941 876280 : lazy_scan_prune(LVRelState *vacrel,
1942 : Buffer buf,
1943 : BlockNumber blkno,
1944 : Page page,
1945 : Buffer vmbuffer,
1946 : bool all_visible_according_to_vm,
1947 : bool *has_lpdead_items,
1948 : bool *vm_page_frozen)
1949 : {
1950 876280 : Relation rel = vacrel->rel;
1951 : PruneFreezeResult presult;
1952 876280 : int prune_options = 0;
1953 :
1954 : Assert(BufferGetBlockNumber(buf) == blkno);
1955 :
1956 : /*
1957 : * Prune all HOT-update chains and potentially freeze tuples on this page.
1958 : *
1959 : * If the relation has no indexes, we can immediately mark would-be dead
1960 : * items LP_UNUSED.
1961 : *
1962 : * The number of tuples removed from the page is returned in
1963 : * presult.ndeleted. It should not be confused with presult.lpdead_items;
1964 : * presult.lpdead_items's final value can be thought of as the number of
1965 : * tuples that were deleted from indexes.
1966 : *
1967 : * We will update the VM after collecting LP_DEAD items and freezing
1968 : * tuples. Pruning will have determined whether or not the page is
1969 : * all-visible.
1970 : */
1971 876280 : prune_options = HEAP_PAGE_PRUNE_FREEZE;
1972 876280 : if (vacrel->nindexes == 0)
1973 34166 : prune_options |= HEAP_PAGE_PRUNE_MARK_UNUSED_NOW;
1974 :
1975 876280 : heap_page_prune_and_freeze(rel, buf, vacrel->vistest, prune_options,
1976 : &vacrel->cutoffs, &presult, PRUNE_VACUUM_SCAN,
1977 : &vacrel->offnum,
1978 : &vacrel->NewRelfrozenXid, &vacrel->NewRelminMxid);
1979 :
1980 : Assert(MultiXactIdIsValid(vacrel->NewRelminMxid));
1981 : Assert(TransactionIdIsValid(vacrel->NewRelfrozenXid));
1982 :
1983 876280 : if (presult.nfrozen > 0)
1984 : {
1985 : /*
1986 : * We don't increment the new_frozen_tuple_pages instrumentation
1987 : * counter when nfrozen == 0, since it only counts pages with newly
1988 : * frozen tuples (don't confuse that with pages newly set all-frozen
1989 : * in VM).
1990 : */
1991 45070 : vacrel->new_frozen_tuple_pages++;
1992 : }
1993 :
1994 : /*
1995 : * VACUUM will call heap_page_is_all_visible() during the second pass over
1996 : * the heap to determine all_visible and all_frozen for the page -- this
1997 : * is a specialized version of the logic from this function. Now that
1998 : * we've finished pruning and freezing, make sure that we're in total
1999 : * agreement with heap_page_is_all_visible() using an assertion.
2000 : */
2001 : #ifdef USE_ASSERT_CHECKING
2002 : /* Note that all_frozen value does not matter when !all_visible */
2003 : if (presult.all_visible)
2004 : {
2005 : TransactionId debug_cutoff;
2006 : bool debug_all_frozen;
2007 :
2008 : Assert(presult.lpdead_items == 0);
2009 :
2010 : if (!heap_page_is_all_visible(vacrel, buf,
2011 : &debug_cutoff, &debug_all_frozen))
2012 : Assert(false);
2013 :
2014 : Assert(presult.all_frozen == debug_all_frozen);
2015 :
2016 : Assert(!TransactionIdIsValid(debug_cutoff) ||
2017 : debug_cutoff == presult.vm_conflict_horizon);
2018 : }
2019 : #endif
2020 :
2021 : /*
2022 : * Now save details of the LP_DEAD items from the page in vacrel
2023 : */
2024 876280 : if (presult.lpdead_items > 0)
2025 : {
2026 29852 : vacrel->lpdead_item_pages++;
2027 :
2028 : /*
2029 : * deadoffsets are collected incrementally in
2030 : * heap_page_prune_and_freeze() as each dead line pointer is recorded,
2031 : * with an indeterminate order, but dead_items_add requires them to be
2032 : * sorted.
2033 : */
2034 29852 : qsort(presult.deadoffsets, presult.lpdead_items, sizeof(OffsetNumber),
2035 : cmpOffsetNumbers);
2036 :
2037 29852 : dead_items_add(vacrel, blkno, presult.deadoffsets, presult.lpdead_items);
2038 : }
2039 :
2040 : /* Finally, add page-local counts to whole-VACUUM counts */
2041 876280 : vacrel->tuples_deleted += presult.ndeleted;
2042 876280 : vacrel->tuples_frozen += presult.nfrozen;
2043 876280 : vacrel->lpdead_items += presult.lpdead_items;
2044 876280 : vacrel->live_tuples += presult.live_tuples;
2045 876280 : vacrel->recently_dead_tuples += presult.recently_dead_tuples;
2046 :
2047 : /* Can't truncate this page */
2048 876280 : if (presult.hastup)
2049 861924 : vacrel->nonempty_pages = blkno + 1;
2050 :
2051 : /* Did we find LP_DEAD items? */
2052 876280 : *has_lpdead_items = (presult.lpdead_items > 0);
2053 :
2054 : Assert(!presult.all_visible || !(*has_lpdead_items));
2055 :
2056 : /*
2057 : * Handle setting visibility map bit based on information from the VM (as
2058 : * of last heap_vac_scan_next_block() call), and from all_visible and
2059 : * all_frozen variables
2060 : */
2061 876280 : if (!all_visible_according_to_vm && presult.all_visible)
2062 69974 : {
2063 : uint8 old_vmbits;
2064 69974 : uint8 flags = VISIBILITYMAP_ALL_VISIBLE;
2065 :
2066 69974 : if (presult.all_frozen)
2067 : {
2068 : Assert(!TransactionIdIsValid(presult.vm_conflict_horizon));
2069 52530 : flags |= VISIBILITYMAP_ALL_FROZEN;
2070 : }
2071 :
2072 : /*
2073 : * It should never be the case that the visibility map page is set
2074 : * while the page-level bit is clear, but the reverse is allowed (if
2075 : * checksums are not enabled). Regardless, set both bits so that we
2076 : * get back in sync.
2077 : *
2078 : * NB: If the heap page is all-visible but the VM bit is not set, we
2079 : * don't need to dirty the heap page. However, if checksums are
2080 : * enabled, we do need to make sure that the heap page is dirtied
2081 : * before passing it to visibilitymap_set(), because it may be logged.
2082 : * Given that this situation should only happen in rare cases after a
2083 : * crash, it is not worth optimizing.
2084 : */
2085 69974 : PageSetAllVisible(page);
2086 69974 : MarkBufferDirty(buf);
2087 69974 : old_vmbits = visibilitymap_set(vacrel->rel, blkno, buf,
2088 : InvalidXLogRecPtr,
2089 : vmbuffer, presult.vm_conflict_horizon,
2090 : flags);
2091 :
2092 : /*
2093 : * If the page wasn't already set all-visible and/or all-frozen in the
2094 : * VM, count it as newly set for logging.
2095 : */
2096 69974 : if ((old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
2097 : {
2098 69974 : vacrel->vm_new_visible_pages++;
2099 69974 : if (presult.all_frozen)
2100 : {
2101 52530 : vacrel->vm_new_visible_frozen_pages++;
2102 52530 : *vm_page_frozen = true;
2103 : }
2104 : }
2105 0 : else if ((old_vmbits & VISIBILITYMAP_ALL_FROZEN) == 0 &&
2106 0 : presult.all_frozen)
2107 : {
2108 0 : vacrel->vm_new_frozen_pages++;
2109 0 : *vm_page_frozen = true;
2110 : }
2111 : }
2112 :
2113 : /*
2114 : * As of PostgreSQL 9.2, the visibility map bit should never be set if the
2115 : * page-level bit is clear. However, it's possible that the bit got
2116 : * cleared after heap_vac_scan_next_block() was called, so we must recheck
2117 : * with buffer lock before concluding that the VM is corrupt.
2118 : */
2119 806306 : else if (all_visible_according_to_vm && !PageIsAllVisible(page) &&
2120 0 : visibilitymap_get_status(vacrel->rel, blkno, &vmbuffer) != 0)
2121 : {
2122 0 : elog(WARNING, "page is not marked all-visible but visibility map bit is set in relation \"%s\" page %u",
2123 : vacrel->relname, blkno);
2124 0 : visibilitymap_clear(vacrel->rel, blkno, vmbuffer,
2125 : VISIBILITYMAP_VALID_BITS);
2126 : }
2127 :
2128 : /*
2129 : * It's possible for the value returned by
2130 : * GetOldestNonRemovableTransactionId() to move backwards, so it's not
2131 : * wrong for us to see tuples that appear to not be visible to everyone
2132 : * yet, while PD_ALL_VISIBLE is already set. The real safe xmin value
2133 : * never moves backwards, but GetOldestNonRemovableTransactionId() is
2134 : * conservative and sometimes returns a value that's unnecessarily small,
2135 : * so if we see that contradiction it just means that the tuples that we
2136 : * think are not visible to everyone yet actually are, and the
2137 : * PD_ALL_VISIBLE flag is correct.
2138 : *
2139 : * There should never be LP_DEAD items on a page with PD_ALL_VISIBLE set,
2140 : * however.
2141 : */
2142 806306 : else if (presult.lpdead_items > 0 && PageIsAllVisible(page))
2143 : {
2144 0 : elog(WARNING, "page containing LP_DEAD items is marked as all-visible in relation \"%s\" page %u",
2145 : vacrel->relname, blkno);
2146 0 : PageClearAllVisible(page);
2147 0 : MarkBufferDirty(buf);
2148 0 : visibilitymap_clear(vacrel->rel, blkno, vmbuffer,
2149 : VISIBILITYMAP_VALID_BITS);
2150 : }
2151 :
2152 : /*
2153 : * If the all-visible page is all-frozen but not marked as such yet, mark
2154 : * it as all-frozen. Note that all_frozen is only valid if all_visible is
2155 : * true, so we must check both all_visible and all_frozen.
2156 : */
2157 806306 : else if (all_visible_according_to_vm && presult.all_visible &&
2158 378790 : presult.all_frozen && !VM_ALL_FROZEN(vacrel->rel, blkno, &vmbuffer))
2159 : {
2160 : uint8 old_vmbits;
2161 :
2162 : /*
2163 : * Avoid relying on all_visible_according_to_vm as a proxy for the
2164 : * page-level PD_ALL_VISIBLE bit being set, since it might have become
2165 : * stale -- even when all_visible is set
2166 : */
2167 44 : if (!PageIsAllVisible(page))
2168 : {
2169 0 : PageSetAllVisible(page);
2170 0 : MarkBufferDirty(buf);
2171 : }
2172 :
2173 : /*
2174 : * Set the page all-frozen (and all-visible) in the VM.
