Line data Source code
1 : /*-------------------------------------------------------------------------
2 : * logical.c
3 : * PostgreSQL logical decoding coordination
4 : *
5 : * Copyright (c) 2012-2026, PostgreSQL Global Development Group
6 : *
7 : * IDENTIFICATION
8 : * src/backend/replication/logical/logical.c
9 : *
10 : * NOTES
11 : * This file coordinates interaction between the various modules that
12 : * together provide logical decoding, primarily by providing so
13 : * called LogicalDecodingContexts. The goal is to encapsulate most of the
14 : * internal complexity for consumers of logical decoding, so they can
15 : * create and consume a changestream with a low amount of code. Builtin
16 : * consumers are the walsender and SQL SRF interface, but it's possible to
17 : * add further ones without changing core code, e.g. to consume changes in
18 : * a bgworker.
19 : *
20 : * The idea is that a consumer provides three callbacks, one to read WAL,
21 : * one to prepare a data write, and a final one for actually writing since
22 : * their implementation depends on the type of consumer. Check
23 : * logicalfuncs.c for an example implementation of a fairly simple consumer
24 : * and an implementation of a WAL reading callback that's suitable for
25 : * simple consumers.
26 : *-------------------------------------------------------------------------
27 : */
28 :
29 : #include "postgres.h"
30 :
31 : #include "access/xact.h"
32 : #include "access/xlog_internal.h"
33 : #include "access/xlogutils.h"
34 : #include "fmgr.h"
35 : #include "miscadmin.h"
36 : #include "pgstat.h"
37 : #include "replication/decode.h"
38 : #include "replication/logical.h"
39 : #include "replication/reorderbuffer.h"
40 : #include "replication/slotsync.h"
41 : #include "replication/snapbuild.h"
42 : #include "storage/proc.h"
43 : #include "storage/procarray.h"
44 : #include "utils/builtins.h"
45 : #include "utils/injection_point.h"
46 : #include "utils/inval.h"
47 : #include "utils/memutils.h"
48 :
49 : /* data for errcontext callback */
50 : typedef struct LogicalErrorCallbackState
51 : {
52 : LogicalDecodingContext *ctx;
53 : const char *callback_name;
54 : XLogRecPtr report_location;
55 : } LogicalErrorCallbackState;
56 :
57 : /* wrappers around output plugin callbacks */
58 : static void output_plugin_error_callback(void *arg);
59 : static void startup_cb_wrapper(LogicalDecodingContext *ctx, OutputPluginOptions *opt,
60 : bool is_init);
61 : static void shutdown_cb_wrapper(LogicalDecodingContext *ctx);
62 : static void begin_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn);
63 : static void commit_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
64 : XLogRecPtr commit_lsn);
65 : static void begin_prepare_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn);
66 : static void prepare_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
67 : XLogRecPtr prepare_lsn);
68 : static void commit_prepared_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
69 : XLogRecPtr commit_lsn);
70 : static void rollback_prepared_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
71 : XLogRecPtr prepare_end_lsn, TimestampTz prepare_time);
72 : static void change_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
73 : Relation relation, ReorderBufferChange *change);
74 : static void truncate_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
75 : int nrelations, Relation relations[], ReorderBufferChange *change);
76 : static void message_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
77 : XLogRecPtr message_lsn, bool transactional,
78 : const char *prefix, Size message_size, const char *message);
79 :
80 : /* streaming callbacks */
81 : static void stream_start_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
82 : XLogRecPtr first_lsn);
83 : static void stream_stop_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
84 : XLogRecPtr last_lsn);
85 : static void stream_abort_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
86 : XLogRecPtr abort_lsn);
87 : static void stream_prepare_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
88 : XLogRecPtr prepare_lsn);
89 : static void stream_commit_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
90 : XLogRecPtr commit_lsn);
91 : static void stream_change_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
92 : Relation relation, ReorderBufferChange *change);
93 : static void stream_message_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
94 : XLogRecPtr message_lsn, bool transactional,
95 : const char *prefix, Size message_size, const char *message);
96 : static void stream_truncate_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
97 : int nrelations, Relation relations[], ReorderBufferChange *change);
98 :
99 : /* callback to update txn's progress */
100 : static void update_progress_txn_cb_wrapper(ReorderBuffer *cache,
101 : ReorderBufferTXN *txn,
102 : XLogRecPtr lsn);
103 :
104 : static void LoadOutputPlugin(OutputPluginCallbacks *callbacks, const char *plugin);
105 :
106 : /*
107 : * Make sure the current settings & environment are capable of doing logical
108 : * decoding.
109 : */
110 : void
111 1712 : CheckLogicalDecodingRequirements(bool repack)
112 : {
113 1712 : CheckSlotRequirements(repack);
114 :
115 : /*
116 : * NB: Adding a new requirement likely means that RestoreSlotFromDisk()
117 : * needs the same check.
