Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * snapbuild.c
4 : *
5 : * Infrastructure for building historic catalog snapshots based on contents
6 : * of the WAL, for the purpose of decoding heapam.c style values in the
7 : * WAL.
8 : *
9 : * NOTES:
10 : *
11 : * We build snapshots which can *only* be used to read catalog contents and we
12 : * do so by reading and interpreting the WAL stream. The aim is to build a
13 : * snapshot that behaves the same as a freshly taken MVCC snapshot would have
14 : * at the time the XLogRecord was generated.
15 : *
16 : * To build the snapshots we reuse the infrastructure built for Hot
17 : * Standby. The in-memory snapshots we build look different than HS' because
18 : * we have different needs. To successfully decode data from the WAL we only
19 : * need to access catalog tables and (sys|rel|cat)cache, not the actual user
20 : * tables since the data we decode is wholly contained in the WAL
21 : * records. Also, our snapshots need to be different in comparison to normal
22 : * MVCC ones because in contrast to those we cannot fully rely on the clog and
23 : * pg_subtrans for information about committed transactions because they might
24 : * commit in the future from the POV of the WAL entry we're currently
25 : * decoding. This definition has the advantage that we only need to prevent
26 : * removal of catalog rows, while normal table's rows can still be
27 : * removed. This is achieved by using the replication slot mechanism.
28 : *
29 : * As the percentage of transactions modifying the catalog normally is fairly
30 : * small in comparisons to ones only manipulating user data, we keep track of
31 : * the committed catalog modifying ones inside [xmin, xmax) instead of keeping
32 : * track of all running transactions like it's done in a normal snapshot. Note
33 : * that we're generally only looking at transactions that have acquired an
34 : * xid. That is we keep a list of transactions between snapshot->(xmin, xmax)
35 : * that we consider committed, everything else is considered aborted/in
36 : * progress. That also allows us not to care about subtransactions before they
37 : * have committed which means this module, in contrast to HS, doesn't have to
38 : * care about suboverflowed subtransactions and similar.
39 : *
40 : * One complexity of doing this is that to e.g. handle mixed DDL/DML
41 : * transactions we need Snapshots that see intermediate versions of the
42 : * catalog in a transaction. During normal operation this is achieved by using
43 : * CommandIds/cmin/cmax. The problem with that however is that for space
44 : * efficiency reasons, the cmin and cmax are not included in WAL records. We
45 : * cannot read the cmin/cmax from the tuple itself, either, because it is
46 : * reset on crash recovery. Even if we could, we could not decode combocids
47 : * which are only tracked in the original backend's memory. To work around
48 : * that, heapam writes an extra WAL record (XLOG_HEAP2_NEW_CID) every time a
49 : * catalog row is modified, which includes the cmin and cmax of the
50 : * tuple. During decoding, we insert the ctid->(cmin,cmax) mappings into the
51 : * reorder buffer, and use them at visibility checks instead of the cmin/cmax
52 : * on the tuple itself. Check the reorderbuffer.c's comment above
53 : * ResolveCminCmaxDuringDecoding() for details.
54 : *
55 : * To facilitate all this we need our own visibility routine, as the normal
56 : * ones are optimized for different usecases.
57 : *
58 : * To replace the normal catalog snapshots with decoding ones use the
59 : * SetupHistoricSnapshot() and TeardownHistoricSnapshot() functions.
60 : *
61 : *
62 : *
63 : * The snapbuild machinery is starting up in several stages, as illustrated
64 : * by the following graph describing the SnapBuild->state transitions:
65 : *
66 : * +-------------------------+
67 : * +----| START |-------------+
68 : * | +-------------------------+ |
69 : * | | |
70 : * | | |
71 : * | running_xacts #1 |
72 : * | | |
73 : * | | |
74 : * | v |
75 : * | +-------------------------+ v
76 : * | | BUILDING_SNAPSHOT |------------>|
77 : * | +-------------------------+ |
78 : * | | |
79 : * | | |
80 : * | running_xacts #2, xacts from #1 finished |
81 : * | | |
82 : * | | |
83 : * | v |
84 : * | +-------------------------+ v
85 : * | | FULL_SNAPSHOT |------------>|
86 : * | +-------------------------+ |
87 : * | | |
88 : * running_xacts | saved snapshot
89 : * with zero xacts | at running_xacts's lsn
90 : * | | |
91 : * | running_xacts with xacts from #2 finished |
92 : * | | |
93 : * | v |
94 : * | +-------------------------+ |
95 : * +--->|SNAPBUILD_CONSISTENT |<------------+
96 : * +-------------------------+
97 : *
98 : * Initially the machinery is in the START stage. When an xl_running_xacts
99 : * record is read that is sufficiently new (above the safe xmin horizon),
100 : * there's a state transition. If there were no running xacts when the
101 : * xl_running_xacts record was generated, we'll directly go into CONSISTENT
102 : * state, otherwise we'll switch to the BUILDING_SNAPSHOT state. Having a full
103 : * snapshot means that all transactions that start henceforth can be decoded
104 : * in their entirety, but transactions that started previously can't. In
105 : * FULL_SNAPSHOT we'll switch into CONSISTENT once all those previously
106 : * running transactions have committed or aborted.
107 : *
108 : * Only transactions that commit after CONSISTENT state has been reached will
109 : * be replayed, even though they might have started while still in
110 : * FULL_SNAPSHOT. That ensures that we'll reach a point where no previous
111 : * changes has been exported, but all the following ones will be. That point
112 : * is a convenient point to initialize replication from, which is why we
113 : * export a snapshot at that point, which *can* be used to read normal data.
114 : *
115 : * Copyright (c) 2012-2025, PostgreSQL Global Development Group
116 : *
117 : * IDENTIFICATION
118 : * src/backend/replication/logical/snapbuild.c
119 : *
120 : *-------------------------------------------------------------------------
121 : */
122 :
123 : #include "postgres.h"
124 :
125 : #include <sys/stat.h>
126 : #include <unistd.h>
127 :
128 : #include "access/heapam_xlog.h"
129 : #include "access/transam.h"
130 : #include "access/xact.h"
131 : #include "common/file_utils.h"
132 : #include "miscadmin.h"
133 : #include "pgstat.h"
134 : #include "replication/logical.h"
135 : #include "replication/reorderbuffer.h"
136 : #include "replication/snapbuild.h"
137 : #include "replication/snapbuild_internal.h"
138 : #include "storage/fd.h"
139 : #include "storage/lmgr.h"
140 : #include "storage/proc.h"
141 : #include "storage/procarray.h"
142 : #include "storage/standby.h"
143 : #include "utils/builtins.h"
144 : #include "utils/memutils.h"
145 : #include "utils/snapmgr.h"
146 : #include "utils/snapshot.h"
147 : /*
148 : * Starting a transaction -- which we need to do while exporting a snapshot --
149 : * removes knowledge about the previously used resowner, so we save it here.
150 : */
151 : static ResourceOwner SavedResourceOwnerDuringExport = NULL;
152 : static bool ExportInProgress = false;
153 :
154 : /* ->committed and ->catchange manipulation */
155 : static void SnapBuildPurgeOlderTxn(SnapBuild *builder);
156 :
157 : /* snapshot building/manipulation/distribution functions */
158 : static Snapshot SnapBuildBuildSnapshot(SnapBuild *builder);
159 :
160 : static void SnapBuildFreeSnapshot(Snapshot snap);
161 :
162 : static void SnapBuildSnapIncRefcount(Snapshot snap);
163 :
164 : static void SnapBuildDistributeSnapshotAndInval(SnapBuild *builder, XLogRecPtr lsn, TransactionId xid);
165 :
166 : static inline bool SnapBuildXidHasCatalogChanges(SnapBuild *builder, TransactionId xid,
167 : uint32 xinfo);
168 :
169 : /* xlog reading helper functions for SnapBuildProcessRunningXacts */
170 : static bool SnapBuildFindSnapshot(SnapBuild *builder, XLogRecPtr lsn, xl_running_xacts *running);
171 : static void SnapBuildWaitSnapshot(xl_running_xacts *running, TransactionId cutoff);
172 :
173 : /* serialization functions */
174 : static void SnapBuildSerialize(SnapBuild *builder, XLogRecPtr lsn);
175 : static bool SnapBuildRestore(SnapBuild *builder, XLogRecPtr lsn);
176 : static void SnapBuildRestoreContents(int fd, void *dest, Size size, const char *path);
177 :
178 : /*
179 : * Allocate a new snapshot builder.
180 : *
181 : * xmin_horizon is the xid >= which we can be sure no catalog rows have been
182 : * removed, start_lsn is the LSN >= we want to replay commits.
183 : */
184 : SnapBuild *
185 2210 : AllocateSnapshotBuilder(ReorderBuffer *reorder,
186 : TransactionId xmin_horizon,
187 : XLogRecPtr start_lsn,
188 : bool need_full_snapshot,
189 : bool in_slot_creation,
190 : XLogRecPtr two_phase_at)
191 : {
192 : MemoryContext context;
193 : MemoryContext oldcontext;
194 : SnapBuild *builder;
195 :
196 : /* allocate memory in own context, to have better accountability */
197 2210 : context = AllocSetContextCreate(CurrentMemoryContext,
198 : "snapshot builder context",
199 : ALLOCSET_DEFAULT_SIZES);
200 2210 : oldcontext = MemoryContextSwitchTo(context);
201 :
202 2210 : builder = palloc0_object(SnapBuild);
203 :
204 2210 : builder->state = SNAPBUILD_START;
205 2210 : builder->context = context;
206 2210 : builder->reorder = reorder;
207 : /* Other struct members initialized by zeroing via palloc0 above */
208 :
209 2210 : builder->committed.xcnt = 0;
210 2210 : builder->committed.xcnt_space = 128; /* arbitrary number */
211 2210 : builder->committed.xip =
212 2210 : palloc0(builder->committed.xcnt_space * sizeof(TransactionId));
213 2210 : builder->committed.includes_all_transactions = true;
214 :
215 2210 : builder->catchange.xcnt = 0;
216 2210 : builder->catchange.xip = NULL;
217 :
218 2210 : builder->initial_xmin_horizon = xmin_horizon;
219 2210 : builder->start_decoding_at = start_lsn;
220 2210 : builder->in_slot_creation = in_slot_creation;
221 2210 : builder->building_full_snapshot = need_full_snapshot;
222 2210 : builder->two_phase_at = two_phase_at;
223 :
224 2210 : MemoryContextSwitchTo(oldcontext);
225 :
226 2210 : return builder;
227 : }
228 :
229 : /*
230 : * Free a snapshot builder.
231 : */
232 : void
233 1754 : FreeSnapshotBuilder(SnapBuild *builder)
234 : {
235 1754 : MemoryContext context = builder->context;
236 :
237 : /* free snapshot explicitly, that contains some error checking */
238 1754 : if (builder->snapshot != NULL)
239 : {
240 426 : SnapBuildSnapDecRefcount(builder->snapshot);
241 426 : builder->snapshot = NULL;
242 : }
243 :
244 : /* other resources are deallocated via memory context reset */
245 1754 : MemoryContextDelete(context);
246 1754 : }
247 :
248 : /*
249 : * Free an unreferenced snapshot that has previously been built by us.