2175 : *
2176 : * We can pass InvalidTransactionId as our cutoff_xid, since a
2177 : * snapshotConflictHorizon sufficient to make everything safe for REDO
2178 : * was logged when the page's tuples were frozen.
2179 : */
2180 : Assert(!TransactionIdIsValid(presult.vm_conflict_horizon));
2181 44 : old_vmbits = visibilitymap_set(vacrel->rel, blkno, buf,
2182 : InvalidXLogRecPtr,
2183 : vmbuffer, InvalidTransactionId,
2184 : VISIBILITYMAP_ALL_VISIBLE |
2185 : VISIBILITYMAP_ALL_FROZEN);
2186 :
2187 : /*
2188 : * The page was likely already set all-visible in the VM. However,
2189 : * there is a small chance that it was modified sometime between
2190 : * setting all_visible_according_to_vm and checking the visibility
2191 : * during pruning. Check the return value of old_vmbits anyway to
2192 : * ensure the visibility map counters used for logging are accurate.
2193 : */
2194 44 : if ((old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
2195 : {
2196 0 : vacrel->vm_new_visible_pages++;
2197 0 : vacrel->vm_new_visible_frozen_pages++;
2198 0 : *vm_page_frozen = true;
2199 : }
2200 :
2201 : /*
2202 : * We already checked that the page was not set all-frozen in the VM
2203 : * above, so we don't need to test the value of old_vmbits.
2204 : */
2205 : else
2206 : {
2207 44 : vacrel->vm_new_frozen_pages++;
2208 44 : *vm_page_frozen = true;
2209 : }
2210 : }
2211 876280 : }
2212 :
2213 : /*
2214 : * lazy_scan_noprune() -- lazy_scan_prune() without pruning or freezing
2215 : *
2216 : * Caller need only hold a pin and share lock on the buffer, unlike
2217 : * lazy_scan_prune, which requires a full cleanup lock. While pruning isn't
2218 : * performed here, it's quite possible that an earlier opportunistic pruning
2219 : * operation left LP_DEAD items behind. We'll at least collect any such items
2220 : * in dead_items for removal from indexes.
2221 : *
2222 : * For aggressive VACUUM callers, we may return false to indicate that a full
2223 : * cleanup lock is required for processing by lazy_scan_prune. This is only
2224 : * necessary when the aggressive VACUUM needs to freeze some tuple XIDs from
2225 : * one or more tuples on the page. We always return true for non-aggressive
2226 : * callers.
2227 : *
2228 : * If this function returns true, *has_lpdead_items gets set to true or false
2229 : * depending on whether, upon return from this function, any LP_DEAD items are
2230 : * present on the page. If this function returns false, *has_lpdead_items
2231 : * is not updated.
2232 : */
2233 : static bool
2234 408 : lazy_scan_noprune(LVRelState *vacrel,
2235 : Buffer buf,
2236 : BlockNumber blkno,
2237 : Page page,
2238 : bool *has_lpdead_items)
2239 : {
2240 : OffsetNumber offnum,
2241 : maxoff;
2242 : int lpdead_items,
2243 : live_tuples,
2244 : recently_dead_tuples,
2245 : missed_dead_tuples;
2246 : bool hastup;
2247 : HeapTupleHeader tupleheader;
2248 408 : TransactionId NoFreezePageRelfrozenXid = vacrel->NewRelfrozenXid;
2249 408 : MultiXactId NoFreezePageRelminMxid = vacrel->NewRelminMxid;
2250 : OffsetNumber deadoffsets[MaxHeapTuplesPerPage];
2251 :
2252 : Assert(BufferGetBlockNumber(buf) == blkno);
2253 :
2254 408 : hastup = false; /* for now */
2255 :
2256 408 : lpdead_items = 0;
2257 408 : live_tuples = 0;
2258 408 : recently_dead_tuples = 0;
2259 408 : missed_dead_tuples = 0;
2260 :
2261 408 : maxoff = PageGetMaxOffsetNumber(page);
2262 408 : for (offnum = FirstOffsetNumber;
2263 12774 : offnum <= maxoff;
2264 12366 : offnum = OffsetNumberNext(offnum))
2265 : {
2266 : ItemId itemid;
2267 : HeapTupleData tuple;
2268 :
2269 12532 : vacrel->offnum = offnum;
2270 12532 : itemid = PageGetItemId(page, offnum);
2271 :
2272 12532 : if (!ItemIdIsUsed(itemid))
2273 2538 : continue;
2274 :
2275 11122 : if (ItemIdIsRedirected(itemid))
2276 : {
2277 1128 : hastup = true;
2278 1128 : continue;
2279 : }
2280 :
2281 9994 : if (ItemIdIsDead(itemid))
2282 : {
2283 : /*
2284 : * Deliberately don't set hastup=true here. See same point in
2285 : * lazy_scan_prune for an explanation.
2286 : */
2287 0 : deadoffsets[lpdead_items++] = offnum;
2288 0 : continue;
2289 : }
2290 :
2291 9994 : hastup = true; /* page prevents rel truncation */
2292 9994 : tupleheader = (HeapTupleHeader) PageGetItem(page, itemid);
2293 9994 : if (heap_tuple_should_freeze(tupleheader, &vacrel->cutoffs,
2294 : &NoFreezePageRelfrozenXid,
2295 : &NoFreezePageRelminMxid))
2296 : {
2297 : /* Tuple with XID < FreezeLimit (or MXID < MultiXactCutoff) */
2298 294 : if (vacrel->aggressive)
2299 : {
2300 : /*
2301 : * Aggressive VACUUMs must always be able to advance rel's
2302 : * relfrozenxid to a value >= FreezeLimit (and be able to
2303 : * advance rel's relminmxid to a value >= MultiXactCutoff).
2304 : * The ongoing aggressive VACUUM won't be able to do that
2305 : * unless it can freeze an XID (or MXID) from this tuple now.
2306 : *
2307 : * The only safe option is to have caller perform processing
2308 : * of this page using lazy_scan_prune. Caller might have to
2309 : * wait a while for a cleanup lock, but it can't be helped.
2310 : */
2311 166 : vacrel->offnum = InvalidOffsetNumber;
2312 166 : return false;
2313 : }
2314 :
2315 : /*
2316 : * Non-aggressive VACUUMs are under no obligation to advance
2317 : * relfrozenxid (even by one XID). We can be much laxer here.
2318 : *
2319 : * Currently we always just accept an older final relfrozenxid
2320 : * and/or relminmxid value. We never make caller wait or work a
2321 : * little harder, even when it likely makes sense to do so.
2322 : */
2323 : }
2324 :
2325 9828 : ItemPointerSet(&(tuple.t_self), blkno, offnum);
2326 9828 : tuple.t_data = (HeapTupleHeader) PageGetItem(page, itemid);
2327 9828 : tuple.t_len = ItemIdGetLength(itemid);
2328 9828 : tuple.t_tableOid = RelationGetRelid(vacrel->rel);
2329 :
2330 9828 : switch (HeapTupleSatisfiesVacuum(&tuple, vacrel->cutoffs.OldestXmin,
2331 : buf))
2332 : {
2333 9820 : case HEAPTUPLE_DELETE_IN_PROGRESS:
2334 : case HEAPTUPLE_LIVE:
2335 :
2336 : /*
2337 : * Count both cases as live, just like lazy_scan_prune
2338 : */
2339 9820 : live_tuples++;
2340 :
2341 9820 : break;
2342 4 : case HEAPTUPLE_DEAD:
2343 :
2344 : /*
2345 : * There is some useful work for pruning to do, that won't be
2346 : * done due to failure to get a cleanup lock.
2347 : */
2348 4 : missed_dead_tuples++;
2349 4 : break;
2350 4 : case HEAPTUPLE_RECENTLY_DEAD:
2351 :
2352 : /*
2353 : * Count in recently_dead_tuples, just like lazy_scan_prune
2354 : */
2355 4 : recently_dead_tuples++;
2356 4 : break;
2357 0 : case HEAPTUPLE_INSERT_IN_PROGRESS:
2358 :
2359 : /*
2360 : * Do not count these rows as live, just like lazy_scan_prune
2361 : */
2362 0 : break;
2363 0 : default:
2364 0 : elog(ERROR, "unexpected HeapTupleSatisfiesVacuum result");
2365 : break;
2366 : }
2367 : }
2368 :
2369 242 : vacrel->offnum = InvalidOffsetNumber;
2370 :
2371 : /*
2372 : * By here we know for sure that caller can put off freezing and pruning
2373 : * this particular page until the next VACUUM. Remember its details now.
2374 : * (lazy_scan_prune expects a clean slate, so we have to do this last.)
2375 : */
2376 242 : vacrel->NewRelfrozenXid = NoFreezePageRelfrozenXid;
2377 242 : vacrel->NewRelminMxid = NoFreezePageRelminMxid;
2378 :
2379 : /* Save any LP_DEAD items found on the page in dead_items */
2380 242 : if (vacrel->nindexes == 0)
2381 : {
2382 : /* Using one-pass strategy (since table has no indexes) */
2383 0 : if (lpdead_items > 0)
2384 : {
2385 : /*
2386 : * Perfunctory handling for the corner case where a single pass
2387 : * strategy VACUUM cannot get a cleanup lock, and it turns out
2388 : * that there is one or more LP_DEAD items: just count the LP_DEAD
2389 : * items as missed_dead_tuples instead. (This is a bit dishonest,
2390 : * but it beats having to maintain specialized heap vacuuming code
2391 : * forever, for vanishingly little benefit.)
2392 : */
2393 0 : hastup = true;
2394 0 : missed_dead_tuples += lpdead_items;
2395 : }
2396 : }
2397 242 : else if (lpdead_items > 0)
2398 : {
2399 : /*
2400 : * Page has LP_DEAD items, and so any references/TIDs that remain in
2401 : * indexes will be deleted during index vacuuming (and then marked
2402 : * LP_UNUSED in the heap)
2403 : */
2404 0 : vacrel->lpdead_item_pages++;
2405 :
2406 0 : dead_items_add(vacrel, blkno, deadoffsets, lpdead_items);
2407 :
2408 0 : vacrel->lpdead_items += lpdead_items;
2409 : }
2410 :
2411 : /*
2412 : * Finally, add relevant page-local counts to whole-VACUUM counts
2413 : */
2414 242 : vacrel->live_tuples += live_tuples;
2415 242 : vacrel->recently_dead_tuples += recently_dead_tuples;
2416 242 : vacrel->missed_dead_tuples += missed_dead_tuples;
2417 242 : if (missed_dead_tuples > 0)
2418 4 : vacrel->missed_dead_pages++;
2419 :
2420 : /* Can't truncate this page */
2421 242 : if (hastup)
2422 242 : vacrel->nonempty_pages = blkno + 1;
2423 :
2424 : /* Did we find LP_DEAD items? */
2425 242 : *has_lpdead_items = (lpdead_items > 0);
2426 :
2427 : /* Caller won't need to call lazy_scan_prune with same page */
2428 242 : return true;
2429 : }
2430 :
2431 : /*
2432 : * Main entry point for index vacuuming and heap vacuuming.