118 : */
119 :
120 1712 : if (MyDatabaseId == InvalidOid)
121 1 : ereport(ERROR,
122 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
123 : errmsg("logical decoding requires a database connection")));
124 :
125 : /* CheckSlotRequirements() has already checked if wal_level >= 'replica' */
126 : Assert(wal_level >= WAL_LEVEL_REPLICA);
127 :
128 : /* Check if logical decoding is available on standby */
129 1711 : if (RecoveryInProgress() && !IsLogicalDecodingEnabled())
130 2 : ereport(ERROR,
131 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
132 : errmsg("logical decoding on standby requires \"effective_wal_level\" >= \"logical\" on the primary"),
133 : errhint("Set \"wal_level\" >= \"logical\" or create at least one logical slot when \"wal_level\" = \"replica\".")));
134 1709 : }
135 :
136 : /*
137 : * Helper function for CreateInitDecodingContext() and
138 : * CreateDecodingContext() performing common tasks.
139 : */
140 : static LogicalDecodingContext *
141 1211 : StartupDecodingContext(List *output_plugin_options,
142 : XLogRecPtr start_lsn,
143 : TransactionId xmin_horizon,
144 : bool need_full_snapshot,
145 : bool fast_forward,
146 : bool in_create,
147 : XLogReaderRoutine *xl_routine,
148 : LogicalOutputPluginWriterPrepareWrite prepare_write,
149 : LogicalOutputPluginWriterWrite do_write,
150 : LogicalOutputPluginWriterUpdateProgress update_progress)
151 : {
152 : ReplicationSlot *slot;
153 : MemoryContext context,
154 : old_context;
155 : LogicalDecodingContext *ctx;
156 :
157 : /* shorter lines... */
158 1211 : slot = MyReplicationSlot;
159 :
160 1211 : context = AllocSetContextCreate(CurrentMemoryContext,
161 : "Logical decoding context",
162 : ALLOCSET_DEFAULT_SIZES);
163 1211 : old_context = MemoryContextSwitchTo(context);
164 1211 : ctx = palloc0_object(LogicalDecodingContext);
165 :
166 1211 : ctx->context = context;
167 :
168 : /*
169 : * (re-)load output plugins, so we detect a bad (removed) output plugin
170 : * now.
171 : */
172 1211 : if (!fast_forward)
173 1186 : LoadOutputPlugin(&ctx->callbacks, NameStr(slot->data.plugin));
174 :
175 : /*
176 : * Now that the slot's xmin has been set, we can announce ourselves as a
177 : * logical decoding backend which doesn't need to be checked individually
178 : * when computing the xmin horizon because the xmin is enforced via
179 : * replication slots.
180 : *
181 : * We can only do so if we're outside of a transaction (i.e. the case when
182 : * streaming changes via walsender), otherwise an already setup
183 : * snapshot/xid would end up being ignored. That's not a particularly
184 : * bothersome restriction since the SQL interface can't be used for
185 : * streaming anyway.
186 : */
187 1210 : if (!IsTransactionOrTransactionBlock())
188 : {
189 619 : LWLockAcquire(ProcArrayLock, LW_EXCLUSIVE);
190 619 : MyProc->statusFlags |= PROC_IN_LOGICAL_DECODING;
191 619 : ProcGlobal->statusFlags[MyProc->pgxactoff] = MyProc->statusFlags;
192 619 : LWLockRelease(ProcArrayLock);
193 : }
194 :
195 1210 : ctx->slot = slot;
196 :
197 1210 : ctx->reader = XLogReaderAllocate(wal_segment_size, NULL, xl_routine, ctx);
198 1210 : if (!ctx->reader)
199 0 : ereport(ERROR,
200 : (errcode(ERRCODE_OUT_OF_MEMORY),
201 : errmsg("out of memory"),
202 : errdetail("Failed while allocating a WAL reading processor.")));
203 :
204 1210 : ctx->reorder = ReorderBufferAllocate();
205 1210 : ctx->snapshot_builder =
206 1210 : AllocateSnapshotBuilder(ctx->reorder, xmin_horizon, start_lsn,
207 : need_full_snapshot, in_create, slot->data.two_phase_at);
208 :
209 1210 : ctx->reorder->private_data = ctx;
210 :
211 : /* wrap output plugin callbacks, so we can add error context information */
212 1210 : ctx->reorder->begin = begin_cb_wrapper;
213 1210 : ctx->reorder->apply_change = change_cb_wrapper;
214 1210 : ctx->reorder->apply_truncate = truncate_cb_wrapper;
215 1210 : ctx->reorder->commit = commit_cb_wrapper;
216 1210 : ctx->reorder->message = message_cb_wrapper;
217 :
218 : /*
219 : * To support streaming, we require start/stop/abort/commit/change
220 : * callbacks. The message and truncate callbacks are optional, similar to
221 : * regular output plugins. We however enable streaming when at least one
222 : * of the methods is enabled so that we can easily identify missing
223 : * methods.
224 : *
225 : * We decide it here, but only check it later in the wrappers.