250 : */
251 : static void
252 2738 : SnapBuildFreeSnapshot(Snapshot snap)
253 : {
254 : /* make sure we don't get passed an external snapshot */
255 : Assert(snap->snapshot_type == SNAPSHOT_HISTORIC_MVCC);
256 :
257 : /* make sure nobody modified our snapshot */
258 : Assert(snap->curcid == FirstCommandId);
259 : Assert(!snap->suboverflowed);
260 : Assert(!snap->takenDuringRecovery);
261 : Assert(snap->regd_count == 0);
262 :
263 : /* slightly more likely, so it's checked even without c-asserts */
264 2738 : if (snap->copied)
265 0 : elog(ERROR, "cannot free a copied snapshot");
266 :
267 2738 : if (snap->active_count)
268 0 : elog(ERROR, "cannot free an active snapshot");
269 :
270 2738 : pfree(snap);
271 2738 : }
272 :
273 : /*
274 : * In which state of snapshot building are we?
275 : */
276 : SnapBuildState
277 4241770 : SnapBuildCurrentState(SnapBuild *builder)
278 : {
279 4241770 : return builder->state;
280 : }
281 :
282 : /*
283 : * Return the LSN at which the two-phase decoding was first enabled.
284 : */
285 : XLogRecPtr
286 62 : SnapBuildGetTwoPhaseAt(SnapBuild *builder)
287 : {
288 62 : return builder->two_phase_at;
289 : }
290 :
291 : /*
292 : * Set the LSN at which two-phase decoding is enabled.
293 : */
294 : void
295 16 : SnapBuildSetTwoPhaseAt(SnapBuild *builder, XLogRecPtr ptr)
296 : {
297 16 : builder->two_phase_at = ptr;
298 16 : }
299 :
300 : /*
301 : * Should the contents of transaction ending at 'ptr' be decoded?
302 : */
303 : bool
304 728384 : SnapBuildXactNeedsSkip(SnapBuild *builder, XLogRecPtr ptr)
305 : {
306 728384 : return ptr < builder->start_decoding_at;
307 : }
308 :
309 : /*
310 : * Increase refcount of a snapshot.
311 : *
312 : * This is used when handing out a snapshot to some external resource or when
313 : * adding a Snapshot as builder->snapshot.
314 : */
315 : static void
316 12118 : SnapBuildSnapIncRefcount(Snapshot snap)
317 : {
318 12118 : snap->active_count++;
319 12118 : }
320 :
321 : /*
322 : * Decrease refcount of a snapshot and free if the refcount reaches zero.
323 : *
324 : * Externally visible, so that external resources that have been handed an
325 : * IncRef'ed Snapshot can adjust its refcount easily.
326 : */
327 : void
328 11560 : SnapBuildSnapDecRefcount(Snapshot snap)
329 : {
330 : /* make sure we don't get passed an external snapshot */
331 : Assert(snap->snapshot_type == SNAPSHOT_HISTORIC_MVCC);
332 :
333 : /* make sure nobody modified our snapshot */
334 : Assert(snap->curcid == FirstCommandId);
335 : Assert(!snap->suboverflowed);
336 : Assert(!snap->takenDuringRecovery);
337 :
338 : Assert(snap->regd_count == 0);
339 :
340 : Assert(snap->active_count > 0);
341 :
342 : /* slightly more likely, so it's checked even without casserts */
343 11560 : if (snap->copied)
344 0 : elog(ERROR, "cannot free a copied snapshot");
345 :
346 11560 : snap->active_count--;
347 11560 : if (snap->active_count == 0)
348 2738 : SnapBuildFreeSnapshot(snap);
349 11560 : }
350 :
351 : /*
352 : * Build a new snapshot, based on currently committed catalog-modifying
353 : * transactions.
354 : *
355 : * In-progress transactions with catalog access are *not* allowed to modify
356 : * these snapshots; they have to copy them and fill in appropriate ->curcid
357 : * and ->subxip/subxcnt values.
358 : */
359 : static Snapshot
360 3608 : SnapBuildBuildSnapshot(SnapBuild *builder)
361 : {
362 : Snapshot snapshot;
363 : Size ssize;
364 :
365 : Assert(builder->state >= SNAPBUILD_FULL_SNAPSHOT);
366 :
367 3608 : ssize = sizeof(SnapshotData)
368 3608 : + sizeof(TransactionId) * builder->committed.xcnt
369 3608 : + sizeof(TransactionId) * 1 /* toplevel xid */ ;
370 :
371 3608 : snapshot = MemoryContextAllocZero(builder->context, ssize);
372 :
373 3608 : snapshot->snapshot_type = SNAPSHOT_HISTORIC_MVCC;
374 :
375 : /*
376 : * We misuse the original meaning of SnapshotData's xip and subxip fields
377 : * to make the more fitting for our needs.
378 : *
379 : * In the 'xip' array we store transactions that have to be treated as
380 : * committed. Since we will only ever look at tuples from transactions
381 : * that have modified the catalog it's more efficient to store those few
382 : * that exist between xmin and xmax (frequently there are none).
383 : *
384 : * Snapshots that are used in transactions that have modified the catalog
385 : * also use the 'subxip' array to store their toplevel xid and all the
386 : * subtransaction xids so we can recognize when we need to treat rows as
387 : * visible that are not in xip but still need to be visible. Subxip only
388 : * gets filled when the transaction is copied into the context of a
389 : * catalog modifying transaction since we otherwise share a snapshot
390 : * between transactions. As long as a txn hasn't modified the catalog it
391 : * doesn't need to treat any uncommitted rows as visible, so there is no
392 : * need for those xids.
393 : *
394 : * Both arrays are qsort'ed so that we can use bsearch() on them.
395 : */
396 : Assert(TransactionIdIsNormal(builder->xmin));
397 : Assert(TransactionIdIsNormal(builder->xmax));
398 :
399 3608 : snapshot->xmin = builder->xmin;
400 3608 : snapshot->xmax = builder->xmax;
401 :
402 : /* store all transactions to be treated as committed by this snapshot */
403 3608 : snapshot->xip =
404 3608 : (TransactionId *) ((char *) snapshot + sizeof(SnapshotData));
405 3608 : snapshot->xcnt = builder->committed.xcnt;
406 3608 : memcpy(snapshot->xip,
407 3608 : builder->committed.xip,
408 3608 : builder->committed.xcnt * sizeof(TransactionId));
409 :
410 : /* sort so we can bsearch() */
411 3608 : qsort(snapshot->xip, snapshot->xcnt, sizeof(TransactionId), xidComparator);
412 :
413 : /*
414 : * Initially, subxip is empty, i.e. it's a snapshot to be used by
415 : * transactions that don't modify the catalog. Will be filled by
416 : * ReorderBufferCopySnap() if necessary.
417 : */
418 3608 : snapshot->subxcnt = 0;
419 3608 : snapshot->subxip = NULL;
420 :
421 3608 : snapshot->suboverflowed = false;
422 3608 : snapshot->takenDuringRecovery = false;
423 3608 : snapshot->copied = false;
424 3608 : snapshot->curcid = FirstCommandId;
425 3608 : snapshot->active_count = 0;
426 3608 : snapshot->regd_count = 0;
427 3608 : snapshot->snapXactCompletionCount = 0;
428 :
429 3608 : return snapshot;
430 : }
431 :
432 : /*
433 : * Build the initial slot snapshot and convert it to a normal snapshot that
434 : * is understood by HeapTupleSatisfiesMVCC.
435 : *
436 : * The snapshot will be usable directly in current transaction or exported
437 : * for loading in different transaction.
438 : */
439 : Snapshot
440 396 : SnapBuildInitialSnapshot(SnapBuild *builder)
441 : {
442 : Snapshot snap;
443 : TransactionId xid;
444 : TransactionId safeXid;
445 : TransactionId *newxip;
446 396 : int newxcnt = 0;
447 :
448 : Assert(XactIsoLevel == XACT_REPEATABLE_READ);
449 : Assert(builder->building_full_snapshot);
450 :
451 : /* don't allow older snapshots */
452 396 : InvalidateCatalogSnapshot(); /* about to overwrite MyProc->xmin */
453 396 : if (HaveRegisteredOrActiveSnapshot())
454 0 : elog(ERROR, "cannot build an initial slot snapshot when snapshots exist");
455 : Assert(!HistoricSnapshotActive());
456 :
457 396 : if (builder->state != SNAPBUILD_CONSISTENT)
458 0 : elog(ERROR, "cannot build an initial slot snapshot before reaching a consistent state");
459 :
460 396 : if (!builder->committed.includes_all_transactions)
461 0 : elog(ERROR, "cannot build an initial slot snapshot, not all transactions are monitored anymore");
462 :
463 : /* so we don't overwrite the existing value */
464 396 : if (TransactionIdIsValid(MyProc->xmin))
465 0 : elog(ERROR, "cannot build an initial slot snapshot when MyProc->xmin already is valid");
466 :
467 396 : snap = SnapBuildBuildSnapshot(builder);
468 :
469 : /*
470 : * We know that snap->xmin is alive, enforced by the logical xmin
471 : * mechanism. Due to that we can do this without locks, we're only
472 : * changing our own value.
473 : *
474 : * Building an initial snapshot is expensive and an unenforced xmin
475 : * horizon would have bad consequences, therefore always double-check that
476 : * the horizon is enforced.
477 : */
478 396 : LWLockAcquire(ProcArrayLock, LW_SHARED);
479 396 : safeXid = GetOldestSafeDecodingTransactionId(false);
480 396 : LWLockRelease(ProcArrayLock);
481 :
482 396 : if (TransactionIdFollows(safeXid, snap->xmin))
483 0 : elog(ERROR, "cannot build an initial slot snapshot as oldest safe xid %u follows snapshot's xmin %u",
484 : safeXid, snap->xmin);
485 :
486 396 : MyProc->xmin = snap->xmin;
487 :
488 : /* allocate in transaction context */
489 396 : newxip = palloc_array(TransactionId, GetMaxSnapshotXidCount());
490 :
491 : /*
492 : * snapbuild.c builds transactions in an "inverted" manner, which means it
493 : * stores committed transactions in ->xip, not ones in progress. Build a
494 : * classical snapshot by marking all non-committed transactions as
495 : * in-progress. This can be expensive.
496 : */
497 396 : for (xid = snap->xmin; NormalTransactionIdPrecedes(xid, snap->xmax);)
498 : {
499 : void *test;
500 :
501 : /*
502 : * Check whether transaction committed using the decoding snapshot
503 : * meaning of ->xip.
504 : */
505 0 : test = bsearch(&xid, snap->xip, snap->xcnt,
506 : sizeof(TransactionId), xidComparator);
507 :
508 0 : if (test == NULL)
509 : {
510 0 : if (newxcnt >= GetMaxSnapshotXidCount())
511 0 : ereport(ERROR,
512 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
513 : errmsg("initial slot snapshot too large")));
514 :
515 0 : newxip[newxcnt++] = xid;
516 : }
517 :
518 0 : TransactionIdAdvance(xid);
519 : }
520 :
521 : /* adjust remaining snapshot fields as needed */
522 396 : snap->snapshot_type = SNAPSHOT_MVCC;
523 396 : snap->xcnt = newxcnt;
524 396 : snap->xip = newxip;
525 :
526 396 : return snap;
527 : }
528 :
529 : /*
530 : * Export a snapshot so it can be set in another session with SET TRANSACTION
531 : * SNAPSHOT.