2433 : *
2434 : * Removes items collected in dead_items from table's indexes, then marks the
2435 : * same items LP_UNUSED in the heap. See the comments above lazy_scan_heap
2436 : * for full details.
2437 : *
2438 : * Also empties dead_items, freeing up space for later TIDs.
2439 : *
2440 : * We may choose to bypass index vacuuming at this point, though only when the
2441 : * ongoing VACUUM operation will definitely only have one index scan/round of
2442 : * index vacuuming.
2443 : */
2444 : static void
2445 1294 : lazy_vacuum(LVRelState *vacrel)
2446 : {
2447 : bool bypass;
2448 :
2449 : /* Should not end up here with no indexes */
2450 : Assert(vacrel->nindexes > 0);
2451 : Assert(vacrel->lpdead_item_pages > 0);
2452 :
2453 1294 : if (!vacrel->do_index_vacuuming)
2454 : {
2455 : Assert(!vacrel->do_index_cleanup);
2456 30 : dead_items_reset(vacrel);
2457 30 : return;
2458 : }
2459 :
2460 : /*
2461 : * Consider bypassing index vacuuming (and heap vacuuming) entirely.
2462 : *
2463 : * We currently only do this in cases where the number of LP_DEAD items
2464 : * for the entire VACUUM operation is close to zero. This avoids sharp
2465 : * discontinuities in the duration and overhead of successive VACUUM
2466 : * operations that run against the same table with a fixed workload.
2467 : * Ideally, successive VACUUM operations will behave as if there are
2468 : * exactly zero LP_DEAD items in cases where there are close to zero.
2469 : *
2470 : * This is likely to be helpful with a table that is continually affected
2471 : * by UPDATEs that can mostly apply the HOT optimization, but occasionally
2472 : * have small aberrations that lead to just a few heap pages retaining
2473 : * only one or two LP_DEAD items. This is pretty common; even when the
2474 : * DBA goes out of their way to make UPDATEs use HOT, it is practically
2475 : * impossible to predict whether HOT will be applied in 100% of cases.
2476 : * It's far easier to ensure that 99%+ of all UPDATEs against a table use
2477 : * HOT through careful tuning.
2478 : */
2479 1264 : bypass = false;
2480 1264 : if (vacrel->consider_bypass_optimization && vacrel->rel_pages > 0)
2481 : {
2482 : BlockNumber threshold;
2483 :
2484 : Assert(vacrel->num_index_scans == 0);
2485 : Assert(vacrel->lpdead_items == vacrel->dead_items_info->num_items);
2486 : Assert(vacrel->do_index_vacuuming);
2487 : Assert(vacrel->do_index_cleanup);
2488 :
2489 : /*
2490 : * This crossover point at which we'll start to do index vacuuming is
2491 : * expressed as a percentage of the total number of heap pages in the
2492 : * table that are known to have at least one LP_DEAD item. This is
2493 : * much more important than the total number of LP_DEAD items, since
2494 : * it's a proxy for the number of heap pages whose visibility map bits
2495 : * cannot be set on account of bypassing index and heap vacuuming.
2496 : *
2497 : * We apply one further precautionary test: the space currently used
2498 : * to store the TIDs (TIDs that now all point to LP_DEAD items) must
2499 : * not exceed 32MB. This limits the risk that we will bypass index
2500 : * vacuuming again and again until eventually there is a VACUUM whose
2501 : * dead_items space is not CPU cache resident.
2502 : *
2503 : * We don't take any special steps to remember the LP_DEAD items (such
2504 : * as counting them in our final update to the stats system) when the
2505 : * optimization is applied. Though the accounting used in analyze.c's
2506 : * acquire_sample_rows() will recognize the same LP_DEAD items as dead
2507 : * rows in its own stats report, that's okay. The discrepancy should
2508 : * be negligible. If this optimization is ever expanded to cover more
2509 : * cases then this may need to be reconsidered.
2510 : */
2511 1234 : threshold = (double) vacrel->rel_pages * BYPASS_THRESHOLD_PAGES;
2512 1236 : bypass = (vacrel->lpdead_item_pages < threshold &&
2513 2 : TidStoreMemoryUsage(vacrel->dead_items) < 32 * 1024 * 1024);
2514 : }
2515 :
2516 1264 : if (bypass)
2517 : {
2518 : /*
2519 : * There are almost zero TIDs. Behave as if there were precisely
2520 : * zero: bypass index vacuuming, but do index cleanup.
2521 : *
2522 : * We expect that the ongoing VACUUM operation will finish very
2523 : * quickly, so there is no point in considering speeding up as a
2524 : * failsafe against wraparound failure. (Index cleanup is expected to
2525 : * finish very quickly in cases where there were no ambulkdelete()
2526 : * calls.)
2527 : */
2528 2 : vacrel->do_index_vacuuming = false;
2529 : }
2530 1262 : else if (lazy_vacuum_all_indexes(vacrel))
2531 : {
2532 : /*
2533 : * We successfully completed a round of index vacuuming. Do related
2534 : * heap vacuuming now.
2535 : */
2536 1262 : lazy_vacuum_heap_rel(vacrel);
2537 : }
2538 : else
2539 : {
2540 : /*
2541 : * Failsafe case.
2542 : *
2543 : * We attempted index vacuuming, but didn't finish a full round/full
2544 : * index scan. This happens when relfrozenxid or relminmxid is too
2545 : * far in the past.
2546 : *
2547 : * From this point on the VACUUM operation will do no further index
2548 : * vacuuming or heap vacuuming. This VACUUM operation won't end up
2549 : * back here again.
2550 : */
2551 : Assert(VacuumFailsafeActive);
2552 : }
2553 :
2554 : /*
2555 : * Forget the LP_DEAD items that we just vacuumed (or just decided to not
2556 : * vacuum)
2557 : */
2558 1264 : dead_items_reset(vacrel);
2559 : }
2560 :
2561 : /*
2562 : * lazy_vacuum_all_indexes() -- Main entry for index vacuuming
2563 : *
2564 : * Returns true in the common case when all indexes were successfully
2565 : * vacuumed. Returns false in rare cases where we determined that the ongoing
2566 : * VACUUM operation is at risk of taking too long to finish, leading to
2567 : * wraparound failure.
2568 : */
2569 : static bool
2570 1262 : lazy_vacuum_all_indexes(LVRelState *vacrel)
2571 : {
2572 1262 : bool allindexes = true;
2573 1262 : double old_live_tuples = vacrel->rel->rd_rel->reltuples;
2574 1262 : const int progress_start_index[] = {
2575 : PROGRESS_VACUUM_PHASE,
2576 : PROGRESS_VACUUM_INDEXES_TOTAL
2577 : };
2578 1262 : const int progress_end_index[] = {
2579 : PROGRESS_VACUUM_INDEXES_TOTAL,
2580 : PROGRESS_VACUUM_INDEXES_PROCESSED,
2581 : PROGRESS_VACUUM_NUM_INDEX_VACUUMS
2582 : };
2583 : int64 progress_start_val[2];
2584 : int64 progress_end_val[3];
2585 :
2586 : Assert(vacrel->nindexes > 0);
2587 : Assert(vacrel->do_index_vacuuming);
2588 : Assert(vacrel->do_index_cleanup);
2589 :
2590 : /* Precheck for XID wraparound emergencies */
2591 1262 : if (lazy_check_wraparound_failsafe(vacrel))
2592 : {
2593 : /* Wraparound emergency -- don't even start an index scan */
2594 0 : return false;
2595 : }
2596 :
2597 : /*
2598 : * Report that we are now vacuuming indexes and the number of indexes to
2599 : * vacuum.
2600 : */
2601 1262 : progress_start_val[0] = PROGRESS_VACUUM_PHASE_VACUUM_INDEX;
2602 1262 : progress_start_val[1] = vacrel->nindexes;
2603 1262 : pgstat_progress_update_multi_param(2, progress_start_index, progress_start_val);
2604 :
2605 1262 : if (!ParallelVacuumIsActive(vacrel))
2606 : {
2607 3666 : for (int idx = 0; idx < vacrel->nindexes; idx++)
2608 : {
2609 2416 : Relation indrel = vacrel->indrels[idx];
2610 2416 : IndexBulkDeleteResult *istat = vacrel->indstats[idx];
2611 :
2612 2416 : vacrel->indstats[idx] = lazy_vacuum_one_index(indrel, istat,
2613 : old_live_tuples,
2614 : vacrel);
2615 :
2616 : /* Report the number of indexes vacuumed */
2617 2416 : pgstat_progress_update_param(PROGRESS_VACUUM_INDEXES_PROCESSED,
2618 2416 : idx + 1);
2619 :
2620 2416 : if (lazy_check_wraparound_failsafe(vacrel))
2621 : {
2622 : /* Wraparound emergency -- end current index scan */
2623 0 : allindexes = false;
2624 0 : break;
2625 : }
2626 : }
2627 : }
2628 : else
2629 : {
2630 : /* Outsource everything to parallel variant */
2631 12 : parallel_vacuum_bulkdel_all_indexes(vacrel->pvs, old_live_tuples,
2632 : vacrel->num_index_scans);
2633 :
2634 : /*
2635 : * Do a postcheck to consider applying wraparound failsafe now. Note
2636 : * that parallel VACUUM only gets the precheck and this postcheck.
2637 : */
2638 12 : if (lazy_check_wraparound_failsafe(vacrel))
2639 0 : allindexes = false;
2640 : }
2641 :
2642 : /*
2643 : * We delete all LP_DEAD items from the first heap pass in all indexes on
2644 : * each call here (except calls where we choose to do the failsafe). This
2645 : * makes the next call to lazy_vacuum_heap_rel() safe (except in the event
2646 : * of the failsafe triggering, which prevents the next call from taking
2647 : * place).