226 : */
227 2447 : ctx->streaming = (ctx->callbacks.stream_start_cb != NULL) ||
228 27 : (ctx->callbacks.stream_stop_cb != NULL) ||
229 27 : (ctx->callbacks.stream_abort_cb != NULL) ||
230 27 : (ctx->callbacks.stream_commit_cb != NULL) ||
231 27 : (ctx->callbacks.stream_change_cb != NULL) ||
232 1264 : (ctx->callbacks.stream_message_cb != NULL) ||
233 27 : (ctx->callbacks.stream_truncate_cb != NULL);
234 :
235 : /*
236 : * streaming callbacks
237 : *
238 : * stream_message and stream_truncate callbacks are optional, so we do not
239 : * fail with ERROR when missing, but the wrappers simply do nothing. We
240 : * must set the ReorderBuffer callbacks to something, otherwise the calls
241 : * from there will crash (we don't want to move the checks there).
242 : */
243 1210 : ctx->reorder->stream_start = stream_start_cb_wrapper;
244 1210 : ctx->reorder->stream_stop = stream_stop_cb_wrapper;
245 1210 : ctx->reorder->stream_abort = stream_abort_cb_wrapper;
246 1210 : ctx->reorder->stream_prepare = stream_prepare_cb_wrapper;
247 1210 : ctx->reorder->stream_commit = stream_commit_cb_wrapper;
248 1210 : ctx->reorder->stream_change = stream_change_cb_wrapper;
249 1210 : ctx->reorder->stream_message = stream_message_cb_wrapper;
250 1210 : ctx->reorder->stream_truncate = stream_truncate_cb_wrapper;
251 :
252 :
253 : /*
254 : * To support two-phase logical decoding, we require
255 : * begin_prepare/prepare/commit-prepare/abort-prepare callbacks. The
256 : * filter_prepare callback is optional. We however enable two-phase
257 : * logical decoding when at least one of the methods is enabled so that we
258 : * can easily identify missing methods.
259 : *
260 : * We decide it here, but only check it later in the wrappers.
261 : */
262 2447 : ctx->twophase = (ctx->callbacks.begin_prepare_cb != NULL) ||
263 27 : (ctx->callbacks.prepare_cb != NULL) ||
264 27 : (ctx->callbacks.commit_prepared_cb != NULL) ||
265 27 : (ctx->callbacks.rollback_prepared_cb != NULL) ||
266 1264 : (ctx->callbacks.stream_prepare_cb != NULL) ||
267 27 : (ctx->callbacks.filter_prepare_cb != NULL);
268 :
269 : /*
270 : * Callback to support decoding at prepare time.
271 : */
272 1210 : ctx->reorder->begin_prepare = begin_prepare_cb_wrapper;
273 1210 : ctx->reorder->prepare = prepare_cb_wrapper;
274 1210 : ctx->reorder->commit_prepared = commit_prepared_cb_wrapper;
275 1210 : ctx->reorder->rollback_prepared = rollback_prepared_cb_wrapper;
276 :
277 : /*
278 : * Callback to support updating progress during sending data of a
279 : * transaction (and its subtransactions) to the output plugin.
280 : */
281 1210 : ctx->reorder->update_progress_txn = update_progress_txn_cb_wrapper;
282 :
283 1210 : ctx->out = makeStringInfo();
284 1210 : ctx->prepare_write = prepare_write;
285 1210 : ctx->write = do_write;
286 1210 : ctx->update_progress = update_progress;
287 :
288 : /* Assume shared catalog access. The startup callback can change it. */
289 1210 : ctx->options.need_shared_catalogs = true;
290 :
291 1210 : ctx->output_plugin_options = output_plugin_options;
292 :
293 1210 : ctx->fast_forward = fast_forward;
294 :
295 1210 : MemoryContextSwitchTo(old_context);
296 :
297 1210 : return ctx;
298 : }
299 :
300 : /*
301 : * Create a new decoding context, for a new logical slot.
302 : *
303 : * plugin -- contains the name of the output plugin
304 : * output_plugin_options -- contains options passed to the output plugin
305 : * need_full_snapshot -- if true, must obtain a snapshot able to read all
306 : * tables; if false, one that can read only catalogs is acceptable.
307 : * for_repack -- if true, we're going to be decoding for REPACK.
308 : * restart_lsn -- if given as invalid, it's this routine's responsibility to
309 : * mark WAL as reserved by setting a convenient restart_lsn for the slot.
310 : * Otherwise, we set for decoding to start from the given LSN without
311 : * marking WAL reserved beforehand. In that scenario, it's up to the
312 : * caller to guarantee that WAL remains available.
313 : * xl_routine -- XLogReaderRoutine for underlying XLogReader
314 : * prepare_write, do_write, update_progress --
315 : * callbacks that perform the use-case dependent, actual, work.
316 : *
317 : * Needs to be called while in a memory context that's at least as long lived
318 : * as the decoding context because further memory contexts will be created
319 : * inside it.
320 : *
321 : * Returns an initialized decoding context after calling the output plugin's
322 : * startup function.