532 : *
533 : * For that we need to start a transaction in the current backend as the
534 : * importing side checks whether the source transaction is still open to make
535 : * sure the xmin horizon hasn't advanced since then.
536 : */
537 : const char *
538 2 : SnapBuildExportSnapshot(SnapBuild *builder)
539 : {
540 : Snapshot snap;
541 : char *snapname;
542 :
543 2 : if (IsTransactionOrTransactionBlock())
544 0 : elog(ERROR, "cannot export a snapshot from within a transaction");
545 :
546 2 : if (SavedResourceOwnerDuringExport)
547 0 : elog(ERROR, "can only export one snapshot at a time");
548 :
549 2 : SavedResourceOwnerDuringExport = CurrentResourceOwner;
550 2 : ExportInProgress = true;
551 :
552 2 : StartTransactionCommand();
553 :
554 : /* There doesn't seem to a nice API to set these */
555 2 : XactIsoLevel = XACT_REPEATABLE_READ;
556 2 : XactReadOnly = true;
557 :
558 2 : snap = SnapBuildInitialSnapshot(builder);
559 :
560 : /*
561 : * now that we've built a plain snapshot, make it active and use the
562 : * normal mechanisms for exporting it
563 : */
564 2 : snapname = ExportSnapshot(snap);
565 :
566 2 : ereport(LOG,
567 : (errmsg_plural("exported logical decoding snapshot: \"%s\" with %u transaction ID",
568 : "exported logical decoding snapshot: \"%s\" with %u transaction IDs",
569 : snap->xcnt,
570 : snapname, snap->xcnt)));
571 2 : return snapname;
572 : }
573 :
574 : /*
575 : * Ensure there is a snapshot and if not build one for current transaction.
576 : */
577 : Snapshot
578 16 : SnapBuildGetOrBuildSnapshot(SnapBuild *builder)
579 : {
580 : Assert(builder->state == SNAPBUILD_CONSISTENT);
581 :
582 : /* only build a new snapshot if we don't have a prebuilt one */
583 16 : if (builder->snapshot == NULL)
584 : {
585 2 : builder->snapshot = SnapBuildBuildSnapshot(builder);
586 : /* increase refcount for the snapshot builder */
587 2 : SnapBuildSnapIncRefcount(builder->snapshot);
588 : }
589 :
590 16 : return builder->snapshot;
591 : }
592 :
593 : /*
594 : * Reset a previously SnapBuildExportSnapshot()'ed snapshot if there is
595 : * any. Aborts the previously started transaction and resets the resource
596 : * owner back to its original value.
597 : */
598 : void
599 10670 : SnapBuildClearExportedSnapshot(void)
600 : {
601 : ResourceOwner tmpResOwner;
602 :
603 : /* nothing exported, that is the usual case */
604 10670 : if (!ExportInProgress)
605 10668 : return;
606 :
607 2 : if (!IsTransactionState())
608 0 : elog(ERROR, "clearing exported snapshot in wrong transaction state");
609 :
610 : /*
611 : * AbortCurrentTransaction() takes care of resetting the snapshot state,
612 : * so remember SavedResourceOwnerDuringExport.
613 : */
614 2 : tmpResOwner = SavedResourceOwnerDuringExport;
615 :
616 : /* make sure nothing could have ever happened */
617 2 : AbortCurrentTransaction();
618 :
619 2 : CurrentResourceOwner = tmpResOwner;
620 : }
621 :
622 : /*
623 : * Clear snapshot export state during transaction abort.
624 : */
625 : void
626 51586 : SnapBuildResetExportedSnapshotState(void)
627 : {
628 51586 : SavedResourceOwnerDuringExport = NULL;
629 51586 : ExportInProgress = false;
630 51586 : }
631 :
632 : /*
633 : * Handle the effects of a single heap change, appropriate to the current state
634 : * of the snapshot builder and returns whether changes made at (xid, lsn) can
635 : * be decoded.
636 : */
637 : bool
638 3113954 : SnapBuildProcessChange(SnapBuild *builder, TransactionId xid, XLogRecPtr lsn)
639 : {
640 : /*
641 : * We can't handle data in transactions if we haven't built a snapshot
642 : * yet, so don't store them.
643 : */
644 3113954 : if (builder->state < SNAPBUILD_FULL_SNAPSHOT)
645 0 : return false;
646 :
647 : /*
648 : * No point in keeping track of changes in transactions that we don't have
649 : * enough information about to decode. This means that they started before
650 : * we got into the SNAPBUILD_FULL_SNAPSHOT state.
651 : */
652 3113960 : if (builder->state < SNAPBUILD_CONSISTENT &&
653 6 : TransactionIdPrecedes(xid, builder->next_phase_at))
654 0 : return false;
655 :
656 : /*
657 : * If the reorderbuffer doesn't yet have a snapshot, add one now, it will
658 : * be needed to decode the change we're currently processing.
659 : */
660 3113954 : if (!ReorderBufferXidHasBaseSnapshot(builder->reorder, xid))
661 : {
662 : /* only build a new snapshot if we don't have a prebuilt one */
663 6484 : if (builder->snapshot == NULL)
664 : {
665 800 : builder->snapshot = SnapBuildBuildSnapshot(builder);
666 : /* increase refcount for the snapshot builder */
667 800 : SnapBuildSnapIncRefcount(builder->snapshot);
668 : }
669 :
670 : /*
671 : * Increase refcount for the transaction we're handing the snapshot
672 : * out to.
673 : */
674 6484 : SnapBuildSnapIncRefcount(builder->snapshot);
675 6484 : ReorderBufferSetBaseSnapshot(builder->reorder, xid, lsn,
676 : builder->snapshot);
677 : }
678 :
679 3113954 : return true;
680 : }
681 :
682 : /*
683 : * Do CommandId/combo CID handling after reading an xl_heap_new_cid record.
684 : * This implies that a transaction has done some form of write to system
685 : * catalogs.
686 : */
687 : void
688 48756 : SnapBuildProcessNewCid(SnapBuild *builder, TransactionId xid,
689 : XLogRecPtr lsn, xl_heap_new_cid *xlrec)
690 : {
691 : CommandId cid;
692 :
693 : /*
694 : * we only log new_cid's if a catalog tuple was modified, so mark the
695 : * transaction as containing catalog modifications
696 : */
697 48756 : ReorderBufferXidSetCatalogChanges(builder->reorder, xid, lsn);
698 :
699 48756 : ReorderBufferAddNewTupleCids(builder->reorder, xlrec->top_xid, lsn,
700 : xlrec->target_locator, xlrec->target_tid,
701 : xlrec->cmin, xlrec->cmax,
702 : xlrec->combocid);
703 :
704 : /* figure out new command id */
705 48756 : if (xlrec->cmin != InvalidCommandId &&
706 40996 : xlrec->cmax != InvalidCommandId)
707 6428 : cid = Max(xlrec->cmin, xlrec->cmax);
708 42328 : else if (xlrec->cmax != InvalidCommandId)
709 7760 : cid = xlrec->cmax;
710 34568 : else if (xlrec->cmin != InvalidCommandId)
711 34568 : cid = xlrec->cmin;
712 : else
713 : {
714 0 : cid = InvalidCommandId; /* silence compiler */
715 0 : elog(ERROR, "xl_heap_new_cid record without a valid CommandId");
716 : }
717 :
718 48756 : ReorderBufferAddNewCommandId(builder->reorder, xid, lsn, cid + 1);
719 48756 : }
720 :
721 : /*
722 : * Add a new Snapshot and invalidation messages to all transactions we're
723 : * decoding that currently are in-progress so they can see new catalog contents
724 : * made by the transaction that just committed. This is necessary because those
725 : * in-progress transactions will use the new catalog's contents from here on
726 : * (at the very least everything they do needs to be compatible with newer
727 : * catalog contents).
728 : */
729 : static void
730 2394 : SnapBuildDistributeSnapshotAndInval(SnapBuild *builder, XLogRecPtr lsn, TransactionId xid)
731 : {
732 : dlist_iter txn_i;
733 : ReorderBufferTXN *txn;
734 :
735 : /*
736 : * Iterate through all toplevel transactions. This can include
737 : * subtransactions which we just don't yet know to be that, but that's
738 : * fine, they will just get an unnecessary snapshot and invalidations
739 : * queued.
740 : */
741 4860 : dlist_foreach(txn_i, &builder->reorder->toplevel_by_lsn)
742 : {
743 2466 : txn = dlist_container(ReorderBufferTXN, node, txn_i.cur);
744 :
745 : Assert(TransactionIdIsValid(txn->xid));
746 :
747 : /*
748 : * If we don't have a base snapshot yet, there are no changes in this
749 : * transaction which in turn implies we don't yet need a snapshot at
750 : * all. We'll add a snapshot when the first change gets queued.
751 : *
752 : * Similarly, we don't need to add invalidations to a transaction
753 : * whose base snapshot is not yet set. Once a base snapshot is built,
754 : * it will include the xids of committed transactions that have
755 : * modified the catalog, thus reflecting the new catalog contents. The
756 : * existing catalog cache will have already been invalidated after
757 : * processing the invalidations in the transaction that modified
758 : * catalogs, ensuring that a fresh cache is constructed during
759 : * decoding.
760 : *
761 : * NB: This works correctly even for subtransactions because
762 : * ReorderBufferAssignChild() takes care to transfer the base snapshot
763 : * to the top-level transaction, and while iterating the changequeue
764 : * we'll get the change from the subtxn.
765 : */
766 2466 : if (!ReorderBufferXidHasBaseSnapshot(builder->reorder, txn->xid))
767 4 : continue;
768 :
769 : /*
770 : * We don't need to add snapshot or invalidations to prepared
771 : * transactions as they should not see the new catalog contents.
772 : */
773 2462 : if (rbtxn_is_prepared(txn))
774 56 : continue;
775 :
776 2406 : elog(DEBUG2, "adding a new snapshot and invalidations to %u at %X/%08X",
777 : txn->xid, LSN_FORMAT_ARGS(lsn));
778 :
779 : /*
780 : * increase the snapshot's refcount for the transaction we are handing
781 : * it out to
782 : */
783 2406 : SnapBuildSnapIncRefcount(builder->snapshot);
784 2406 : ReorderBufferAddSnapshot(builder->reorder, txn->xid, lsn,
785 : builder->snapshot);
786 :
787 : /*
788 : * Add invalidation messages to the reorder buffer of in-progress
789 : * transactions except the current committed transaction, for which we
790 : * will execute invalidations at the end.
791 : *
792 : * It is required, otherwise, we will end up using the stale catcache
793 : * contents built by the current transaction even after its decoding,
794 : * which should have been invalidated due to concurrent catalog
795 : * changing transaction.
796 : *
797 : * Distribute only the invalidation messages generated by the current
798 : * committed transaction. Invalidation messages received from other
799 : * transactions would have already been propagated to the relevant
800 : * in-progress transactions. This transaction would have processed
801 : * those invalidations, ensuring that subsequent transactions observe
802 : * a consistent cache state.