2648 : */
2649 : Assert(vacrel->num_index_scans > 0 ||
2650 : vacrel->dead_items_info->num_items == vacrel->lpdead_items);
2651 : Assert(allindexes || VacuumFailsafeActive);
2652 :
2653 : /*
2654 : * Increase and report the number of index scans. Also, we reset
2655 : * PROGRESS_VACUUM_INDEXES_TOTAL and PROGRESS_VACUUM_INDEXES_PROCESSED.
2656 : *
2657 : * We deliberately include the case where we started a round of bulk
2658 : * deletes that we weren't able to finish due to the failsafe triggering.
2659 : */
2660 1262 : vacrel->num_index_scans++;
2661 1262 : progress_end_val[0] = 0;
2662 1262 : progress_end_val[1] = 0;
2663 1262 : progress_end_val[2] = vacrel->num_index_scans;
2664 1262 : pgstat_progress_update_multi_param(3, progress_end_index, progress_end_val);
2665 :
2666 1262 : return allindexes;
2667 : }
2668 :
2669 : /*
2670 : * Read stream callback for vacuum's third phase (second pass over the heap).
2671 : * Gets the next block from the TID store and returns it or InvalidBlockNumber
2672 : * if there are no further blocks to vacuum.
2673 : *
2674 : * NB: Assumed to be safe to use with READ_STREAM_USE_BATCHING.
2675 : */
2676 : static BlockNumber
2677 26490 : vacuum_reap_lp_read_stream_next(ReadStream *stream,
2678 : void *callback_private_data,
2679 : void *per_buffer_data)
2680 : {
2681 26490 : TidStoreIter *iter = callback_private_data;
2682 : TidStoreIterResult *iter_result;
2683 :
2684 26490 : iter_result = TidStoreIterateNext(iter);
2685 26490 : if (iter_result == NULL)
2686 1262 : return InvalidBlockNumber;
2687 :
2688 : /*
2689 : * Save the TidStoreIterResult for later, so we can extract the offsets.
2690 : * It is safe to copy the result, according to TidStoreIterateNext().
2691 : */
2692 25228 : memcpy(per_buffer_data, iter_result, sizeof(*iter_result));
2693 :
2694 25228 : return iter_result->blkno;
2695 : }
2696 :
2697 : /*
2698 : * lazy_vacuum_heap_rel() -- second pass over the heap for two pass strategy
2699 : *
2700 : * This routine marks LP_DEAD items in vacrel->dead_items as LP_UNUSED. Pages
2701 : * that never had lazy_scan_prune record LP_DEAD items are not visited at all.
2702 : *
2703 : * We may also be able to truncate the line pointer array of the heap pages we
2704 : * visit. If there is a contiguous group of LP_UNUSED items at the end of the
2705 : * array, it can be reclaimed as free space. These LP_UNUSED items usually
2706 : * start out as LP_DEAD items recorded by lazy_scan_prune (we set items from
2707 : * each page to LP_UNUSED, and then consider if it's possible to truncate the
2708 : * page's line pointer array).
2709 : *
2710 : * Note: the reason for doing this as a second pass is we cannot remove the
2711 : * tuples until we've removed their index entries, and we want to process
2712 : * index entry removal in batches as large as possible.
2713 : */
2714 : static void
2715 1262 : lazy_vacuum_heap_rel(LVRelState *vacrel)
2716 : {
2717 : ReadStream *stream;
2718 1262 : BlockNumber vacuumed_pages = 0;
2719 1262 : Buffer vmbuffer = InvalidBuffer;
2720 : LVSavedErrInfo saved_err_info;
2721 : TidStoreIter *iter;
2722 :
2723 : Assert(vacrel->do_index_vacuuming);
2724 : Assert(vacrel->do_index_cleanup);
2725 : Assert(vacrel->num_index_scans > 0);
2726 :
2727 : /* Report that we are now vacuuming the heap */
2728 1262 : pgstat_progress_update_param(PROGRESS_VACUUM_PHASE,
2729 : PROGRESS_VACUUM_PHASE_VACUUM_HEAP);
2730 :
2731 : /* Update error traceback information */
2732 1262 : update_vacuum_error_info(vacrel, &saved_err_info,
2733 : VACUUM_ERRCB_PHASE_VACUUM_HEAP,
2734 : InvalidBlockNumber, InvalidOffsetNumber);
2735 :
2736 1262 : iter = TidStoreBeginIterate(vacrel->dead_items);
2737 :
2738 : /*
2739 : * Set up the read stream for vacuum's second pass through the heap.
2740 : *
2741 : * It is safe to use batchmode, as vacuum_reap_lp_read_stream_next() does
2742 : * not need to wait for IO and does not perform locking. Once we support
2743 : * parallelism it should still be fine, as presumably the holder of locks
2744 : * would never be blocked by IO while holding the lock.
2745 : */
2746 1262 : stream = read_stream_begin_relation(READ_STREAM_MAINTENANCE |
2747 : READ_STREAM_USE_BATCHING,
2748 : vacrel->bstrategy,
2749 : vacrel->rel,
2750 : MAIN_FORKNUM,
2751 : vacuum_reap_lp_read_stream_next,
2752 : iter,
2753 : sizeof(TidStoreIterResult));
2754 :
2755 : while (true)
2756 25228 : {
2757 : BlockNumber blkno;
2758 : Buffer buf;
2759 : Page page;
2760 : TidStoreIterResult *iter_result;
2761 : Size freespace;
2762 : OffsetNumber offsets[MaxOffsetNumber];
2763 : int num_offsets;
2764 :
2765 26490 : vacuum_delay_point(false);
2766 :
2767 26490 : buf = read_stream_next_buffer(stream, (void **) &iter_result);
2768 :
2769 : /* The relation is exhausted */
2770 26490 : if (!BufferIsValid(buf))
2771 1262 : break;
2772 :
2773 25228 : vacrel->blkno = blkno = BufferGetBlockNumber(buf);
2774 :
2775 : Assert(iter_result);
2776 25228 : num_offsets = TidStoreGetBlockOffsets(iter_result, offsets, lengthof(offsets));
2777 : Assert(num_offsets <= lengthof(offsets));
2778 :
2779 : /*
2780 : * Pin the visibility map page in case we need to mark the page
2781 : * all-visible. In most cases this will be very cheap, because we'll
2782 : * already have the correct page pinned anyway.
2783 : */
2784 25228 : visibilitymap_pin(vacrel->rel, blkno, &vmbuffer);
2785 :
2786 : /* We need a non-cleanup exclusive lock to mark dead_items unused */
2787 25228 : LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
2788 25228 : lazy_vacuum_heap_page(vacrel, blkno, buf, offsets,
2789 : num_offsets, vmbuffer);
2790 :
2791 : /* Now that we've vacuumed the page, record its available space */
2792 25228 : page = BufferGetPage(buf);
2793 25228 : freespace = PageGetHeapFreeSpace(page);
2794 :
2795 25228 : UnlockReleaseBuffer(buf);
2796 25228 : RecordPageWithFreeSpace(vacrel->rel, blkno, freespace);
2797 25228 : vacuumed_pages++;
2798 : }
2799 :
2800 1262 : read_stream_end(stream);
2801 1262 : TidStoreEndIterate(iter);
2802 :
2803 1262 : vacrel->blkno = InvalidBlockNumber;
2804 1262 : if (BufferIsValid(vmbuffer))
2805 1262 : ReleaseBuffer(vmbuffer);
2806 :
2807 : /*
2808 : * We set all LP_DEAD items from the first heap pass to LP_UNUSED during
2809 : * the second heap pass. No more, no less.
2810 : */
2811 : Assert(vacrel->num_index_scans > 1 ||
2812 : (vacrel->dead_items_info->num_items == vacrel->lpdead_items &&
2813 : vacuumed_pages == vacrel->lpdead_item_pages));
2814 :
2815 1262 : ereport(DEBUG2,
2816 : (errmsg("table \"%s\": removed %" PRId64 " dead item identifiers in %u pages",
2817 : vacrel->relname, vacrel->dead_items_info->num_items,
2818 : vacuumed_pages)));
2819 :
2820 : /* Revert to the previous phase information for error traceback */
2821 1262 : restore_vacuum_error_info(vacrel, &saved_err_info);
2822 1262 : }
2823 :
2824 : /*
2825 : * lazy_vacuum_heap_page() -- free page's LP_DEAD items listed in the
2826 : * vacrel->dead_items store.
2827 : *
2828 : * Caller must have an exclusive buffer lock on the buffer (though a full
2829 : * cleanup lock is also acceptable). vmbuffer must be valid and already have
2830 : * a pin on blkno's visibility map page.
2831 : */
2832 : static void
2833 25228 : lazy_vacuum_heap_page(LVRelState *vacrel, BlockNumber blkno, Buffer buffer,
2834 : OffsetNumber *deadoffsets, int num_offsets,
2835 : Buffer vmbuffer)
2836 : {
2837 25228 : Page page = BufferGetPage(buffer);
2838 : OffsetNumber unused[MaxHeapTuplesPerPage];
2839 25228 : int nunused = 0;
2840 : TransactionId visibility_cutoff_xid;
2841 : bool all_frozen;
2842 : LVSavedErrInfo saved_err_info;
2843 :
2844 : Assert(vacrel->do_index_vacuuming);
2845 :
2846 25228 : pgstat_progress_update_param(PROGRESS_VACUUM_HEAP_BLKS_VACUUMED, blkno);
2847 :
2848 : /* Update error traceback information */
2849 25228 : update_vacuum_error_info(vacrel, &saved_err_info,
2850 : VACUUM_ERRCB_PHASE_VACUUM_HEAP, blkno,
2851 : InvalidOffsetNumber);
2852 :
2853 25228 : START_CRIT_SECTION();
2854 :
2855 1542238 : for (int i = 0; i < num_offsets; i++)
2856 : {
2857 : ItemId itemid;
2858 1517010 : OffsetNumber toff = deadoffsets[i];
2859 :
2860 1517010 : itemid = PageGetItemId(page, toff);
2861 :
2862 : Assert(ItemIdIsDead(itemid) && !ItemIdHasStorage(itemid));
2863 1517010 : ItemIdSetUnused(itemid);
2864 1517010 : unused[nunused++] = toff;
2865 : }
2866 :
2867 : Assert(nunused > 0);
2868 :
2869 : /* Attempt to truncate line pointer array now */
2870 25228 : PageTruncateLinePointerArray(page);
2871 :
2872 : /*
2873 : * Mark buffer dirty before we write WAL.