323 : */
324 : LogicalDecodingContext *
325 502 : CreateInitDecodingContext(const char *plugin,
326 : List *output_plugin_options,
327 : bool need_full_snapshot,
328 : bool for_repack,
329 : XLogRecPtr restart_lsn,
330 : XLogReaderRoutine *xl_routine,
331 : LogicalOutputPluginWriterPrepareWrite prepare_write,
332 : LogicalOutputPluginWriterWrite do_write,
333 : LogicalOutputPluginWriterUpdateProgress update_progress)
334 : {
335 502 : TransactionId xmin_horizon = InvalidTransactionId;
336 : ReplicationSlot *slot;
337 : NameData plugin_name;
338 : LogicalDecodingContext *ctx;
339 : MemoryContext old_context;
340 :
341 : /*
342 : * On a standby, this check is also required while creating the slot.
343 : * Check the comments in the function.
344 : */
345 502 : CheckLogicalDecodingRequirements(for_repack);
346 :
347 : /* shorter lines... */
348 502 : slot = MyReplicationSlot;
349 :
350 : /* first some sanity checks that are unlikely to be violated */
351 502 : if (slot == NULL)
352 0 : elog(ERROR, "cannot perform logical decoding without an acquired slot");
353 :
354 502 : if (plugin == NULL)
355 0 : elog(ERROR, "cannot initialize logical decoding without a specified plugin");
356 :
357 : /* Make sure the passed slot is suitable. These are user facing errors. */
358 502 : if (SlotIsPhysical(slot))
359 0 : ereport(ERROR,
360 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
361 : errmsg("cannot use physical replication slot for logical decoding")));
362 :
363 502 : if (slot->data.database != MyDatabaseId)
364 0 : ereport(ERROR,
365 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
366 : errmsg("replication slot \"%s\" was not created in this database",
367 : NameStr(slot->data.name))));
368 :
369 854 : if (IsTransactionState() &&
370 352 : GetTopTransactionIdIfAny() != InvalidTransactionId)
371 2 : ereport(ERROR,
372 : (errcode(ERRCODE_ACTIVE_SQL_TRANSACTION),
373 : errmsg("cannot create logical replication slot in transaction that has performed writes")));
374 :
375 : /*
376 : * Register output plugin name with slot. We need the mutex to avoid
377 : * concurrent reading of a partially copied string. But we don't want any
378 : * complicated code while holding a spinlock, so do namestrcpy() outside.
379 : */
380 500 : namestrcpy(&plugin_name, plugin);
381 500 : SpinLockAcquire(&slot->mutex);
382 500 : slot->data.plugin = plugin_name;
383 500 : SpinLockRelease(&slot->mutex);
384 :
385 500 : if (!XLogRecPtrIsValid(restart_lsn))
386 493 : ReplicationSlotReserveWal();
387 : else
388 : {
389 7 : SpinLockAcquire(&slot->mutex);
390 7 : slot->data.restart_lsn = restart_lsn;
391 7 : SpinLockRelease(&slot->mutex);
392 : }
393 :
394 : /* ----
395 : * This is a bit tricky: We need to determine a safe xmin horizon to start
396 : * decoding from, to avoid starting from a running xacts record referring
397 : * to xids whose rows have been vacuumed or pruned
398 : * already. GetOldestSafeDecodingTransactionId() returns such a value, but
399 : * without further interlock its return value might immediately be out of
400 : * date.
401 : *
402 : * So we have to acquire both the ReplicationSlotControlLock and the
403 : * ProcArrayLock to prevent concurrent computation and update of new xmin
404 : * horizons by other backends, get the safe decoding xid, and inform the
405 : * slot machinery about the new limit. Once that's done both locks can be
406 : * released as the slot machinery now is protecting against vacuum.
407 : *
408 : * Note that, temporarily, the data, not just the catalog, xmin has to be
409 : * reserved if a data snapshot is to be exported. Otherwise the initial
410 : * data snapshot created here is not guaranteed to be valid. After that
411 : * the data xmin doesn't need to be managed anymore and the global xmin
412 : * should be recomputed. As we are fine with losing the pegged data xmin
413 : * after crash - no chance a snapshot would get exported anymore - we can
414 : * get away with just setting the slot's
415 : * effective_xmin. ReplicationSlotRelease will reset it again.
416 : *
417 : * ----
418 : */
419 500 : LWLockAcquire(ReplicationSlotControlLock, LW_EXCLUSIVE);
420 500 : LWLockAcquire(ProcArrayLock, LW_EXCLUSIVE);
421 :
422 500 : xmin_horizon = GetOldestSafeDecodingTransactionId(!need_full_snapshot);
423 :
424 500 : SpinLockAcquire(&slot->mutex);
425 500 : slot->effective_catalog_xmin = xmin_horizon;
426 500 : slot->data.catalog_xmin = xmin_horizon;
427 500 : if (need_full_snapshot)
428 217 : slot->effective_xmin = xmin_horizon;
429 500 : SpinLockRelease(&slot->mutex);
430 :
431 500 : ReplicationSlotsComputeRequiredXmin(true);
432 :
433 500 : LWLockRelease(ProcArrayLock);
434 500 : LWLockRelease(ReplicationSlotControlLock);
435 :
436 500 : ReplicationSlotMarkDirty();
437 500 : ReplicationSlotSave();
438 :
439 500 : ctx = StartupDecodingContext(NIL, restart_lsn, xmin_horizon,
440 : need_full_snapshot, false, true,
441 : xl_routine, prepare_write, do_write,
442 : update_progress);
443 :
444 : /* call output plugin initialization callback */
445 499 : old_context = MemoryContextSwitchTo(ctx->context);
446 499 : if (ctx->callbacks.startup_cb != NULL)
447 499 : startup_cb_wrapper(ctx, &ctx->options, true);
448 499 : MemoryContextSwitchTo(old_context);
449 :
450 : /*
451 : * We allow decoding of prepared transactions when the two_phase is
452 : * enabled at the time of slot creation, or when the two_phase option is
453 : * given at the streaming start, provided the plugin supports all the
454 : * callbacks for two-phase.