803 : */
804 2406 : if (txn->xid != xid)
805 : {
806 : uint32 ninvalidations;
807 68 : SharedInvalidationMessage *msgs = NULL;
808 :
809 68 : ninvalidations = ReorderBufferGetInvalidations(builder->reorder,
810 : xid, &msgs);
811 :
812 68 : if (ninvalidations > 0)
813 : {
814 : Assert(msgs != NULL);
815 :
816 60 : ReorderBufferAddDistributedInvalidations(builder->reorder,
817 : txn->xid, lsn,
818 : ninvalidations, msgs);
819 : }
820 : }
821 : }
822 2394 : }
823 :
824 : /*
825 : * Keep track of a new catalog changing transaction that has committed.
826 : */
827 : static void
828 2412 : SnapBuildAddCommittedTxn(SnapBuild *builder, TransactionId xid)
829 : {
830 : Assert(TransactionIdIsValid(xid));
831 :
832 2412 : if (builder->committed.xcnt == builder->committed.xcnt_space)
833 : {
834 0 : builder->committed.xcnt_space = builder->committed.xcnt_space * 2 + 1;
835 :
836 0 : elog(DEBUG1, "increasing space for committed transactions to %u",
837 : (uint32) builder->committed.xcnt_space);
838 :
839 0 : builder->committed.xip = repalloc(builder->committed.xip,
840 0 : builder->committed.xcnt_space * sizeof(TransactionId));
841 : }
842 :
843 : /*
844 : * TODO: It might make sense to keep the array sorted here instead of
845 : * doing it every time we build a new snapshot. On the other hand this
846 : * gets called repeatedly when a transaction with subtransactions commits.
847 : */
848 2412 : builder->committed.xip[builder->committed.xcnt++] = xid;
849 2412 : }
850 :
851 : /*
852 : * Remove knowledge about transactions we treat as committed or containing catalog
853 : * changes that are smaller than ->xmin. Those won't ever get checked via
854 : * the ->committed or ->catchange array, respectively. The committed xids will
855 : * get checked via the clog machinery.
856 : *
857 : * We can ideally remove the transaction from catchange array once it is
858 : * finished (committed/aborted) but that could be costly as we need to maintain
859 : * the xids order in the array.
860 : */
861 : static void
862 828 : SnapBuildPurgeOlderTxn(SnapBuild *builder)
863 : {
864 : int off;
865 : TransactionId *workspace;
866 828 : int surviving_xids = 0;
867 :
868 : /* not ready yet */
869 828 : if (!TransactionIdIsNormal(builder->xmin))
870 0 : return;
871 :
872 : /* TODO: Neater algorithm than just copying and iterating? */
873 : workspace =
874 828 : MemoryContextAlloc(builder->context,
875 828 : builder->committed.xcnt * sizeof(TransactionId));
876 :
877 : /* copy xids that still are interesting to workspace */
878 1272 : for (off = 0; off < builder->committed.xcnt; off++)
879 : {
880 444 : if (NormalTransactionIdPrecedes(builder->committed.xip[off],
881 : builder->xmin))
882 : ; /* remove */
883 : else
884 2 : workspace[surviving_xids++] = builder->committed.xip[off];
885 : }
886 :
887 : /* copy workspace back to persistent state */
888 828 : memcpy(builder->committed.xip, workspace,
889 : surviving_xids * sizeof(TransactionId));
890 :
891 828 : elog(DEBUG3, "purged committed transactions from %u to %u, xmin: %u, xmax: %u",
892 : (uint32) builder->committed.xcnt, (uint32) surviving_xids,
893 : builder->xmin, builder->xmax);
894 828 : builder->committed.xcnt = surviving_xids;
895 :
896 828 : pfree(workspace);
897 :
898 : /*
899 : * Purge xids in ->catchange as well. The purged array must also be sorted
900 : * in xidComparator order.
901 : */
902 828 : if (builder->catchange.xcnt > 0)
903 : {
904 : /*
905 : * Since catchange.xip is sorted, we find the lower bound of xids that
906 : * are still interesting.
907 : */
908 18 : for (off = 0; off < builder->catchange.xcnt; off++)
909 : {
910 12 : if (TransactionIdFollowsOrEquals(builder->catchange.xip[off],
911 : builder->xmin))
912 2 : break;
913 : }
914 :
915 8 : surviving_xids = builder->catchange.xcnt - off;
916 :
917 8 : if (surviving_xids > 0)
918 : {
919 2 : memmove(builder->catchange.xip, &(builder->catchange.xip[off]),
920 : surviving_xids * sizeof(TransactionId));
921 : }
922 : else
923 : {
924 6 : pfree(builder->catchange.xip);
925 6 : builder->catchange.xip = NULL;
926 : }
927 :
928 8 : elog(DEBUG3, "purged catalog modifying transactions from %u to %u, xmin: %u, xmax: %u",
929 : (uint32) builder->catchange.xcnt, (uint32) surviving_xids,
930 : builder->xmin, builder->xmax);
931 8 : builder->catchange.xcnt = surviving_xids;
932 : }
933 : }
934 :
935 : /*
936 : * Handle everything that needs to be done when a transaction commits
937 : */
938 : void
939 6216 : SnapBuildCommitTxn(SnapBuild *builder, XLogRecPtr lsn, TransactionId xid,
940 : int nsubxacts, TransactionId *subxacts, uint32 xinfo)
941 : {
942 : int nxact;
943 :
944 6216 : bool needs_snapshot = false;
945 6216 : bool needs_timetravel = false;
946 6216 : bool sub_needs_timetravel = false;
947 :
948 6216 : TransactionId xmax = xid;
949 :
950 : /*
951 : * Transactions preceding BUILDING_SNAPSHOT will neither be decoded, nor
952 : * will they be part of a snapshot. So we don't need to record anything.
953 : */
954 6216 : if (builder->state == SNAPBUILD_START ||
955 6216 : (builder->state == SNAPBUILD_BUILDING_SNAPSHOT &&
956 0 : TransactionIdPrecedes(xid, builder->next_phase_at)))
957 : {
958 : /* ensure that only commits after this are getting replayed */
959 0 : if (builder->start_decoding_at <= lsn)
960 0 : builder->start_decoding_at = lsn + 1;
961 0 : return;
962 : }
963 :
964 6216 : if (builder->state < SNAPBUILD_CONSISTENT)
965 : {
966 : /* ensure that only commits after this are getting replayed */
967 10 : if (builder->start_decoding_at <= lsn)
968 4 : builder->start_decoding_at = lsn + 1;
969 :
970 : /*
971 : * If building an exportable snapshot, force xid to be tracked, even
972 : * if the transaction didn't modify the catalog.
973 : */
974 10 : if (builder->building_full_snapshot)
975 : {
976 0 : needs_timetravel = true;
977 : }
978 : }
979 :
980 8714 : for (nxact = 0; nxact < nsubxacts; nxact++)
981 : {
982 2498 : TransactionId subxid = subxacts[nxact];
983 :
984 : /*
985 : * Add subtransaction to base snapshot if catalog modifying, we don't
986 : * distinguish to toplevel transactions there.
987 : */
988 2498 : if (SnapBuildXidHasCatalogChanges(builder, subxid, xinfo))
989 : {
990 18 : sub_needs_timetravel = true;
991 18 : needs_snapshot = true;
992 :
993 18 : elog(DEBUG1, "found subtransaction %u:%u with catalog changes",
994 : xid, subxid);
995 :
996 18 : SnapBuildAddCommittedTxn(builder, subxid);
997 :
998 18 : if (NormalTransactionIdFollows(subxid, xmax))
999 18 : xmax = subxid;
1000 : }
1001 :
1002 : /*
1003 : * If we're forcing timetravel we also need visibility information
1004 : * about subtransaction, so keep track of subtransaction's state, even
1005 : * if not catalog modifying. Don't need to distribute a snapshot in
1006 : * that case.
1007 : */
1008 2480 : else if (needs_timetravel)
1009 : {
1010 0 : SnapBuildAddCommittedTxn(builder, subxid);
1011 0 : if (NormalTransactionIdFollows(subxid, xmax))
1012 0 : xmax = subxid;
1013 : }
1014 : }
1015 :
1016 : /* if top-level modified catalog, it'll need a snapshot */
1017 6216 : if (SnapBuildXidHasCatalogChanges(builder, xid, xinfo))
1018 : {
1019 2392 : elog(DEBUG2, "found top level transaction %u, with catalog changes",
1020 : xid);
1021 2392 : needs_snapshot = true;
1022 2392 : needs_timetravel = true;
1023 2392 : SnapBuildAddCommittedTxn(builder, xid);
1024 : }
1025 3824 : else if (sub_needs_timetravel)
1026 : {
1027 : /* track toplevel txn as well, subxact alone isn't meaningful */
1028 2 : elog(DEBUG2, "forced transaction %u to do timetravel due to one of its subtransactions",
1029 : xid);
1030 2 : needs_timetravel = true;
1031 2 : SnapBuildAddCommittedTxn(builder, xid);
1032 : }
1033 3822 : else if (needs_timetravel)
1034 : {
1035 0 : elog(DEBUG2, "forced transaction %u to do timetravel", xid);
1036 :
1037 0 : SnapBuildAddCommittedTxn(builder, xid);
1038 : }
1039 :
1040 6216 : if (!needs_timetravel)
1041 : {
1042 : /* record that we cannot export a general snapshot anymore */
1043 3822 : builder->committed.includes_all_transactions = false;
1044 : }
1045 :
1046 : Assert(!needs_snapshot || needs_timetravel);
1047 :
1048 : /*
1049 : * Adjust xmax of the snapshot builder, we only do that for committed,
1050 : * catalog modifying, transactions, everything else isn't interesting for
1051 : * us since we'll never look at the respective rows.
1052 : */
1053 6216 : if (needs_timetravel &&
1054 4788 : (!TransactionIdIsValid(builder->xmax) ||
1055 2394 : TransactionIdFollowsOrEquals(xmax, builder->xmax)))
1056 : {
1057 2392 : builder->xmax = xmax;
1058 2392 : TransactionIdAdvance(builder->xmax);
1059 : }
1060 :
1061 : /* if there's any reason to build a historic snapshot, do so now */
1062 6216 : if (needs_snapshot)
1063 : {
1064 : /*
1065 : * If we haven't built a complete snapshot yet there's no need to hand
1066 : * it out, it wouldn't (and couldn't) be used anyway.
1067 : */
1068 2394 : if (builder->state < SNAPBUILD_FULL_SNAPSHOT)
1069 0 : return;
1070 :
1071 : /*
1072 : * Decrease the snapshot builder's refcount of the old snapshot, note
1073 : * that it still will be used if it has been handed out to the
1074 : * reorderbuffer earlier.