2874 : */
2875 25228 : MarkBufferDirty(buffer);
2876 :
2877 : /* XLOG stuff */
2878 25228 : if (RelationNeedsWAL(vacrel->rel))
2879 : {
2880 23528 : log_heap_prune_and_freeze(vacrel->rel, buffer,
2881 : InvalidTransactionId,
2882 : false, /* no cleanup lock required */
2883 : PRUNE_VACUUM_CLEANUP,
2884 : NULL, 0, /* frozen */
2885 : NULL, 0, /* redirected */
2886 : NULL, 0, /* dead */
2887 : unused, nunused);
2888 : }
2889 :
2890 : /*
2891 : * End critical section, so we safely can do visibility tests (which
2892 : * possibly need to perform IO and allocate memory!). If we crash now the
2893 : * page (including the corresponding vm bit) might not be marked all
2894 : * visible, but that's fine. A later vacuum will fix that.
2895 : */
2896 25228 : END_CRIT_SECTION();
2897 :
2898 : /*
2899 : * Now that we have removed the LP_DEAD items from the page, once again
2900 : * check if the page has become all-visible. The page is already marked
2901 : * dirty, exclusively locked, and, if needed, a full page image has been
2902 : * emitted.
2903 : */
2904 : Assert(!PageIsAllVisible(page));
2905 25228 : if (heap_page_is_all_visible(vacrel, buffer, &visibility_cutoff_xid,
2906 : &all_frozen))
2907 : {
2908 24938 : uint8 flags = VISIBILITYMAP_ALL_VISIBLE;
2909 :
2910 24938 : if (all_frozen)
2911 : {
2912 : Assert(!TransactionIdIsValid(visibility_cutoff_xid));
2913 19994 : flags |= VISIBILITYMAP_ALL_FROZEN;
2914 : }
2915 :
2916 24938 : PageSetAllVisible(page);
2917 24938 : visibilitymap_set(vacrel->rel, blkno, buffer,
2918 : InvalidXLogRecPtr,
2919 : vmbuffer, visibility_cutoff_xid,
2920 : flags);
2921 :
2922 : /* Count the newly set VM page for logging */
2923 24938 : vacrel->vm_new_visible_pages++;
2924 24938 : if (all_frozen)
2925 19994 : vacrel->vm_new_visible_frozen_pages++;
2926 : }
2927 :
2928 : /* Revert to the previous phase information for error traceback */
2929 25228 : restore_vacuum_error_info(vacrel, &saved_err_info);
2930 25228 : }
2931 :
2932 : /*
2933 : * Trigger the failsafe to avoid wraparound failure when vacrel table has a
2934 : * relfrozenxid and/or relminmxid that is dangerously far in the past.
2935 : * Triggering the failsafe makes the ongoing VACUUM bypass any further index
2936 : * vacuuming and heap vacuuming. Truncating the heap is also bypassed.
2937 : *
2938 : * Any remaining work (work that VACUUM cannot just bypass) is typically sped
2939 : * up when the failsafe triggers. VACUUM stops applying any cost-based delay
2940 : * that it started out with.
2941 : *
2942 : * Returns true when failsafe has been triggered.
2943 : */
2944 : static bool
2945 242590 : lazy_check_wraparound_failsafe(LVRelState *vacrel)
2946 : {
2947 : /* Don't warn more than once per VACUUM */
2948 242590 : if (VacuumFailsafeActive)
2949 0 : return true;
2950 :
2951 242590 : if (unlikely(vacuum_xid_failsafe_check(&vacrel->cutoffs)))
2952 : {
2953 62864 : const int progress_index[] = {
2954 : PROGRESS_VACUUM_INDEXES_TOTAL,
2955 : PROGRESS_VACUUM_INDEXES_PROCESSED
2956 : };
2957 62864 : int64 progress_val[2] = {0, 0};
2958 :
2959 62864 : VacuumFailsafeActive = true;
2960 :
2961 : /*
2962 : * Abandon use of a buffer access strategy to allow use of all of
2963 : * shared buffers. We assume the caller who allocated the memory for
2964 : * the BufferAccessStrategy will free it.
2965 : */
2966 62864 : vacrel->bstrategy = NULL;
2967 :
2968 : /* Disable index vacuuming, index cleanup, and heap rel truncation */
2969 62864 : vacrel->do_index_vacuuming = false;
2970 62864 : vacrel->do_index_cleanup = false;
2971 62864 : vacrel->do_rel_truncate = false;
2972 :
2973 : /* Reset the progress counters */
2974 62864 : pgstat_progress_update_multi_param(2, progress_index, progress_val);
2975 :
2976 62864 : ereport(WARNING,
2977 : (errmsg("bypassing nonessential maintenance of table \"%s.%s.%s\" as a failsafe after %d index scans",
2978 : vacrel->dbname, vacrel->relnamespace, vacrel->relname,
2979 : vacrel->num_index_scans),
2980 : errdetail("The table's relfrozenxid or relminmxid is too far in the past."),
2981 : errhint("Consider increasing configuration parameter \"maintenance_work_mem\" or \"autovacuum_work_mem\".\n"
2982 : "You might also need to consider other ways for VACUUM to keep up with the allocation of transaction IDs.")));
2983 :
2984 : /* Stop applying cost limits from this point on */
2985 62864 : VacuumCostActive = false;
2986 62864 : VacuumCostBalance = 0;
2987 :
2988 62864 : return true;
2989 : }
2990 :
2991 179726 : return false;
2992 : }
2993 :
2994 : /*
2995 : * lazy_cleanup_all_indexes() -- cleanup all indexes of relation.
2996 : */
2997 : static void
2998 167132 : lazy_cleanup_all_indexes(LVRelState *vacrel)
2999 : {
3000 167132 : double reltuples = vacrel->new_rel_tuples;
3001 167132 : bool estimated_count = vacrel->scanned_pages < vacrel->rel_pages;
3002 167132 : const int progress_start_index[] = {
3003 : PROGRESS_VACUUM_PHASE,
3004 : PROGRESS_VACUUM_INDEXES_TOTAL
3005 : };
3006 167132 : const int progress_end_index[] = {
3007 : PROGRESS_VACUUM_INDEXES_TOTAL,
3008 : PROGRESS_VACUUM_INDEXES_PROCESSED
3009 : };
3010 : int64 progress_start_val[2];
3011 167132 : int64 progress_end_val[2] = {0, 0};
3012 :
3013 : Assert(vacrel->do_index_cleanup);
3014 : Assert(vacrel->nindexes > 0);
3015 :
3016 : /*
3017 : * Report that we are now cleaning up indexes and the number of indexes to
3018 : * cleanup.
3019 : */
3020 167132 : progress_start_val[0] = PROGRESS_VACUUM_PHASE_INDEX_CLEANUP;
3021 167132 : progress_start_val[1] = vacrel->nindexes;
3022 167132 : pgstat_progress_update_multi_param(2, progress_start_index, progress_start_val);
3023 :
3024 167132 : if (!ParallelVacuumIsActive(vacrel))
3025 : {
3026 430438 : for (int idx = 0; idx < vacrel->nindexes; idx++)
3027 : {
3028 263340 : Relation indrel = vacrel->indrels[idx];
3029 263340 : IndexBulkDeleteResult *istat = vacrel->indstats[idx];
3030 :
3031 526680 : vacrel->indstats[idx] =
3032 263340 : lazy_cleanup_one_index(indrel, istat, reltuples,
3033 : estimated_count, vacrel);
3034 :
3035 : /* Report the number of indexes cleaned up */
3036 263340 : pgstat_progress_update_param(PROGRESS_VACUUM_INDEXES_PROCESSED,
3037 263340 : idx + 1);
3038 : }
3039 : }
3040 : else
3041 : {
3042 : /* Outsource everything to parallel variant */
3043 34 : parallel_vacuum_cleanup_all_indexes(vacrel->pvs, reltuples,
3044 : vacrel->num_index_scans,
3045 : estimated_count);
3046 : }
3047 :
3048 : /* Reset the progress counters */
3049 167132 : pgstat_progress_update_multi_param(2, progress_end_index, progress_end_val);
3050 167132 : }
3051 :
3052 : /*
3053 : * lazy_vacuum_one_index() -- vacuum index relation.
3054 : *
3055 : * Delete all the index tuples containing a TID collected in
3056 : * vacrel->dead_items. Also update running statistics. Exact
3057 : * details depend on index AM's ambulkdelete routine.
3058 : *
3059 : * reltuples is the number of heap tuples to be passed to the
3060 : * bulkdelete callback. It's always assumed to be estimated.
3061 : * See indexam.sgml for more info.
3062 : *
3063 : * Returns bulk delete stats derived from input stats
3064 : */
3065 : static IndexBulkDeleteResult *
3066 2416 : lazy_vacuum_one_index(Relation indrel, IndexBulkDeleteResult *istat,
3067 : double reltuples, LVRelState *vacrel)
3068 : {
3069 : IndexVacuumInfo ivinfo;
3070 : LVSavedErrInfo saved_err_info;
3071 :
3072 2416 : ivinfo.index = indrel;
3073 2416 : ivinfo.heaprel = vacrel->rel;
3074 2416 : ivinfo.analyze_only = false;
3075 2416 : ivinfo.report_progress = false;
3076 2416 : ivinfo.estimated_count = true;
3077 2416 : ivinfo.message_level = DEBUG2;
3078 2416 : ivinfo.num_heap_tuples = reltuples;
3079 2416 : ivinfo.strategy = vacrel->bstrategy;
3080 :
3081 : /*
3082 : * Update error traceback information.
3083 : *
3084 : * The index name is saved during this phase and restored immediately
3085 : * after this phase. See vacuum_error_callback.
3086 : */
3087 : Assert(vacrel->indname == NULL);
3088 2416 : vacrel->indname = pstrdup(RelationGetRelationName(indrel));
3089 2416 : update_vacuum_error_info(vacrel, &saved_err_info,
3090 : VACUUM_ERRCB_PHASE_VACUUM_INDEX,
3091 : InvalidBlockNumber, InvalidOffsetNumber);
3092 :
3093 : /* Do bulk deletion */
3094 2416 : istat = vac_bulkdel_one_index(&ivinfo, istat, vacrel->dead_items,
3095 : vacrel->dead_items_info);
3096 :
3097 : /* Revert to the previous phase information for error traceback */
3098 2416 : restore_vacuum_error_info(vacrel, &saved_err_info);
3099 2416 : pfree(vacrel->indname);
3100 2416 : vacrel->indname = NULL;
3101 :
3102 2416 : return istat;
3103 : }
3104 :
3105 : /*
3106 : * lazy_cleanup_one_index() -- do post-vacuum cleanup for index relation.
3107 : *
3108 : * Calls index AM's amvacuumcleanup routine. reltuples is the number
3109 : * of heap tuples and estimated_count is true if reltuples is an
3110 : * estimated value. See indexam.sgml for more info.