455 : */
456 499 : ctx->twophase &= slot->data.two_phase;
457 :
458 499 : ctx->reorder->output_rewrites = ctx->options.receive_rewrites;
459 :
460 499 : return ctx;
461 : }
462 :
463 : /*
464 : * Create a new decoding context, for a logical slot that has previously been
465 : * used already.
466 : *
467 : * start_lsn
468 : * The LSN at which to start decoding. If InvalidXLogRecPtr, restart
469 : * from the slot's confirmed_flush; otherwise, start from the specified
470 : * location (but move it forwards to confirmed_flush if it's older than
471 : * that, see below).
472 : *
473 : * output_plugin_options
474 : * options passed to the output plugin.
475 : *
476 : * fast_forward
477 : * bypass the generation of logical changes.
478 : *
479 : * xl_routine
480 : * XLogReaderRoutine used by underlying xlogreader
481 : *
482 : * prepare_write, do_write, update_progress
483 : * callbacks that have to be filled to perform the use-case dependent,
484 : * actual work.
485 : *
486 : * Needs to be called while in a memory context that's at least as long lived
487 : * as the decoding context because further memory contexts will be created
488 : * inside it.
489 : *
490 : * Returns an initialized decoding context after calling the output plugin's
491 : * startup function.
492 : */
493 : LogicalDecodingContext *
494 716 : CreateDecodingContext(XLogRecPtr start_lsn,
495 : List *output_plugin_options,
496 : bool fast_forward,
497 : XLogReaderRoutine *xl_routine,
498 : LogicalOutputPluginWriterPrepareWrite prepare_write,
499 : LogicalOutputPluginWriterWrite do_write,
500 : LogicalOutputPluginWriterUpdateProgress update_progress)
501 : {
502 : LogicalDecodingContext *ctx;
503 : ReplicationSlot *slot;
504 : MemoryContext old_context;
505 :
506 : /* shorter lines... */
507 716 : slot = MyReplicationSlot;
508 :
509 : /* first some sanity checks that are unlikely to be violated */
510 716 : if (slot == NULL)
511 0 : elog(ERROR, "cannot perform logical decoding without an acquired slot");
512 :
513 : /* make sure the passed slot is suitable, these are user facing errors */
514 716 : if (SlotIsPhysical(slot))
515 1 : ereport(ERROR,
516 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
517 : errmsg("cannot use physical replication slot for logical decoding")));
518 :
519 : /*
520 : * We need to access the system tables during decoding to build the
521 : * logical changes unless we are in fast_forward mode where no changes are
522 : * generated.
523 : */
524 715 : if (slot->data.database != MyDatabaseId && !fast_forward)
525 3 : ereport(ERROR,
526 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
527 : errmsg("replication slot \"%s\" was not created in this database",
528 : NameStr(slot->data.name))));
529 :
530 : /*
531 : * The slots being synced from the primary can't be used for decoding as
532 : * they are used after failover. However, we do allow advancing the LSNs
533 : * during the synchronization of slots. See update_local_synced_slot.
534 : */
535 712 : if (RecoveryInProgress() && slot->data.synced && !IsSyncingReplicationSlots())
536 1 : ereport(ERROR,
537 : errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
538 : errmsg("cannot use replication slot \"%s\" for logical decoding",
539 : NameStr(slot->data.name)),
540 : errdetail("This replication slot is being synchronized from the primary server."),
541 : errhint("Specify another replication slot."));
542 :
543 : /* slot must be valid to allow decoding */
544 : Assert(slot->data.invalidated == RS_INVAL_NONE);
545 : Assert(XLogRecPtrIsValid(slot->data.restart_lsn));
546 :
547 711 : if (!XLogRecPtrIsValid(start_lsn))
548 : {
549 : /* continue from last position */
550 393 : start_lsn = slot->data.confirmed_flush;
551 : }
552 318 : else if (start_lsn < slot->data.confirmed_flush)
553 : {
554 : /*
555 : * It might seem like we should error out in this case, but it's
556 : * pretty common for a client to acknowledge a LSN it doesn't have to
557 : * do anything for, and thus didn't store persistently, because the
558 : * xlog records didn't result in anything relevant for logical
559 : * decoding. Clients have to be able to do that to support synchronous
560 : * replication.