1075 : */
1076 2394 : if (builder->snapshot)
1077 2394 : SnapBuildSnapDecRefcount(builder->snapshot);
1078 :
1079 2394 : builder->snapshot = SnapBuildBuildSnapshot(builder);
1080 :
1081 : /* we might need to execute invalidations, add snapshot */
1082 2394 : if (!ReorderBufferXidHasBaseSnapshot(builder->reorder, xid))
1083 : {
1084 16 : SnapBuildSnapIncRefcount(builder->snapshot);
1085 16 : ReorderBufferSetBaseSnapshot(builder->reorder, xid, lsn,
1086 : builder->snapshot);
1087 : }
1088 :
1089 : /* refcount of the snapshot builder for the new snapshot */
1090 2394 : SnapBuildSnapIncRefcount(builder->snapshot);
1091 :
1092 : /*
1093 : * Add a new catalog snapshot and invalidations messages to all
1094 : * currently running transactions.
1095 : */
1096 2394 : SnapBuildDistributeSnapshotAndInval(builder, lsn, xid);
1097 : }
1098 : }
1099 :
1100 : /*
1101 : * Check the reorder buffer and the snapshot to see if the given transaction has
1102 : * modified catalogs.
1103 : */
1104 : static inline bool
1105 8714 : SnapBuildXidHasCatalogChanges(SnapBuild *builder, TransactionId xid,
1106 : uint32 xinfo)
1107 : {
1108 8714 : if (ReorderBufferXidHasCatalogChanges(builder->reorder, xid))
1109 2402 : return true;
1110 :
1111 : /*
1112 : * The transactions that have changed catalogs must have invalidation
1113 : * info.
1114 : */
1115 6312 : if (!(xinfo & XACT_XINFO_HAS_INVALS))
1116 6296 : return false;
1117 :
1118 : /* Check the catchange XID array */
1119 24 : return ((builder->catchange.xcnt > 0) &&
1120 8 : (bsearch(&xid, builder->catchange.xip, builder->catchange.xcnt,
1121 : sizeof(TransactionId), xidComparator) != NULL));
1122 : }
1123 :
1124 : /* -----------------------------------
1125 : * Snapshot building functions dealing with xlog records
1126 : * -----------------------------------
1127 : */
1128 :
1129 : /*
1130 : * Process a running xacts record, and use its information to first build a
1131 : * historic snapshot and later to release resources that aren't needed
1132 : * anymore.
1133 : */
1134 : void
1135 2764 : SnapBuildProcessRunningXacts(SnapBuild *builder, XLogRecPtr lsn, xl_running_xacts *running)
1136 : {
1137 : ReorderBufferTXN *txn;
1138 : TransactionId xmin;
1139 :
1140 : /*
1141 : * If we're not consistent yet, inspect the record to see whether it
1142 : * allows to get closer to being consistent. If we are consistent, dump
1143 : * our snapshot so others or we, after a restart, can use it.
1144 : */
1145 2764 : if (builder->state < SNAPBUILD_CONSISTENT)
1146 : {
1147 : /* returns false if there's no point in performing cleanup just yet */
1148 1982 : if (!SnapBuildFindSnapshot(builder, lsn, running))
1149 1932 : return;
1150 : }
1151 : else
1152 782 : SnapBuildSerialize(builder, lsn);
1153 :
1154 : /*
1155 : * Update range of interesting xids based on the running xacts
1156 : * information. We don't increase ->xmax using it, because once we are in
1157 : * a consistent state we can do that ourselves and much more efficiently
1158 : * so, because we only need to do it for catalog transactions since we
1159 : * only ever look at those.
1160 : *
1161 : * NB: We only increase xmax when a catalog modifying transaction commits
1162 : * (see SnapBuildCommitTxn). Because of this, xmax can be lower than
1163 : * xmin, which looks odd but is correct and actually more efficient, since
1164 : * we hit fast paths in heapam_visibility.c.
1165 : */
1166 828 : builder->xmin = running->oldestRunningXid;
1167 :
1168 : /* Remove transactions we don't need to keep track off anymore */
1169 828 : SnapBuildPurgeOlderTxn(builder);
1170 :
1171 : /*
1172 : * Advance the xmin limit for the current replication slot, to allow
1173 : * vacuum to clean up the tuples this slot has been protecting.
1174 : *
1175 : * The reorderbuffer might have an xmin among the currently running
1176 : * snapshots; use it if so. If not, we need only consider the snapshots
1177 : * we'll produce later, which can't be less than the oldest running xid in
1178 : * the record we're reading now.
1179 : */
1180 828 : xmin = ReorderBufferGetOldestXmin(builder->reorder);
1181 828 : if (xmin == InvalidTransactionId)
1182 734 : xmin = running->oldestRunningXid;
1183 828 : elog(DEBUG3, "xmin: %u, xmax: %u, oldest running: %u, oldest xmin: %u",
1184 : builder->xmin, builder->xmax, running->oldestRunningXid, xmin);
1185 828 : LogicalIncreaseXminForSlot(lsn, xmin);
1186 :
1187 : /*
1188 : * Also tell the slot where we can restart decoding from. We don't want to
1189 : * do that after every commit because changing that implies an fsync of
1190 : * the logical slot's state file, so we only do it every time we see a
1191 : * running xacts record.
1192 : *
1193 : * Do so by looking for the oldest in progress transaction (determined by
1194 : * the first LSN of any of its relevant records). Every transaction
1195 : * remembers the last location we stored the snapshot to disk before its
1196 : * beginning. That point is where we can restart from.
1197 : */
1198 :
1199 : /*
1200 : * Can't know about a serialized snapshot's location if we're not
1201 : * consistent.
1202 : */
1203 828 : if (builder->state < SNAPBUILD_CONSISTENT)
1204 36 : return;
1205 :
1206 792 : txn = ReorderBufferGetOldestTXN(builder->reorder);
1207 :
1208 : /*
1209 : * oldest ongoing txn might have started when we didn't yet serialize
1210 : * anything because we hadn't reached a consistent state yet.
1211 : */
1212 792 : if (txn != NULL && XLogRecPtrIsValid(txn->restart_decoding_lsn))
1213 46 : LogicalIncreaseRestartDecodingForSlot(lsn, txn->restart_decoding_lsn);
1214 :
1215 : /*
1216 : * No in-progress transaction, can reuse the last serialized snapshot if
1217 : * we have one.
1218 : */
1219 746 : else if (txn == NULL &&
1220 680 : XLogRecPtrIsValid(builder->reorder->current_restart_decoding_lsn) &&
1221 676 : XLogRecPtrIsValid(builder->last_serialized_snapshot))
1222 676 : LogicalIncreaseRestartDecodingForSlot(lsn,
1223 : builder->last_serialized_snapshot);
1224 : }
1225 :
1226 :
1227 : /*
1228 : * Build the start of a snapshot that's capable of decoding the catalog.
1229 : *
1230 : * Helper function for SnapBuildProcessRunningXacts() while we're not yet
1231 : * consistent.
1232 : *
1233 : * Returns true if there is a point in performing internal maintenance/cleanup
1234 : * using the xl_running_xacts record.
1235 : */
1236 : static bool
1237 1982 : SnapBuildFindSnapshot(SnapBuild *builder, XLogRecPtr lsn, xl_running_xacts *running)
1238 : {
1239 : /* ---
1240 : * Build catalog decoding snapshot incrementally using information about
1241 : * the currently running transactions. There are several ways to do that:
1242 : *
1243 : * a) There were no running transactions when the xl_running_xacts record
1244 : * was inserted, jump to CONSISTENT immediately. We might find such a
1245 : * state while waiting on c)'s sub-states.
1246 : *
1247 : * b) This (in a previous run) or another decoding slot serialized a
1248 : * snapshot to disk that we can use. Can't use this method while finding
1249 : * the start point for decoding changes as the restart LSN would be an
1250 : * arbitrary LSN but we need to find the start point to extract changes
1251 : * where we won't see the data for partial transactions. Also, we cannot
1252 : * use this method when a slot needs a full snapshot for export or direct
1253 : * use, as that snapshot will only contain catalog modifying transactions.
1254 : *
1255 : * c) First incrementally build a snapshot for catalog tuples
1256 : * (BUILDING_SNAPSHOT), that requires all, already in-progress,
1257 : * transactions to finish. Every transaction starting after that
1258 : * (FULL_SNAPSHOT state), has enough information to be decoded. But
1259 : * for older running transactions no viable snapshot exists yet, so
1260 : * CONSISTENT will only be reached once all of those have finished.
1261 : * ---
1262 : */
1263 :
1264 : /*
1265 : * xl_running_xacts record is older than what we can use, we might not
1266 : * have all necessary catalog rows anymore.
1267 : */
1268 1982 : if (TransactionIdIsNormal(builder->initial_xmin_horizon) &&
1269 926 : NormalTransactionIdPrecedes(running->oldestRunningXid,
1270 : builder->initial_xmin_horizon))
1271 : {
1272 0 : ereport(DEBUG1,
1273 : errmsg_internal("skipping snapshot at %X/%08X while building logical decoding snapshot, xmin horizon too low",
1274 : LSN_FORMAT_ARGS(lsn)),
1275 : errdetail_internal("initial xmin horizon of %u vs the snapshot's %u",
1276 : builder->initial_xmin_horizon, running->oldestRunningXid));
1277 :
1278 :
1279 0 : SnapBuildWaitSnapshot(running, builder->initial_xmin_horizon);
1280 :
1281 0 : return true;
1282 : }
1283 :
1284 : /*
1285 : * a) No transaction were running, we can jump to consistent.
1286 : *
1287 : * This is not affected by races around xl_running_xacts, because we can
1288 : * miss transaction commits, but currently not transactions starting.
1289 : *
1290 : * NB: We might have already started to incrementally assemble a snapshot,
1291 : * so we need to be careful to deal with that.
1292 : */
1293 1982 : if (running->oldestRunningXid == running->nextXid)
1294 : {
1295 1916 : if (!XLogRecPtrIsValid(builder->start_decoding_at) ||
1296 1008 : builder->start_decoding_at <= lsn)
1297 : /* can decode everything after this */
1298 910 : builder->start_decoding_at = lsn + 1;
1299 :
1300 : /* As no transactions were running xmin/xmax can be trivially set. */
1301 1916 : builder->xmin = running->nextXid; /* < are finished */
1302 1916 : builder->xmax = running->nextXid; /* >= are running */
1303 :
1304 : /* so we can safely use the faster comparisons */
1305 : Assert(TransactionIdIsNormal(builder->xmin));
1306 : Assert(TransactionIdIsNormal(builder->xmax));
1307 :
1308 1916 : builder->state = SNAPBUILD_CONSISTENT;
1309 1916 : builder->next_phase_at = InvalidTransactionId;
1310 :
1311 1916 : ereport(LOG,
1312 : errmsg("logical decoding found consistent point at %X/%08X",
1313 : LSN_FORMAT_ARGS(lsn)),
1314 : errdetail("There are no running transactions."));
1315 :
1316 1916 : return false;
1317 : }
1318 :
1319 : /*
1320 : * b) valid on disk state and while neither building full snapshot nor
1321 : * creating a slot.
1322 : */
1323 66 : else if (!builder->building_full_snapshot &&
1324 100 : !builder->in_slot_creation &&
1325 38 : SnapBuildRestore(builder, lsn))
1326 : {
1327 : /* there won't be any state to cleanup */
1328 16 : return false;
1329 : }
1330 :
1331 : /*
1332 : * c) transition from START to BUILDING_SNAPSHOT.