3111 : *
3112 : * Returns bulk delete stats derived from input stats
3113 : */
3114 : static IndexBulkDeleteResult *
3115 263340 : lazy_cleanup_one_index(Relation indrel, IndexBulkDeleteResult *istat,
3116 : double reltuples, bool estimated_count,
3117 : LVRelState *vacrel)
3118 : {
3119 : IndexVacuumInfo ivinfo;
3120 : LVSavedErrInfo saved_err_info;
3121 :
3122 263340 : ivinfo.index = indrel;
3123 263340 : ivinfo.heaprel = vacrel->rel;
3124 263340 : ivinfo.analyze_only = false;
3125 263340 : ivinfo.report_progress = false;
3126 263340 : ivinfo.estimated_count = estimated_count;
3127 263340 : ivinfo.message_level = DEBUG2;
3128 :
3129 263340 : ivinfo.num_heap_tuples = reltuples;
3130 263340 : ivinfo.strategy = vacrel->bstrategy;
3131 :
3132 : /*
3133 : * Update error traceback information.
3134 : *
3135 : * The index name is saved during this phase and restored immediately
3136 : * after this phase. See vacuum_error_callback.
3137 : */
3138 : Assert(vacrel->indname == NULL);
3139 263340 : vacrel->indname = pstrdup(RelationGetRelationName(indrel));
3140 263340 : update_vacuum_error_info(vacrel, &saved_err_info,
3141 : VACUUM_ERRCB_PHASE_INDEX_CLEANUP,
3142 : InvalidBlockNumber, InvalidOffsetNumber);
3143 :
3144 263340 : istat = vac_cleanup_one_index(&ivinfo, istat);
3145 :
3146 : /* Revert to the previous phase information for error traceback */
3147 263340 : restore_vacuum_error_info(vacrel, &saved_err_info);
3148 263340 : pfree(vacrel->indname);
3149 263340 : vacrel->indname = NULL;
3150 :
3151 263340 : return istat;
3152 : }
3153 :
3154 : /*
3155 : * should_attempt_truncation - should we attempt to truncate the heap?
3156 : *
3157 : * Don't even think about it unless we have a shot at releasing a goodly
3158 : * number of pages. Otherwise, the time taken isn't worth it, mainly because
3159 : * an AccessExclusive lock must be replayed on any hot standby, where it can
3160 : * be particularly disruptive.
3161 : *
3162 : * Also don't attempt it if wraparound failsafe is in effect. The entire
3163 : * system might be refusing to allocate new XIDs at this point. The system
3164 : * definitely won't return to normal unless and until VACUUM actually advances
3165 : * the oldest relfrozenxid -- which hasn't happened for target rel just yet.
3166 : * If lazy_truncate_heap attempted to acquire an AccessExclusiveLock to
3167 : * truncate the table under these circumstances, an XID exhaustion error might
3168 : * make it impossible for VACUUM to fix the underlying XID exhaustion problem.
3169 : * There is very little chance of truncation working out when the failsafe is
3170 : * in effect in any case. lazy_scan_prune makes the optimistic assumption
3171 : * that any LP_DEAD items it encounters will always be LP_UNUSED by the time
3172 : * we're called.
3173 : */
3174 : static bool
3175 238900 : should_attempt_truncation(LVRelState *vacrel)
3176 : {
3177 : BlockNumber possibly_freeable;
3178 :
3179 238900 : if (!vacrel->do_rel_truncate || VacuumFailsafeActive)
3180 63154 : return false;
3181 :
3182 175746 : possibly_freeable = vacrel->rel_pages - vacrel->nonempty_pages;
3183 175746 : if (possibly_freeable > 0 &&
3184 306 : (possibly_freeable >= REL_TRUNCATE_MINIMUM ||
3185 306 : possibly_freeable >= vacrel->rel_pages / REL_TRUNCATE_FRACTION))
3186 284 : return true;
3187 :
3188 175462 : return false;
3189 : }
3190 :
3191 : /*
3192 : * lazy_truncate_heap - try to truncate off any empty pages at the end
3193 : */
3194 : static void
3195 284 : lazy_truncate_heap(LVRelState *vacrel)
3196 : {
3197 284 : BlockNumber orig_rel_pages = vacrel->rel_pages;
3198 : BlockNumber new_rel_pages;
3199 : bool lock_waiter_detected;
3200 : int lock_retry;
3201 :
3202 : /* Report that we are now truncating */
3203 284 : pgstat_progress_update_param(PROGRESS_VACUUM_PHASE,
3204 : PROGRESS_VACUUM_PHASE_TRUNCATE);
3205 :
3206 : /* Update error traceback information one last time */
3207 284 : update_vacuum_error_info(vacrel, NULL, VACUUM_ERRCB_PHASE_TRUNCATE,
3208 : vacrel->nonempty_pages, InvalidOffsetNumber);
3209 :
3210 : /*
3211 : * Loop until no more truncating can be done.
3212 : */
3213 : do
3214 : {
3215 : /*
3216 : * We need full exclusive lock on the relation in order to do
3217 : * truncation. If we can't get it, give up rather than waiting --- we
3218 : * don't want to block other backends, and we don't want to deadlock
3219 : * (which is quite possible considering we already hold a lower-grade
3220 : * lock).
3221 : */
3222 284 : lock_waiter_detected = false;
3223 284 : lock_retry = 0;
3224 : while (true)
3225 : {
3226 684 : if (ConditionalLockRelation(vacrel->rel, AccessExclusiveLock))
3227 280 : break;
3228 :
3229 : /*
3230 : * Check for interrupts while trying to (re-)acquire the exclusive
3231 : * lock.
3232 : */
3233 404 : CHECK_FOR_INTERRUPTS();
3234 :
3235 404 : if (++lock_retry > (VACUUM_TRUNCATE_LOCK_TIMEOUT /
3236 : VACUUM_TRUNCATE_LOCK_WAIT_INTERVAL))
3237 : {
3238 : /*
3239 : * We failed to establish the lock in the specified number of
3240 : * retries. This means we give up truncating.
3241 : */
3242 4 : ereport(vacrel->verbose ? INFO : DEBUG2,
3243 : (errmsg("\"%s\": stopping truncate due to conflicting lock request",
3244 : vacrel->relname)));
3245 6 : return;
3246 : }
3247 :
3248 400 : (void) WaitLatch(MyLatch,
3249 : WL_LATCH_SET | WL_TIMEOUT | WL_EXIT_ON_PM_DEATH,
3250 : VACUUM_TRUNCATE_LOCK_WAIT_INTERVAL,
3251 : WAIT_EVENT_VACUUM_TRUNCATE);
3252 400 : ResetLatch(MyLatch);
3253 : }
3254 :
3255 : /*
3256 : * Now that we have exclusive lock, look to see if the rel has grown
3257 : * whilst we were vacuuming with non-exclusive lock. If so, give up;
3258 : * the newly added pages presumably contain non-deletable tuples.
3259 : */
3260 280 : new_rel_pages = RelationGetNumberOfBlocks(vacrel->rel);
3261 280 : if (new_rel_pages != orig_rel_pages)
3262 : {
3263 : /*
3264 : * Note: we intentionally don't update vacrel->rel_pages with the
3265 : * new rel size here. If we did, it would amount to assuming that
3266 : * the new pages are empty, which is unlikely. Leaving the numbers
3267 : * alone amounts to assuming that the new pages have the same
3268 : * tuple density as existing ones, which is less unlikely.
3269 : */
3270 0 : UnlockRelation(vacrel->rel, AccessExclusiveLock);
3271 0 : return;
3272 : }
3273 :
3274 : /*
3275 : * Scan backwards from the end to verify that the end pages actually
3276 : * contain no tuples. This is *necessary*, not optional, because
3277 : * other backends could have added tuples to these pages whilst we
3278 : * were vacuuming.
3279 : */
3280 280 : new_rel_pages = count_nondeletable_pages(vacrel, &lock_waiter_detected);
3281 280 : vacrel->blkno = new_rel_pages;
3282 :
3283 280 : if (new_rel_pages >= orig_rel_pages)
3284 : {
3285 : /* can't do anything after all */
3286 2 : UnlockRelation(vacrel->rel, AccessExclusiveLock);
3287 2 : return;
3288 : }
3289 :
3290 : /*
3291 : * Okay to truncate.
3292 : */
3293 278 : RelationTruncate(vacrel->rel, new_rel_pages);
3294 :
3295 : /*
3296 : * We can release the exclusive lock as soon as we have truncated.
3297 : * Other backends can't safely access the relation until they have
3298 : * processed the smgr invalidation that smgrtruncate sent out ... but
3299 : * that should happen as part of standard invalidation processing once
3300 : * they acquire lock on the relation.
3301 : */
3302 278 : UnlockRelation(vacrel->rel, AccessExclusiveLock);
3303 :
3304 : /*
3305 : * Update statistics. Here, it *is* correct to adjust rel_pages
3306 : * without also touching reltuples, since the tuple count wasn't
3307 : * changed by the truncation.
3308 : */
3309 278 : vacrel->removed_pages += orig_rel_pages - new_rel_pages;
3310 278 : vacrel->rel_pages = new_rel_pages;
3311 :
3312 278 : ereport(vacrel->verbose ? INFO : DEBUG2,
3313 : (errmsg("table \"%s\": truncated %u to %u pages",
3314 : vacrel->relname,
3315 : orig_rel_pages, new_rel_pages)));
3316 278 : orig_rel_pages = new_rel_pages;
3317 278 : } while (new_rel_pages > vacrel->nonempty_pages && lock_waiter_detected);
3318 : }
3319 :
3320 : /*
3321 : * Rescan end pages to verify that they are (still) empty of tuples.
3322 : *
3323 : * Returns number of nondeletable pages (last nonempty page + 1).
3324 : */
3325 : static BlockNumber
3326 280 : count_nondeletable_pages(LVRelState *vacrel, bool *lock_waiter_detected)
3327 : {
3328 : BlockNumber blkno;
3329 : BlockNumber prefetchedUntil;
3330 : instr_time starttime;
3331 :
3332 : /* Initialize the starttime if we check for conflicting lock requests */
3333 280 : INSTR_TIME_SET_CURRENT(starttime);
3334 :
3335 : /*
3336 : * Start checking blocks at what we believe relation end to be and move
3337 : * backwards. (Strange coding of loop control is needed because blkno is
3338 : * unsigned.) To make the scan faster, we prefetch a few blocks at a time
3339 : * in forward direction, so that OS-level readahead can kick in.