561 : *
562 : * Starting at a different LSN than requested might not catch certain
563 : * kinds of client errors; so the client may wish to check that
564 : * confirmed_flush_lsn matches its expectations.
565 : */
566 33 : elog(LOG, "%X/%08X has been already streamed, forwarding to %X/%08X",
567 : LSN_FORMAT_ARGS(start_lsn),
568 : LSN_FORMAT_ARGS(slot->data.confirmed_flush));
569 :
570 33 : start_lsn = slot->data.confirmed_flush;
571 : }
572 :
573 711 : ctx = StartupDecodingContext(output_plugin_options,
574 : start_lsn, InvalidTransactionId, false,
575 : fast_forward, false, xl_routine, prepare_write,
576 : do_write, update_progress);
577 :
578 : /* call output plugin initialization callback */
579 711 : old_context = MemoryContextSwitchTo(ctx->context);
580 711 : if (ctx->callbacks.startup_cb != NULL)
581 686 : startup_cb_wrapper(ctx, &ctx->options, false);
582 708 : MemoryContextSwitchTo(old_context);
583 :
584 : /*
585 : * We allow decoding of prepared transactions when the two_phase is
586 : * enabled at the time of slot creation, or when the two_phase option is
587 : * given at the streaming start, provided the plugin supports all the
588 : * callbacks for two-phase.
589 : */
590 708 : ctx->twophase &= (slot->data.two_phase || ctx->twophase_opt_given);
591 :
592 : /* Mark slot to allow two_phase decoding if not already marked */
593 708 : if (ctx->twophase && !slot->data.two_phase)
594 : {
595 8 : SpinLockAcquire(&slot->mutex);
596 8 : slot->data.two_phase = true;
597 8 : slot->data.two_phase_at = start_lsn;
598 8 : SpinLockRelease(&slot->mutex);
599 8 : ReplicationSlotMarkDirty();
600 8 : ReplicationSlotSave();
601 8 : SnapBuildSetTwoPhaseAt(ctx->snapshot_builder, start_lsn);
602 : }
603 :
604 708 : ctx->reorder->output_rewrites = ctx->options.receive_rewrites;
605 :
606 708 : ereport(LogicalDecodingLogLevel(),
607 : (errmsg("starting logical decoding for slot \"%s\"",
608 : NameStr(slot->data.name)),
609 : errdetail("Streaming transactions committing after %X/%08X, reading WAL from %X/%08X.",
610 : LSN_FORMAT_ARGS(slot->data.confirmed_flush),
611 : LSN_FORMAT_ARGS(slot->data.restart_lsn))));
612 :
613 708 : return ctx;
614 : }
615 :
616 : /*
617 : * Returns true if a consistent initial decoding snapshot has been built.
618 : */
619 : bool
620 1991 : DecodingContextReady(LogicalDecodingContext *ctx)
621 : {
622 1991 : return SnapBuildCurrentState(ctx->snapshot_builder) == SNAPBUILD_CONSISTENT;
623 : }
624 :
625 : /*
626 : * Read from the decoding slot, until it is ready to start extracting changes.
627 : */
628 : void
629 492 : DecodingContextFindStartpoint(LogicalDecodingContext *ctx)
630 : {
631 492 : ReplicationSlot *slot = ctx->slot;
632 :
633 : /* Initialize from where to start reading WAL. */
634 492 : XLogBeginRead(ctx->reader, slot->data.restart_lsn);
635 :
636 492 : elog(DEBUG1, "searching for logical decoding starting point, starting at %X/%08X",
637 : LSN_FORMAT_ARGS(slot->data.restart_lsn));
638 :
639 : /* Wait for a consistent starting point */
640 : for (;;)
641 1490 : {
642 : XLogRecord *record;
643 1982 : char *err = NULL;
644 :
645 : /* the read_page callback waits for new WAL */
646 1982 : record = XLogReadRecord(ctx->reader, &err);
647 1982 : if (err)
648 0 : elog(ERROR, "could not find logical decoding starting point: %s", err);
649 1982 : if (!record)
650 0 : elog(ERROR, "could not find logical decoding starting point");
651 :
652 1982 : LogicalDecodingProcessRecord(ctx, ctx->reader);
653 :
654 : /* only continue till we found a consistent spot */
655 1980 : if (DecodingContextReady(ctx))
656 490 : break;
657 :
658 1490 : CHECK_FOR_INTERRUPTS();
659 : }
660 :
661 490 : SpinLockAcquire(&slot->mutex);
662 490 : slot->data.confirmed_flush = ctx->reader->EndRecPtr;
663 490 : if (slot->data.two_phase)
664 9 : slot->data.two_phase_at = ctx->reader->EndRecPtr;
665 490 : SpinLockRelease(&slot->mutex);
666 490 : }
667 :
668 : /*
669 : * Free a previously allocated decoding context, invoking the shutdown
670 : * callback if necessary.