1333 : *
1334 : * In START state, and a xl_running_xacts record with running xacts is
1335 : * encountered. In that case, switch to BUILDING_SNAPSHOT state, and
1336 : * record xl_running_xacts->nextXid. Once all running xacts have finished
1337 : * (i.e. they're all >= nextXid), we have a complete catalog snapshot. It
1338 : * might look that we could use xl_running_xacts's ->xids information to
1339 : * get there quicker, but that is problematic because transactions marked
1340 : * as running, might already have inserted their commit record - it's
1341 : * infeasible to change that with locking.
1342 : */
1343 50 : else if (builder->state == SNAPBUILD_START)
1344 : {
1345 28 : builder->state = SNAPBUILD_BUILDING_SNAPSHOT;
1346 28 : builder->next_phase_at = running->nextXid;
1347 :
1348 : /*
1349 : * Start with an xmin/xmax that's correct for future, when all the
1350 : * currently running transactions have finished. We'll update both
1351 : * while waiting for the pending transactions to finish.
1352 : */
1353 28 : builder->xmin = running->nextXid; /* < are finished */
1354 28 : builder->xmax = running->nextXid; /* >= are running */
1355 :
1356 : /* so we can safely use the faster comparisons */
1357 : Assert(TransactionIdIsNormal(builder->xmin));
1358 : Assert(TransactionIdIsNormal(builder->xmax));
1359 :
1360 28 : ereport(LOG,
1361 : errmsg("logical decoding found initial starting point at %X/%08X",
1362 : LSN_FORMAT_ARGS(lsn)),
1363 : errdetail("Waiting for transactions (approximately %d) older than %u to end.",
1364 : running->xcnt, running->nextXid));
1365 :
1366 28 : SnapBuildWaitSnapshot(running, running->nextXid);
1367 : }
1368 :
1369 : /*
1370 : * c) transition from BUILDING_SNAPSHOT to FULL_SNAPSHOT.
1371 : *
1372 : * In BUILDING_SNAPSHOT state, and this xl_running_xacts' oldestRunningXid
1373 : * is >= than nextXid from when we switched to BUILDING_SNAPSHOT. This
1374 : * means all transactions starting afterwards have enough information to
1375 : * be decoded. Switch to FULL_SNAPSHOT.
1376 : */
1377 34 : else if (builder->state == SNAPBUILD_BUILDING_SNAPSHOT &&
1378 12 : TransactionIdPrecedesOrEquals(builder->next_phase_at,
1379 : running->oldestRunningXid))
1380 : {
1381 10 : builder->state = SNAPBUILD_FULL_SNAPSHOT;
1382 10 : builder->next_phase_at = running->nextXid;
1383 :
1384 10 : ereport(LOG,
1385 : errmsg("logical decoding found initial consistent point at %X/%08X",
1386 : LSN_FORMAT_ARGS(lsn)),
1387 : errdetail("Waiting for transactions (approximately %d) older than %u to end.",
1388 : running->xcnt, running->nextXid));
1389 :
1390 10 : SnapBuildWaitSnapshot(running, running->nextXid);
1391 : }
1392 :
1393 : /*
1394 : * c) transition from FULL_SNAPSHOT to CONSISTENT.
1395 : *
1396 : * In FULL_SNAPSHOT state, and this xl_running_xacts' oldestRunningXid is
1397 : * >= than nextXid from when we switched to FULL_SNAPSHOT. This means all
1398 : * transactions that are currently in progress have a catalog snapshot,
1399 : * and all their changes have been collected. Switch to CONSISTENT.
1400 : */
1401 22 : else if (builder->state == SNAPBUILD_FULL_SNAPSHOT &&
1402 10 : TransactionIdPrecedesOrEquals(builder->next_phase_at,
1403 : running->oldestRunningXid))
1404 : {
1405 10 : builder->state = SNAPBUILD_CONSISTENT;
1406 10 : builder->next_phase_at = InvalidTransactionId;
1407 :
1408 10 : ereport(LOG,
1409 : errmsg("logical decoding found consistent point at %X/%08X",
1410 : LSN_FORMAT_ARGS(lsn)),
1411 : errdetail("There are no old transactions anymore."));
1412 : }
1413 :
1414 : /*
1415 : * We already started to track running xacts and need to wait for all
1416 : * in-progress ones to finish. We fall through to the normal processing of
1417 : * records so incremental cleanup can be performed.
1418 : */
1419 46 : return true;
1420 : }
1421 :
1422 : /* ---
1423 : * Iterate through xids in record, wait for all older than the cutoff to
1424 : * finish. Then, if possible, log a new xl_running_xacts record.
1425 : *
1426 : * This isn't required for the correctness of decoding, but to:
1427 : * a) allow isolationtester to notice that we're currently waiting for
1428 : * something.
1429 : * b) log a new xl_running_xacts record where it'd be helpful, without having
1430 : * to wait for bgwriter or checkpointer.
1431 : * ---
1432 : */
1433 : static void
1434 38 : SnapBuildWaitSnapshot(xl_running_xacts *running, TransactionId cutoff)
1435 : {
1436 : int off;
1437 :
1438 72 : for (off = 0; off < running->xcnt; off++)
1439 : {
1440 38 : TransactionId xid = running->xids[off];
1441 :
1442 : /*
1443 : * Upper layers should prevent that we ever need to wait on ourselves.
1444 : * Check anyway, since failing to do so would either result in an
1445 : * endless wait or an Assert() failure.
1446 : */
1447 38 : if (TransactionIdIsCurrentTransactionId(xid))
1448 0 : elog(ERROR, "waiting for ourselves");
1449 :
1450 38 : if (TransactionIdFollows(xid, cutoff))
1451 0 : continue;
1452 :
1453 38 : XactLockTableWait(xid, NULL, NULL, XLTW_None);
1454 : }
1455 :
1456 : /*
1457 : * All transactions we needed to finish finished - try to ensure there is
1458 : * another xl_running_xacts record in a timely manner, without having to
1459 : * wait for bgwriter or checkpointer to log one. During recovery we can't
1460 : * enforce that, so we'll have to wait.
1461 : */
1462 34 : if (!RecoveryInProgress())
1463 : {
1464 34 : LogStandbySnapshot();
1465 : }
1466 34 : }
1467 :
1468 : #define SnapBuildOnDiskConstantSize \
1469 : offsetof(SnapBuildOnDisk, builder)
1470 : #define SnapBuildOnDiskNotChecksummedSize \
1471 : offsetof(SnapBuildOnDisk, version)
1472 :
1473 : #define SNAPBUILD_MAGIC 0x51A1E001
1474 : #define SNAPBUILD_VERSION 6
1475 :
1476 : /*
1477 : * Store/Load a snapshot from disk, depending on the snapshot builder's state.
1478 : *
1479 : * Supposed to be used by external (i.e. not snapbuild.c) code that just read
1480 : * a record that's a potential location for a serialized snapshot.
1481 : */
1482 : void
1483 104 : SnapBuildSerializationPoint(SnapBuild *builder, XLogRecPtr lsn)
1484 : {
1485 104 : if (builder->state < SNAPBUILD_CONSISTENT)
1486 0 : SnapBuildRestore(builder, lsn);
1487 : else
1488 104 : SnapBuildSerialize(builder, lsn);
1489 104 : }
1490 :
1491 : /*
1492 : * Serialize the snapshot 'builder' at the location 'lsn' if it hasn't already
1493 : * been done by another decoding process.
1494 : */
1495 : static void
1496 886 : SnapBuildSerialize(SnapBuild *builder, XLogRecPtr lsn)
1497 : {
1498 : Size needed_length;
1499 886 : SnapBuildOnDisk *ondisk = NULL;
1500 886 : TransactionId *catchange_xip = NULL;
1501 : MemoryContext old_ctx;
1502 : size_t catchange_xcnt;
1503 : char *ondisk_c;
1504 : int fd;
1505 : char tmppath[MAXPGPATH];
1506 : char path[MAXPGPATH];
1507 : int ret;
1508 : struct stat stat_buf;
1509 : Size sz;
1510 :
1511 : Assert(XLogRecPtrIsValid(lsn));
1512 : Assert(!XLogRecPtrIsValid(builder->last_serialized_snapshot) ||
1513 : builder->last_serialized_snapshot <= lsn);
1514 :
1515 : /*
1516 : * no point in serializing if we cannot continue to work immediately after
1517 : * restoring the snapshot
1518 : */
1519 886 : if (builder->state < SNAPBUILD_CONSISTENT)
1520 0 : return;
1521 :
1522 : /* consistent snapshots have no next phase */
1523 : Assert(builder->next_phase_at == InvalidTransactionId);
1524 :
1525 : /*
1526 : * We identify snapshots by the LSN they are valid for. We don't need to
1527 : * include timelines in the name as each LSN maps to exactly one timeline
1528 : * unless the user used pg_resetwal or similar. If a user did so, there's
1529 : * no hope continuing to decode anyway.
1530 : */
1531 886 : sprintf(path, "%s/%X-%X.snap",
1532 : PG_LOGICAL_SNAPSHOTS_DIR,
1533 886 : LSN_FORMAT_ARGS(lsn));
1534 :
1535 : /*
1536 : * first check whether some other backend already has written the snapshot
1537 : * for this LSN. It's perfectly fine if there's none, so we accept ENOENT
1538 : * as a valid state. Everything else is an unexpected error.
1539 : */
1540 886 : ret = stat(path, &stat_buf);
1541 :
1542 886 : if (ret != 0 && errno != ENOENT)
1543 0 : ereport(ERROR,
1544 : (errcode_for_file_access(),
1545 : errmsg("could not stat file \"%s\": %m", path)));
1546 :
1547 886 : else if (ret == 0)
1548 : {
1549 : /*
1550 : * somebody else has already serialized to this point, don't overwrite
1551 : * but remember location, so we don't need to read old data again.
1552 : *
1553 : * To be sure it has been synced to disk after the rename() from the
1554 : * tempfile filename to the real filename, we just repeat the fsync.
1555 : * That ought to be cheap because in most scenarios it should already
1556 : * be safely on disk.
1557 : */
1558 266 : fsync_fname(path, false);
1559 266 : fsync_fname(PG_LOGICAL_SNAPSHOTS_DIR, true);
1560 :
1561 266 : builder->last_serialized_snapshot = lsn;
1562 266 : goto out;
1563 : }
1564 :
1565 : /*
1566 : * there is an obvious race condition here between the time we stat(2) the
1567 : * file and us writing the file. But we rename the file into place
1568 : * atomically and all files created need to contain the same data anyway,
1569 : * so this is perfectly fine, although a bit of a resource waste. Locking
1570 : * seems like pointless complication.
1571 : */
1572 620 : elog(DEBUG1, "serializing snapshot to %s", path);
1573 :
1574 : /* to make sure only we will write to this tempfile, include pid */
1575 620 : sprintf(tmppath, "%s/%X-%X.snap.%d.tmp",
1576 : PG_LOGICAL_SNAPSHOTS_DIR,
1577 620 : LSN_FORMAT_ARGS(lsn), MyProcPid);
1578 :
1579 : /*
1580 : * Unlink temporary file if it already exists, needs to have been before a
1581 : * crash/error since we won't enter this function twice from within a
1582 : * single decoding slot/backend and the temporary file contains the pid of
1583 : * the current process.