3340 : */
3341 280 : blkno = vacrel->rel_pages;
3342 : StaticAssertStmt((PREFETCH_SIZE & (PREFETCH_SIZE - 1)) == 0,
3343 : "prefetch size must be power of 2");
3344 280 : prefetchedUntil = InvalidBlockNumber;
3345 4142 : while (blkno > vacrel->nonempty_pages)
3346 : {
3347 : Buffer buf;
3348 : Page page;
3349 : OffsetNumber offnum,
3350 : maxoff;
3351 : bool hastup;
3352 :
3353 : /*
3354 : * Check if another process requests a lock on our relation. We are
3355 : * holding an AccessExclusiveLock here, so they will be waiting. We
3356 : * only do this once per VACUUM_TRUNCATE_LOCK_CHECK_INTERVAL, and we
3357 : * only check if that interval has elapsed once every 32 blocks to
3358 : * keep the number of system calls and actual shared lock table
3359 : * lookups to a minimum.
3360 : */
3361 3868 : if ((blkno % 32) == 0)
3362 : {
3363 : instr_time currenttime;
3364 : instr_time elapsed;
3365 :
3366 124 : INSTR_TIME_SET_CURRENT(currenttime);
3367 124 : elapsed = currenttime;
3368 124 : INSTR_TIME_SUBTRACT(elapsed, starttime);
3369 124 : if ((INSTR_TIME_GET_MICROSEC(elapsed) / 1000)
3370 : >= VACUUM_TRUNCATE_LOCK_CHECK_INTERVAL)
3371 : {
3372 0 : if (LockHasWaitersRelation(vacrel->rel, AccessExclusiveLock))
3373 : {
3374 0 : ereport(vacrel->verbose ? INFO : DEBUG2,
3375 : (errmsg("table \"%s\": suspending truncate due to conflicting lock request",
3376 : vacrel->relname)));
3377 :
3378 0 : *lock_waiter_detected = true;
3379 0 : return blkno;
3380 : }
3381 0 : starttime = currenttime;
3382 : }
3383 : }
3384 :
3385 : /*
3386 : * We don't insert a vacuum delay point here, because we have an
3387 : * exclusive lock on the table which we want to hold for as short a
3388 : * time as possible. We still need to check for interrupts however.
3389 : */
3390 3868 : CHECK_FOR_INTERRUPTS();
3391 :
3392 3868 : blkno--;
3393 :
3394 : /* If we haven't prefetched this lot yet, do so now. */
3395 3868 : if (prefetchedUntil > blkno)
3396 : {
3397 : BlockNumber prefetchStart;
3398 : BlockNumber pblkno;
3399 :
3400 368 : prefetchStart = blkno & ~(PREFETCH_SIZE - 1);
3401 5678 : for (pblkno = prefetchStart; pblkno <= blkno; pblkno++)
3402 : {
3403 5310 : PrefetchBuffer(vacrel->rel, MAIN_FORKNUM, pblkno);
3404 5310 : CHECK_FOR_INTERRUPTS();
3405 : }
3406 368 : prefetchedUntil = prefetchStart;
3407 : }
3408 :
3409 3868 : buf = ReadBufferExtended(vacrel->rel, MAIN_FORKNUM, blkno, RBM_NORMAL,
3410 : vacrel->bstrategy);
3411 :
3412 : /* In this phase we only need shared access to the buffer */
3413 3868 : LockBuffer(buf, BUFFER_LOCK_SHARE);
3414 :
3415 3868 : page = BufferGetPage(buf);
3416 :
3417 3868 : if (PageIsNew(page) || PageIsEmpty(page))
3418 : {
3419 1478 : UnlockReleaseBuffer(buf);
3420 1478 : continue;
3421 : }
3422 :
3423 2390 : hastup = false;
3424 2390 : maxoff = PageGetMaxOffsetNumber(page);
3425 2390 : for (offnum = FirstOffsetNumber;
3426 5080 : offnum <= maxoff;
3427 2690 : offnum = OffsetNumberNext(offnum))
3428 : {
3429 : ItemId itemid;
3430 :
3431 2696 : itemid = PageGetItemId(page, offnum);
3432 :
3433 : /*
3434 : * Note: any non-unused item should be taken as a reason to keep
3435 : * this page. Even an LP_DEAD item makes truncation unsafe, since
3436 : * we must not have cleaned out its index entries.
3437 : */
3438 2696 : if (ItemIdIsUsed(itemid))
3439 : {
3440 6 : hastup = true;
3441 6 : break; /* can stop scanning */
3442 : }
3443 : } /* scan along page */
3444 :
3445 2390 : UnlockReleaseBuffer(buf);
3446 :
3447 : /* Done scanning if we found a tuple here */
3448 2390 : if (hastup)
3449 6 : return blkno + 1;
3450 : }
3451 :
3452 : /*
3453 : * If we fall out of the loop, all the previously-thought-to-be-empty
3454 : * pages still are; we need not bother to look at the last known-nonempty
3455 : * page.
3456 : */
3457 274 : return vacrel->nonempty_pages;
3458 : }
3459 :
3460 : /*
3461 : * Allocate dead_items and dead_items_info (either using palloc, or in dynamic
3462 : * shared memory). Sets both in vacrel for caller.
3463 : *
3464 : * Also handles parallel initialization as part of allocating dead_items in
3465 : * DSM when required.
3466 : */
3467 : static void
3468 238900 : dead_items_alloc(LVRelState *vacrel, int nworkers)
3469 : {
3470 : VacDeadItemsInfo *dead_items_info;
3471 690738 : int vac_work_mem = AmAutoVacuumWorkerProcess() &&
3472 212938 : autovacuum_work_mem != -1 ?
3473 451838 : autovacuum_work_mem : maintenance_work_mem;
3474 :
3475 : /*
3476 : * Initialize state for a parallel vacuum. As of now, only one worker can
3477 : * be used for an index, so we invoke parallelism only if there are at
3478 : * least two indexes on a table.
3479 : */
3480 238900 : if (nworkers >= 0 && vacrel->nindexes > 1 && vacrel->do_index_vacuuming)
3481 : {
3482 : /*
3483 : * Since parallel workers cannot access data in temporary tables, we
3484 : * can't perform parallel vacuum on them.
3485 : */
3486 10296 : if (RelationUsesLocalBuffers(vacrel->rel))
3487 : {
3488 : /*
3489 : * Give warning only if the user explicitly tries to perform a
3490 : * parallel vacuum on the temporary table.
3491 : */
3492 6 : if (nworkers > 0)
3493 6 : ereport(WARNING,
3494 : (errmsg("disabling parallel option of vacuum on \"%s\" --- cannot vacuum temporary tables in parallel",
3495 : vacrel->relname)));
3496 : }
3497 : else
3498 10290 : vacrel->pvs = parallel_vacuum_init(vacrel->rel, vacrel->indrels,
3499 : vacrel->nindexes, nworkers,
3500 : vac_work_mem,
3501 10290 : vacrel->verbose ? INFO : DEBUG2,
3502 : vacrel->bstrategy);
3503 :
3504 : /*
3505 : * If parallel mode started, dead_items and dead_items_info spaces are
3506 : * allocated in DSM.
3507 : */
3508 10296 : if (ParallelVacuumIsActive(vacrel))
3509 : {
3510 34 : vacrel->dead_items = parallel_vacuum_get_dead_items(vacrel->pvs,
3511 : &vacrel->dead_items_info);
3512 34 : return;
3513 : }
3514 : }
3515 :
3516 : /*
3517 : * Serial VACUUM case. Allocate both dead_items and dead_items_info
3518 : * locally.
3519 : */
3520 :
3521 238866 : dead_items_info = (VacDeadItemsInfo *) palloc(sizeof(VacDeadItemsInfo));
3522 238866 : dead_items_info->max_bytes = vac_work_mem * (Size) 1024;
3523 238866 : dead_items_info->num_items = 0;
3524 238866 : vacrel->dead_items_info = dead_items_info;
3525 :
3526 238866 : vacrel->dead_items = TidStoreCreateLocal(dead_items_info->max_bytes, true);
3527 : }
3528 :
3529 : /*
3530 : * Add the given block number and offset numbers to dead_items.
3531 : */
3532 : static void
3533 29852 : dead_items_add(LVRelState *vacrel, BlockNumber blkno, OffsetNumber *offsets,
3534 : int num_offsets)
3535 : {
3536 29852 : const int prog_index[2] = {
3537 : PROGRESS_VACUUM_NUM_DEAD_ITEM_IDS,
3538 : PROGRESS_VACUUM_DEAD_TUPLE_BYTES
3539 : };
3540 : int64 prog_val[2];
3541 :
3542 29852 : TidStoreSetBlockOffsets(vacrel->dead_items, blkno, offsets, num_offsets);
3543 29852 : vacrel->dead_items_info->num_items += num_offsets;
3544 :
3545 : /* update the progress information */
3546 29852 : prog_val[0] = vacrel->dead_items_info->num_items;
3547 29852 : prog_val[1] = TidStoreMemoryUsage(vacrel->dead_items);
3548 29852 : pgstat_progress_update_multi_param(2, prog_index, prog_val);
3549 29852 : }
3550 :
3551 : /*
3552 : * Forget all collected dead items.
3553 : */
3554 : static void
3555 1294 : dead_items_reset(LVRelState *vacrel)
3556 : {
3557 1294 : if (ParallelVacuumIsActive(vacrel))
3558 : {
3559 12 : parallel_vacuum_reset_dead_items(vacrel->pvs);
3560 12 : return;
3561 : }
3562 :
3563 : /* Recreate the tidstore with the same max_bytes limitation */
3564 1282 : TidStoreDestroy(vacrel->dead_items);
3565 1282 : vacrel->dead_items = TidStoreCreateLocal(vacrel->dead_items_info->max_bytes, true);
3566 :
3567 : /* Reset the counter */
3568 1282 : vacrel->dead_items_info->num_items = 0;
3569 : }
3570 :
3571 : /*
3572 : * Perform cleanup for resources allocated in dead_items_alloc
3573 : */
3574 : static void
3575 238900 : dead_items_cleanup(LVRelState *vacrel)
3576 : {
3577 238900 : if (!ParallelVacuumIsActive(vacrel))
3578 : {
3579 : /* Don't bother with pfree here */
3580 238866 : return;
3581 : }
3582 :
3583 : /* End parallel mode */
3584 34 : parallel_vacuum_end(vacrel->pvs, vacrel->indstats);
3585 34 : vacrel->pvs = NULL;
3586 : }
3587 :
3588 : /*
3589 : * Check if every tuple in the given page is visible to all current and future
3590 : * transactions. Also return the visibility_cutoff_xid which is the highest
3591 : * xmin amongst the visible tuples. Set *all_frozen to true if every tuple
3592 : * on this page is frozen.
3593 : *
3594 : * This is a stripped down version of lazy_scan_prune(). If you change
3595 : * anything here, make sure that everything stays in sync. Note that an
3596 : * assertion calls us to verify that everybody still agrees. Be sure to avoid
3597 : * introducing new side-effects here.