671 : */
672 : void
673 943 : FreeDecodingContext(LogicalDecodingContext *ctx)
674 : {
675 943 : if (ctx->callbacks.shutdown_cb != NULL)
676 918 : shutdown_cb_wrapper(ctx);
677 :
678 943 : ReorderBufferFree(ctx->reorder);
679 943 : FreeSnapshotBuilder(ctx->snapshot_builder);
680 943 : XLogReaderFree(ctx->reader);
681 943 : MemoryContextDelete(ctx->context);
682 943 : }
683 :
684 : /*
685 : * Prepare a write using the context's output routine.
686 : */
687 : void
688 362405 : OutputPluginPrepareWrite(struct LogicalDecodingContext *ctx, bool last_write)
689 : {
690 362405 : if (!ctx->accept_writes)
691 0 : elog(ERROR, "writes are only accepted in commit, begin and change callbacks");
692 :
693 362405 : ctx->prepare_write(ctx, ctx->write_location, ctx->write_xid, last_write);
694 362405 : ctx->prepared_write = true;
695 362405 : }
696 :
697 : /*
698 : * Perform a write using the context's output routine.
699 : */
700 : void
701 362405 : OutputPluginWrite(struct LogicalDecodingContext *ctx, bool last_write)
702 : {
703 362405 : if (!ctx->prepared_write)
704 0 : elog(ERROR, "OutputPluginPrepareWrite needs to be called before OutputPluginWrite");
705 :
706 362405 : ctx->write(ctx, ctx->write_location, ctx->write_xid, last_write);
707 362391 : ctx->prepared_write = false;
708 362391 : }
709 :
710 : /*
711 : * Update progress tracking (if supported).
712 : */
713 : void
714 4838 : OutputPluginUpdateProgress(struct LogicalDecodingContext *ctx,
715 : bool skipped_xact)
716 : {
717 4838 : if (!ctx->update_progress)
718 1594 : return;
719 :
720 3244 : ctx->update_progress(ctx, ctx->write_location, ctx->write_xid,
721 : skipped_xact);
722 : }
723 :
724 : /*
725 : * Load the output plugin, lookup its output plugin init function, and check
726 : * that it provides the required callbacks.
727 : */
728 : static void
729 1186 : LoadOutputPlugin(OutputPluginCallbacks *callbacks, const char *plugin)
730 : {
731 : LogicalOutputPluginInit plugin_init;
732 :
733 1185 : plugin_init = (LogicalOutputPluginInit)
734 1186 : load_external_function(plugin, "_PG_output_plugin_init", false, NULL);
735 :
736 1185 : if (plugin_init == NULL)
737 0 : elog(ERROR, "output plugins have to declare the _PG_output_plugin_init symbol");
738 :
739 : /* ask the output plugin to fill the callback struct */
740 1185 : plugin_init(callbacks);
741 :
742 1185 : if (callbacks->begin_cb == NULL)
743 0 : elog(ERROR, "output plugins have to register a begin callback");
744 1185 : if (callbacks->change_cb == NULL)
745 0 : elog(ERROR, "output plugins have to register a change callback");
746 1185 : if (callbacks->commit_cb == NULL)
747 0 : elog(ERROR, "output plugins have to register a commit callback");
748 1185 : }
749 :
750 : static void
751 481 : output_plugin_error_callback(void *arg)
752 : {
753 481 : LogicalErrorCallbackState *state = (LogicalErrorCallbackState *) arg;
754 :
755 : /* not all callbacks have an associated LSN */
756 481 : if (XLogRecPtrIsValid(state->report_location))
757 478 : errcontext("slot \"%s\", output plugin \"%s\", in the %s callback, associated LSN %X/%08X",
758 478 : NameStr(state->ctx->slot->data.name),
759 478 : NameStr(state->ctx->slot->data.plugin),
760 : state->callback_name,
761 478 : LSN_FORMAT_ARGS(state->report_location));
762 : else
763 3 : errcontext("slot \"%s\", output plugin \"%s\", in the %s callback",
764 3 : NameStr(state->ctx->slot->data.name),
765 3 : NameStr(state->ctx->slot->data.plugin),
766 : state->callback_name);
767 481 : }
768 :
769 : static void
770 1185 : startup_cb_wrapper(LogicalDecodingContext *ctx, OutputPluginOptions *opt, bool is_init)
771 : {
772 : LogicalErrorCallbackState state;
773 : ErrorContextCallback errcallback;
774 :
775 : Assert(!ctx->fast_forward);
776 :
777 : /* Push callback + info on the error context stack */
778 1185 : state.ctx = ctx;
779 1185 : state.callback_name = "startup";
780 1185 : state.report_location = InvalidXLogRecPtr;
781 1185 : errcallback.callback = output_plugin_error_callback;
782 1185 : errcallback.arg = &state;
783 1185 : errcallback.previous = error_context_stack;
784 1185 : error_context_stack = &errcallback;
785 :
786 : /* set output state */
787 1185 : ctx->accept_writes = false;
788 1185 : ctx->end_xact = false;
789 :
790 : /* do the actual work: call callback */
791 1185 : ctx->callbacks.startup_cb(ctx, opt, is_init);
792 :
793 : /* Pop the error context stack */
794 1182 : error_context_stack = errcallback.previous;
795 1182 : }
796 :
797 : static void
798 918 : shutdown_cb_wrapper(LogicalDecodingContext *ctx)
799 : {
800 : LogicalErrorCallbackState state;
801 : ErrorContextCallback errcallback;
802 :
803 : Assert(!ctx->fast_forward);
804 :
805 : /* Push callback + info on the error context stack */
806 918 : state.ctx = ctx;
807 918 : state.callback_name = "shutdown";
808 918 : state.report_location = InvalidXLogRecPtr;
809 918 : errcallback.callback = output_plugin_error_callback;
810 918 : errcallback.arg = &state;
811 918 : errcallback.previous = error_context_stack;
812 918 : error_context_stack = &errcallback;
813 :
814 : /* set output state */
815 918 : ctx->accept_writes = false;
816 918 : ctx->end_xact = false;
817 :
818 : /* do the actual work: call callback */
819 918 : ctx->callbacks.shutdown_cb(ctx);
820 :
821 : /* Pop the error context stack */
822 918 : error_context_stack = errcallback.previous;
823 918 : }
824 :
825 :
826 : /*
827 : * Callbacks for ReorderBuffer which add in some more information and then call
828 : * output_plugin.h plugins.