1584 : */
1585 620 : if (unlink(tmppath) != 0 && errno != ENOENT)
1586 0 : ereport(ERROR,
1587 : (errcode_for_file_access(),
1588 : errmsg("could not remove file \"%s\": %m", tmppath)));
1589 :
1590 620 : old_ctx = MemoryContextSwitchTo(builder->context);
1591 :
1592 : /* Get the catalog modifying transactions that are yet not committed */
1593 620 : catchange_xip = ReorderBufferGetCatalogChangesXacts(builder->reorder);
1594 620 : catchange_xcnt = dclist_count(&builder->reorder->catchange_txns);
1595 :
1596 620 : needed_length = sizeof(SnapBuildOnDisk) +
1597 620 : sizeof(TransactionId) * (builder->committed.xcnt + catchange_xcnt);
1598 :
1599 620 : ondisk_c = palloc0(needed_length);
1600 620 : ondisk = (SnapBuildOnDisk *) ondisk_c;
1601 620 : ondisk->magic = SNAPBUILD_MAGIC;
1602 620 : ondisk->version = SNAPBUILD_VERSION;
1603 620 : ondisk->length = needed_length;
1604 620 : INIT_CRC32C(ondisk->checksum);
1605 620 : COMP_CRC32C(ondisk->checksum,
1606 : ((char *) ondisk) + SnapBuildOnDiskNotChecksummedSize,
1607 : SnapBuildOnDiskConstantSize - SnapBuildOnDiskNotChecksummedSize);
1608 620 : ondisk_c += sizeof(SnapBuildOnDisk);
1609 :
1610 620 : memcpy(&ondisk->builder, builder, sizeof(SnapBuild));
1611 : /* NULL-ify memory-only data */
1612 620 : ondisk->builder.context = NULL;
1613 620 : ondisk->builder.snapshot = NULL;
1614 620 : ondisk->builder.reorder = NULL;
1615 620 : ondisk->builder.committed.xip = NULL;
1616 620 : ondisk->builder.catchange.xip = NULL;
1617 : /* update catchange only on disk data */
1618 620 : ondisk->builder.catchange.xcnt = catchange_xcnt;
1619 :
1620 620 : COMP_CRC32C(ondisk->checksum,
1621 : &ondisk->builder,
1622 : sizeof(SnapBuild));
1623 :
1624 : /* copy committed xacts */
1625 620 : if (builder->committed.xcnt > 0)
1626 : {
1627 112 : sz = sizeof(TransactionId) * builder->committed.xcnt;
1628 112 : memcpy(ondisk_c, builder->committed.xip, sz);
1629 112 : COMP_CRC32C(ondisk->checksum, ondisk_c, sz);
1630 112 : ondisk_c += sz;
1631 : }
1632 :
1633 : /* copy catalog modifying xacts */
1634 620 : if (catchange_xcnt > 0)
1635 : {
1636 20 : sz = sizeof(TransactionId) * catchange_xcnt;
1637 20 : memcpy(ondisk_c, catchange_xip, sz);
1638 20 : COMP_CRC32C(ondisk->checksum, ondisk_c, sz);
1639 20 : ondisk_c += sz;
1640 : }
1641 :
1642 620 : FIN_CRC32C(ondisk->checksum);
1643 :
1644 : /* we have valid data now, open tempfile and write it there */
1645 620 : fd = OpenTransientFile(tmppath,
1646 : O_CREAT | O_EXCL | O_WRONLY | PG_BINARY);
1647 620 : if (fd < 0)
1648 0 : ereport(ERROR,
1649 : (errcode_for_file_access(),
1650 : errmsg("could not open file \"%s\": %m", tmppath)));
1651 :
1652 620 : errno = 0;
1653 620 : pgstat_report_wait_start(WAIT_EVENT_SNAPBUILD_WRITE);
1654 620 : if ((write(fd, ondisk, needed_length)) != needed_length)
1655 : {
1656 0 : int save_errno = errno;
1657 :
1658 0 : CloseTransientFile(fd);
1659 :
1660 : /* if write didn't set errno, assume problem is no disk space */
1661 0 : errno = save_errno ? save_errno : ENOSPC;
1662 0 : ereport(ERROR,
1663 : (errcode_for_file_access(),
1664 : errmsg("could not write to file \"%s\": %m", tmppath)));
1665 : }
1666 620 : pgstat_report_wait_end();
1667 :
1668 : /*
1669 : * fsync the file before renaming so that even if we crash after this we
1670 : * have either a fully valid file or nothing.
1671 : *
1672 : * It's safe to just ERROR on fsync() here because we'll retry the whole
1673 : * operation including the writes.
1674 : *
1675 : * TODO: Do the fsync() via checkpoints/restartpoints, doing it here has
1676 : * some noticeable overhead since it's performed synchronously during
1677 : * decoding?
1678 : */
1679 620 : pgstat_report_wait_start(WAIT_EVENT_SNAPBUILD_SYNC);
1680 620 : if (pg_fsync(fd) != 0)
1681 : {
1682 0 : int save_errno = errno;
1683 :
1684 0 : CloseTransientFile(fd);
1685 0 : errno = save_errno;
1686 0 : ereport(ERROR,
1687 : (errcode_for_file_access(),
1688 : errmsg("could not fsync file \"%s\": %m", tmppath)));
1689 : }
1690 620 : pgstat_report_wait_end();
1691 :
1692 620 : if (CloseTransientFile(fd) != 0)
1693 0 : ereport(ERROR,
1694 : (errcode_for_file_access(),
1695 : errmsg("could not close file \"%s\": %m", tmppath)));
1696 :
1697 620 : fsync_fname(PG_LOGICAL_SNAPSHOTS_DIR, true);
1698 :
1699 : /*
1700 : * We may overwrite the work from some other backend, but that's ok, our
1701 : * snapshot is valid as well, we'll just have done some superfluous work.
1702 : */
1703 620 : if (rename(tmppath, path) != 0)
1704 : {
1705 0 : ereport(ERROR,
1706 : (errcode_for_file_access(),
1707 : errmsg("could not rename file \"%s\" to \"%s\": %m",
1708 : tmppath, path)));
1709 : }
1710 :
1711 : /* make sure we persist */
1712 620 : fsync_fname(path, false);
1713 620 : fsync_fname(PG_LOGICAL_SNAPSHOTS_DIR, true);
1714 :
1715 : /*
1716 : * Now there's no way we can lose the dumped state anymore, remember this
1717 : * as a serialization point.
1718 : */
1719 620 : builder->last_serialized_snapshot = lsn;
1720 :
1721 620 : MemoryContextSwitchTo(old_ctx);
1722 :
1723 886 : out:
1724 886 : ReorderBufferSetRestartPoint(builder->reorder,
1725 : builder->last_serialized_snapshot);
1726 : /* be tidy */
1727 886 : if (ondisk)
1728 620 : pfree(ondisk);
1729 886 : if (catchange_xip)
1730 20 : pfree(catchange_xip);
1731 : }
1732 :
1733 : /*
1734 : * Restore the logical snapshot file contents to 'ondisk'.
1735 : *
1736 : * 'context' is the memory context where the catalog modifying/committed xid
1737 : * will live.
1738 : * If 'missing_ok' is true, will not throw an error if the file is not found.
1739 : */
1740 : bool
1741 42 : SnapBuildRestoreSnapshot(SnapBuildOnDisk *ondisk, XLogRecPtr lsn,
1742 : MemoryContext context, bool missing_ok)
1743 : {
1744 : int fd;
1745 : pg_crc32c checksum;
1746 : Size sz;
1747 : char path[MAXPGPATH];
1748 :
1749 42 : sprintf(path, "%s/%X-%X.snap",
1750 : PG_LOGICAL_SNAPSHOTS_DIR,
1751 42 : LSN_FORMAT_ARGS(lsn));
1752 :
1753 42 : fd = OpenTransientFile(path, O_RDONLY | PG_BINARY);
1754 :
1755 42 : if (fd < 0)
1756 : {
1757 22 : if (missing_ok && errno == ENOENT)
1758 22 : return false;
1759 :
1760 0 : ereport(ERROR,
1761 : (errcode_for_file_access(),
1762 : errmsg("could not open file \"%s\": %m", path)));
1763 : }
1764 :
1765 : /* ----
1766 : * Make sure the snapshot had been stored safely to disk, that's normally
1767 : * cheap.
1768 : * Note that we do not need PANIC here, nobody will be able to use the
1769 : * slot without fsyncing, and saving it won't succeed without an fsync()
1770 : * either...
1771 : * ----
1772 : */
1773 20 : fsync_fname(path, false);
1774 20 : fsync_fname(PG_LOGICAL_SNAPSHOTS_DIR, true);
1775 :
1776 : /* read statically sized portion of snapshot */
1777 20 : SnapBuildRestoreContents(fd, ondisk, SnapBuildOnDiskConstantSize, path);
1778 :
1779 20 : if (ondisk->magic != SNAPBUILD_MAGIC)
1780 0 : ereport(ERROR,
1781 : (errcode(ERRCODE_DATA_CORRUPTED),
1782 : errmsg("snapbuild state file \"%s\" has wrong magic number: %u instead of %u",
1783 : path, ondisk->magic, SNAPBUILD_MAGIC)));
1784 :
1785 20 : if (ondisk->version != SNAPBUILD_VERSION)
1786 0 : ereport(ERROR,
1787 : (errcode(ERRCODE_DATA_CORRUPTED),
1788 : errmsg("snapbuild state file \"%s\" has unsupported version: %u instead of %u",
1789 : path, ondisk->version, SNAPBUILD_VERSION)));
1790 :
1791 20 : INIT_CRC32C(checksum);
1792 20 : COMP_CRC32C(checksum,
1793 : ((char *) ondisk) + SnapBuildOnDiskNotChecksummedSize,
1794 : SnapBuildOnDiskConstantSize - SnapBuildOnDiskNotChecksummedSize);
1795 :
1796 : /* read SnapBuild */
1797 20 : SnapBuildRestoreContents(fd, &ondisk->builder, sizeof(SnapBuild), path);
1798 20 : COMP_CRC32C(checksum, &ondisk->builder, sizeof(SnapBuild));
1799 :
1800 : /* restore committed xacts information */
1801 20 : if (ondisk->builder.committed.xcnt > 0)
1802 : {
1803 8 : sz = sizeof(TransactionId) * ondisk->builder.committed.xcnt;
1804 8 : ondisk->builder.committed.xip = MemoryContextAllocZero(context, sz);
1805 8 : SnapBuildRestoreContents(fd, ondisk->builder.committed.xip, sz, path);
1806 8 : COMP_CRC32C(checksum, ondisk->builder.committed.xip, sz);
1807 : }
1808 :
1809 : /* restore catalog modifying xacts information */
1810 20 : if (ondisk->builder.catchange.xcnt > 0)
1811 : {
1812 10 : sz = sizeof(TransactionId) * ondisk->builder.catchange.xcnt;
1813 10 : ondisk->builder.catchange.xip = MemoryContextAllocZero(context, sz);
1814 10 : SnapBuildRestoreContents(fd, ondisk->builder.catchange.xip, sz, path);
1815 10 : COMP_CRC32C(checksum, ondisk->builder.catchange.xip, sz);
1816 : }
1817 :
1818 20 : if (CloseTransientFile(fd) != 0)
1819 0 : ereport(ERROR,
1820 : (errcode_for_file_access(),
1821 : errmsg("could not close file \"%s\": %m", path)));
1822 :
1823 20 : FIN_CRC32C(checksum);
1824 :
1825 : /* verify checksum of what we've read */
1826 20 : if (!EQ_CRC32C(checksum, ondisk->checksum))
1827 0 : ereport(ERROR,
1828 : (errcode(ERRCODE_DATA_CORRUPTED),
1829 : errmsg("checksum mismatch for snapbuild state file \"%s\": is %u, should be %u",
1830 : path, checksum, ondisk->checksum)));
1831 :
1832 20 : return true;
1833 : }
1834 :
1835 : /*
1836 : * Restore a snapshot into 'builder' if previously one has been stored at the
1837 : * location indicated by 'lsn'. Returns true if successful, false otherwise.