3598 : */
3599 : static bool
3600 25228 : heap_page_is_all_visible(LVRelState *vacrel, Buffer buf,
3601 : TransactionId *visibility_cutoff_xid,
3602 : bool *all_frozen)
3603 : {
3604 25228 : Page page = BufferGetPage(buf);
3605 25228 : BlockNumber blockno = BufferGetBlockNumber(buf);
3606 : OffsetNumber offnum,
3607 : maxoff;
3608 25228 : bool all_visible = true;
3609 :
3610 25228 : *visibility_cutoff_xid = InvalidTransactionId;
3611 25228 : *all_frozen = true;
3612 :
3613 25228 : maxoff = PageGetMaxOffsetNumber(page);
3614 25228 : for (offnum = FirstOffsetNumber;
3615 1246106 : offnum <= maxoff && all_visible;
3616 1220878 : offnum = OffsetNumberNext(offnum))
3617 : {
3618 : ItemId itemid;
3619 : HeapTupleData tuple;
3620 :
3621 : /*
3622 : * Set the offset number so that we can display it along with any
3623 : * error that occurred while processing this tuple.
3624 : */
3625 1220878 : vacrel->offnum = offnum;
3626 1220878 : itemid = PageGetItemId(page, offnum);
3627 :
3628 : /* Unused or redirect line pointers are of no interest */
3629 1220878 : if (!ItemIdIsUsed(itemid) || ItemIdIsRedirected(itemid))
3630 320926 : continue;
3631 :
3632 899952 : ItemPointerSet(&(tuple.t_self), blockno, offnum);
3633 :
3634 : /*
3635 : * Dead line pointers can have index pointers pointing to them. So
3636 : * they can't be treated as visible
3637 : */
3638 899952 : if (ItemIdIsDead(itemid))
3639 : {
3640 0 : all_visible = false;
3641 0 : *all_frozen = false;
3642 0 : break;
3643 : }
3644 :
3645 : Assert(ItemIdIsNormal(itemid));
3646 :
3647 899952 : tuple.t_data = (HeapTupleHeader) PageGetItem(page, itemid);
3648 899952 : tuple.t_len = ItemIdGetLength(itemid);
3649 899952 : tuple.t_tableOid = RelationGetRelid(vacrel->rel);
3650 :
3651 899952 : switch (HeapTupleSatisfiesVacuum(&tuple, vacrel->cutoffs.OldestXmin,
3652 : buf))
3653 : {
3654 899762 : case HEAPTUPLE_LIVE:
3655 : {
3656 : TransactionId xmin;
3657 :
3658 : /* Check comments in lazy_scan_prune. */
3659 899762 : if (!HeapTupleHeaderXminCommitted(tuple.t_data))
3660 : {
3661 0 : all_visible = false;
3662 0 : *all_frozen = false;
3663 0 : break;
3664 : }
3665 :
3666 : /*
3667 : * The inserter definitely committed. But is it old enough
3668 : * that everyone sees it as committed?
3669 : */
3670 899762 : xmin = HeapTupleHeaderGetXmin(tuple.t_data);
3671 899762 : if (!TransactionIdPrecedes(xmin,
3672 : vacrel->cutoffs.OldestXmin))
3673 : {
3674 100 : all_visible = false;
3675 100 : *all_frozen = false;
3676 100 : break;
3677 : }
3678 :
3679 : /* Track newest xmin on page. */
3680 899662 : if (TransactionIdFollows(xmin, *visibility_cutoff_xid) &&
3681 : TransactionIdIsNormal(xmin))
3682 16048 : *visibility_cutoff_xid = xmin;
3683 :
3684 : /* Check whether this tuple is already frozen or not */
3685 1181904 : if (all_visible && *all_frozen &&
3686 282242 : heap_tuple_needs_eventual_freeze(tuple.t_data))
3687 4986 : *all_frozen = false;
3688 : }
3689 899662 : break;
3690 :
3691 190 : case HEAPTUPLE_DEAD:
3692 : case HEAPTUPLE_RECENTLY_DEAD:
3693 : case HEAPTUPLE_INSERT_IN_PROGRESS:
3694 : case HEAPTUPLE_DELETE_IN_PROGRESS:
3695 : {
3696 190 : all_visible = false;
3697 190 : *all_frozen = false;
3698 190 : break;
3699 : }
3700 0 : default:
3701 0 : elog(ERROR, "unexpected HeapTupleSatisfiesVacuum result");
3702 : break;
3703 : }
3704 : } /* scan along page */
3705 :
3706 : /* Clear the offset information once we have processed the given page. */
3707 25228 : vacrel->offnum = InvalidOffsetNumber;
3708 :
3709 25228 : return all_visible;
3710 : }
3711 :
3712 : /*
3713 : * Update index statistics in pg_class if the statistics are accurate.
3714 : */
3715 : static void
3716 175776 : update_relstats_all_indexes(LVRelState *vacrel)
3717 : {
3718 175776 : Relation *indrels = vacrel->indrels;
3719 175776 : int nindexes = vacrel->nindexes;
3720 175776 : IndexBulkDeleteResult **indstats = vacrel->indstats;
3721 :
3722 : Assert(vacrel->do_index_cleanup);
3723 :
3724 439238 : for (int idx = 0; idx < nindexes; idx++)
3725 : {
3726 263462 : Relation indrel = indrels[idx];
3727 263462 : IndexBulkDeleteResult *istat = indstats[idx];
3728 :
3729 263462 : if (istat == NULL || istat->estimated_count)
3730 260772 : continue;
3731 :
3732 : /* Update index statistics */
3733 2690 : vac_update_relstats(indrel,
3734 : istat->num_pages,
3735 : istat->num_index_tuples,
3736 : 0, 0,
3737 : false,
3738 : InvalidTransactionId,
3739 : InvalidMultiXactId,
3740 : NULL, NULL, false);
3741 : }
3742 175776 : }
3743 :
3744 : /*
3745 : * Error context callback for errors occurring during vacuum. The error
3746 : * context messages for index phases should match the messages set in parallel
3747 : * vacuum. If you change this function for those phases, change
3748 : * parallel_vacuum_error_callback() as well.
3749 : */
3750 : static void
3751 236764 : vacuum_error_callback(void *arg)
3752 : {
3753 236764 : LVRelState *errinfo = arg;
3754 :
3755 236764 : switch (errinfo->phase)
3756 : {
3757 0 : case VACUUM_ERRCB_PHASE_SCAN_HEAP:
3758 0 : if (BlockNumberIsValid(errinfo->blkno))
3759 : {
3760 0 : if (OffsetNumberIsValid(errinfo->offnum))
3761 0 : errcontext("while scanning block %u offset %u of relation \"%s.%s\"",
3762 0 : errinfo->blkno, errinfo->offnum, errinfo->relnamespace, errinfo->relname);
3763 : else
3764 0 : errcontext("while scanning block %u of relation \"%s.%s\"",
3765 : errinfo->blkno, errinfo->relnamespace, errinfo->relname);
3766 : }
3767 : else
3768 0 : errcontext("while scanning relation \"%s.%s\"",
3769 : errinfo->relnamespace, errinfo->relname);
3770 0 : break;
3771 :
3772 0 : case VACUUM_ERRCB_PHASE_VACUUM_HEAP:
3773 0 : if (BlockNumberIsValid(errinfo->blkno))
3774 : {
3775 0 : if (OffsetNumberIsValid(errinfo->offnum))
3776 0 : errcontext("while vacuuming block %u offset %u of relation \"%s.%s\"",
3777 0 : errinfo->blkno, errinfo->offnum, errinfo->relnamespace, errinfo->relname);
3778 : else
3779 0 : errcontext("while vacuuming block %u of relation \"%s.%s\"",
3780 : errinfo->blkno, errinfo->relnamespace, errinfo->relname);
3781 : }
3782 : else
3783 0 : errcontext("while vacuuming relation \"%s.%s\"",
3784 : errinfo->relnamespace, errinfo->relname);
3785 0 : break;
3786 :
3787 0 : case VACUUM_ERRCB_PHASE_VACUUM_INDEX:
3788 0 : errcontext("while vacuuming index \"%s\" of relation \"%s.%s\"",
3789 : errinfo->indname, errinfo->relnamespace, errinfo->relname);
3790 0 : break;
3791 :
3792 0 : case VACUUM_ERRCB_PHASE_INDEX_CLEANUP:
3793 0 : errcontext("while cleaning up index \"%s\" of relation \"%s.%s\"",
3794 : errinfo->indname, errinfo->relnamespace, errinfo->relname);
3795 0 : break;
3796 :
3797 6 : case VACUUM_ERRCB_PHASE_TRUNCATE:
3798 6 : if (BlockNumberIsValid(errinfo->blkno))
3799 6 : errcontext("while truncating relation \"%s.%s\" to %u blocks",
3800 : errinfo->relnamespace, errinfo->relname, errinfo->blkno);
3801 6 : break;
3802 :
3803 236758 : case VACUUM_ERRCB_PHASE_UNKNOWN:
3804 : default:
3805 236758 : return; /* do nothing; the errinfo may not be
3806 : * initialized */
3807 : }
3808 : }
3809 :
3810 : /*
3811 : * Updates the information required for vacuum error callback. This also saves
3812 : * the current information which can be later restored via restore_vacuum_error_info.
3813 : */
3814 : static void
3815 1170276 : update_vacuum_error_info(LVRelState *vacrel, LVSavedErrInfo *saved_vacrel,
3816 : int phase, BlockNumber blkno, OffsetNumber offnum)
3817 : {
3818 1170276 : if (saved_vacrel)
3819 : {
3820 292246 : saved_vacrel->offnum = vacrel->offnum;
3821 292246 : saved_vacrel->blkno = vacrel->blkno;
3822 292246 : saved_vacrel->phase = vacrel->phase;
3823 : }
3824 :
3825 1170276 : vacrel->blkno = blkno;
3826 1170276 : vacrel->offnum = offnum;
3827 1170276 : vacrel->phase = phase;
3828 1170276 : }
3829 :
3830 : /*
3831 : * Restores the vacuum information saved via a prior call to update_vacuum_error_info.
3832 : */
3833 : static void
3834 292246 : restore_vacuum_error_info(LVRelState *vacrel,
3835 : const LVSavedErrInfo *saved_vacrel)
3836 : {
3837 292246 : vacrel->blkno = saved_vacrel->blkno;
3838 292246 : vacrel->offnum = saved_vacrel->offnum;
3839 292246 : vacrel->phase = saved_vacrel->phase;
3840 292246 : }
|