1838 : */
1839 : static bool
1840 38 : SnapBuildRestore(SnapBuild *builder, XLogRecPtr lsn)
1841 : {
1842 : SnapBuildOnDisk ondisk;
1843 :
1844 : /* no point in loading a snapshot if we're already there */
1845 38 : if (builder->state == SNAPBUILD_CONSISTENT)
1846 0 : return false;
1847 :
1848 : /* validate and restore the snapshot to 'ondisk' */
1849 38 : if (!SnapBuildRestoreSnapshot(&ondisk, lsn, builder->context, true))
1850 22 : return false;
1851 :
1852 : /*
1853 : * ok, we now have a sensible snapshot here, figure out if it has more
1854 : * information than we have.
1855 : */
1856 :
1857 : /*
1858 : * We are only interested in consistent snapshots for now, comparing
1859 : * whether one incomplete snapshot is more "advanced" seems to be
1860 : * unnecessarily complex.
1861 : */
1862 16 : if (ondisk.builder.state < SNAPBUILD_CONSISTENT)
1863 0 : goto snapshot_not_interesting;
1864 :
1865 : /*
1866 : * Don't use a snapshot that requires an xmin that we cannot guarantee to
1867 : * be available.
1868 : */
1869 16 : if (TransactionIdPrecedes(ondisk.builder.xmin, builder->initial_xmin_horizon))
1870 0 : goto snapshot_not_interesting;
1871 :
1872 : /*
1873 : * Consistent snapshots have no next phase. Reset next_phase_at as it is
1874 : * possible that an old value may remain.
1875 : */
1876 : Assert(ondisk.builder.next_phase_at == InvalidTransactionId);
1877 16 : builder->next_phase_at = InvalidTransactionId;
1878 :
1879 : /* ok, we think the snapshot is sensible, copy over everything important */
1880 16 : builder->xmin = ondisk.builder.xmin;
1881 16 : builder->xmax = ondisk.builder.xmax;
1882 16 : builder->state = ondisk.builder.state;
1883 :
1884 16 : builder->committed.xcnt = ondisk.builder.committed.xcnt;
1885 : /* We only allocated/stored xcnt, not xcnt_space xids ! */
1886 : /* don't overwrite preallocated xip, if we don't have anything here */
1887 16 : if (builder->committed.xcnt > 0)
1888 : {
1889 4 : pfree(builder->committed.xip);
1890 4 : builder->committed.xcnt_space = ondisk.builder.committed.xcnt;
1891 4 : builder->committed.xip = ondisk.builder.committed.xip;
1892 : }
1893 16 : ondisk.builder.committed.xip = NULL;
1894 :
1895 : /* set catalog modifying transactions */
1896 16 : if (builder->catchange.xip)
1897 0 : pfree(builder->catchange.xip);
1898 16 : builder->catchange.xcnt = ondisk.builder.catchange.xcnt;
1899 16 : builder->catchange.xip = ondisk.builder.catchange.xip;
1900 16 : ondisk.builder.catchange.xip = NULL;
1901 :
1902 : /* our snapshot is not interesting anymore, build a new one */
1903 16 : if (builder->snapshot != NULL)
1904 : {
1905 0 : SnapBuildSnapDecRefcount(builder->snapshot);
1906 : }
1907 16 : builder->snapshot = SnapBuildBuildSnapshot(builder);
1908 16 : SnapBuildSnapIncRefcount(builder->snapshot);
1909 :
1910 16 : ReorderBufferSetRestartPoint(builder->reorder, lsn);
1911 :
1912 : Assert(builder->state == SNAPBUILD_CONSISTENT);
1913 :
1914 16 : ereport(LOG,
1915 : errmsg("logical decoding found consistent point at %X/%08X",
1916 : LSN_FORMAT_ARGS(lsn)),
1917 : errdetail("Logical decoding will begin using saved snapshot."));
1918 16 : return true;
1919 :
1920 0 : snapshot_not_interesting:
1921 0 : if (ondisk.builder.committed.xip != NULL)
1922 0 : pfree(ondisk.builder.committed.xip);
1923 0 : if (ondisk.builder.catchange.xip != NULL)
1924 0 : pfree(ondisk.builder.catchange.xip);
1925 0 : return false;
1926 : }
1927 :
1928 : /*
1929 : * Read the contents of the serialized snapshot to 'dest'.
1930 : */
1931 : static void
1932 58 : SnapBuildRestoreContents(int fd, void *dest, Size size, const char *path)
1933 : {
1934 : int readBytes;
1935 :
1936 58 : pgstat_report_wait_start(WAIT_EVENT_SNAPBUILD_READ);
1937 58 : readBytes = read(fd, dest, size);
1938 58 : pgstat_report_wait_end();
1939 58 : if (readBytes != size)
1940 : {
1941 0 : int save_errno = errno;
1942 :
1943 0 : CloseTransientFile(fd);
1944 :
1945 0 : if (readBytes < 0)
1946 : {
1947 0 : errno = save_errno;
1948 0 : ereport(ERROR,
1949 : (errcode_for_file_access(),
1950 : errmsg("could not read file \"%s\": %m", path)));
1951 : }
1952 : else
1953 0 : ereport(ERROR,
1954 : (errcode(ERRCODE_DATA_CORRUPTED),
1955 : errmsg("could not read file \"%s\": read %d of %zu",
1956 : path, readBytes, size)));
1957 : }
1958 58 : }
1959 :
1960 : /*
1961 : * Remove all serialized snapshots that are not required anymore because no
1962 : * slot can need them. This doesn't actually have to run during a checkpoint,
1963 : * but it's a convenient point to schedule this.
1964 : *
1965 : * NB: We run this during checkpoints even if logical decoding is disabled so
1966 : * we cleanup old slots at some point after it got disabled.
1967 : */
1968 : void
1969 3492 : CheckPointSnapBuild(void)
1970 : {
1971 : XLogRecPtr cutoff;
1972 : XLogRecPtr redo;
1973 : DIR *snap_dir;
1974 : struct dirent *snap_de;
1975 : char path[MAXPGPATH + sizeof(PG_LOGICAL_SNAPSHOTS_DIR)];
1976 :
1977 : /*
1978 : * We start off with a minimum of the last redo pointer. No new
1979 : * replication slot will start before that, so that's a safe upper bound
1980 : * for removal.
1981 : */
1982 3492 : redo = GetRedoRecPtr();
1983 :
1984 : /* now check for the restart ptrs from existing slots */
1985 3492 : cutoff = ReplicationSlotsComputeLogicalRestartLSN();
1986 :
1987 : /* don't start earlier than the restart lsn */
1988 3492 : if (redo < cutoff)
1989 2 : cutoff = redo;
1990 :
1991 3492 : snap_dir = AllocateDir(PG_LOGICAL_SNAPSHOTS_DIR);
1992 11020 : while ((snap_de = ReadDir(snap_dir, PG_LOGICAL_SNAPSHOTS_DIR)) != NULL)
1993 : {
1994 : uint32 hi;
1995 : uint32 lo;
1996 : XLogRecPtr lsn;
1997 : PGFileType de_type;
1998 :
1999 7528 : if (strcmp(snap_de->d_name, ".") == 0 ||
2000 4036 : strcmp(snap_de->d_name, "..") == 0)
2001 6984 : continue;
2002 :
2003 544 : snprintf(path, sizeof(path), "%s/%s", PG_LOGICAL_SNAPSHOTS_DIR, snap_de->d_name);
2004 544 : de_type = get_dirent_type(path, snap_de, false, DEBUG1);
2005 :
2006 544 : if (de_type != PGFILETYPE_ERROR && de_type != PGFILETYPE_REG)
2007 : {
2008 0 : elog(DEBUG1, "only regular files expected: %s", path);
2009 0 : continue;
2010 : }
2011 :
2012 : /*
2013 : * temporary filenames from SnapBuildSerialize() include the LSN and
2014 : * everything but are postfixed by .$pid.tmp. We can just remove them
2015 : * the same as other files because there can be none that are
2016 : * currently being written that are older than cutoff.
2017 : *
2018 : * We just log a message if a file doesn't fit the pattern, it's
2019 : * probably some editors lock/state file or similar...
2020 : */
2021 544 : if (sscanf(snap_de->d_name, "%X-%X.snap", &hi, &lo) != 2)
2022 : {
2023 0 : ereport(LOG,
2024 : (errmsg("could not parse file name \"%s\"", path)));
2025 0 : continue;
2026 : }
2027 :
2028 544 : lsn = ((uint64) hi) << 32 | lo;
2029 :
2030 : /* check whether we still need it */
2031 544 : if (lsn < cutoff || !XLogRecPtrIsValid(cutoff))
2032 : {
2033 366 : elog(DEBUG1, "removing snapbuild snapshot %s", path);
2034 :
2035 : /*
2036 : * It's not particularly harmful, though strange, if we can't
2037 : * remove the file here. Don't prevent the checkpoint from
2038 : * completing, that'd be a cure worse than the disease.
2039 : */
2040 366 : if (unlink(path) < 0)
2041 : {
2042 0 : ereport(LOG,
2043 : (errcode_for_file_access(),
2044 : errmsg("could not remove file \"%s\": %m",
2045 : path)));
2046 0 : continue;
2047 : }
2048 : }
2049 : }
2050 3492 : FreeDir(snap_dir);
2051 3492 : }
2052 :
2053 : /*
2054 : * Check if a logical snapshot at the specified point has been serialized.
2055 : */
2056 : bool
2057 24 : SnapBuildSnapshotExists(XLogRecPtr lsn)
2058 : {
2059 : char path[MAXPGPATH];
2060 : int ret;
2061 : struct stat stat_buf;
2062 :
2063 24 : sprintf(path, "%s/%X-%X.snap",
2064 : PG_LOGICAL_SNAPSHOTS_DIR,
2065 24 : LSN_FORMAT_ARGS(lsn));
2066 :
2067 24 : ret = stat(path, &stat_buf);
2068 :
2069 24 : if (ret != 0 && errno != ENOENT)
2070 0 : ereport(ERROR,
2071 : (errcode_for_file_access(),
2072 : errmsg("could not stat file \"%s\": %m", path)));
2073 :
2074 24 : return ret == 0;
2075 : }
|