829 : */
830 : static void
831 1822 : begin_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn)
832 : {
833 1822 : LogicalDecodingContext *ctx = cache->private_data;
834 : LogicalErrorCallbackState state;
835 : ErrorContextCallback errcallback;
836 :
837 : Assert(!ctx->fast_forward);
838 :
839 : /* Push callback + info on the error context stack */
840 1822 : state.ctx = ctx;
841 1822 : state.callback_name = "begin";
842 1822 : state.report_location = txn->first_lsn;
843 1822 : errcallback.callback = output_plugin_error_callback;
844 1822 : errcallback.arg = &state;
845 1822 : errcallback.previous = error_context_stack;
846 1822 : error_context_stack = &errcallback;
847 :
848 : /* set output state */
849 1822 : ctx->accept_writes = true;
850 1822 : ctx->write_xid = txn->xid;
851 1822 : ctx->write_location = txn->first_lsn;
852 1822 : ctx->end_xact = false;
853 :
854 : /* do the actual work: call callback */
855 1822 : ctx->callbacks.begin_cb(ctx, txn);
856 :
857 : /* Pop the error context stack */
858 1822 : error_context_stack = errcallback.previous;
859 1822 : }
860 :
861 : static void
862 1817 : commit_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
863 : XLogRecPtr commit_lsn)
864 : {
865 1817 : LogicalDecodingContext *ctx = cache->private_data;
866 : LogicalErrorCallbackState state;
867 : ErrorContextCallback errcallback;
868 :
869 : Assert(!ctx->fast_forward);
870 :
871 : /* Push callback + info on the error context stack */
872 1817 : state.ctx = ctx;
873 1817 : state.callback_name = "commit";
874 1817 : state.report_location = txn->final_lsn; /* beginning of commit record */
875 1817 : errcallback.callback = output_plugin_error_callback;
876 1817 : errcallback.arg = &state;
877 1817 : errcallback.previous = error_context_stack;
878 1817 : error_context_stack = &errcallback;
879 :
880 : /* set output state */
881 1817 : ctx->accept_writes = true;
882 1817 : ctx->write_xid = txn->xid;
883 1817 : ctx->write_location = txn->end_lsn; /* points to the end of the record */
884 1817 : ctx->end_xact = true;
885 :
886 : /* do the actual work: call callback */
887 1817 : ctx->callbacks.commit_cb(ctx, txn, commit_lsn);
888 :
889 : /* Pop the error context stack */
890 1807 : error_context_stack = errcallback.previous;
891 1807 : }
892 :
893 : /*
894 : * The functionality of begin_prepare is quite similar to begin with the
895 : * exception that this will have gid (global transaction id) information which
896 : * can be used by plugin. Now, we thought about extending the existing begin
897 : * but that would break the replication protocol and additionally this looks
898 : * cleaner.
899 : */
900 : static void
901 30 : begin_prepare_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn)
902 : {
903 30 : LogicalDecodingContext *ctx = cache->private_data;
904 : LogicalErrorCallbackState state;
905 : ErrorContextCallback errcallback;
906 :
907 : Assert(!ctx->fast_forward);
908 :
909 : /* We're only supposed to call this when two-phase commits are supported */
910 : Assert(ctx->twophase);
911 :
912 : /* Push callback + info on the error context stack */
913 30 : state.ctx = ctx;
914 30 : state.callback_name = "begin_prepare";
915 30 : state.report_location = txn->first_lsn;
916 30 : errcallback.callback = output_plugin_error_callback;
917 30 : errcallback.arg = &state;
918 30 : errcallback.previous = error_context_stack;
919 30 : error_context_stack = &errcallback;
920 :
921 : /* set output state */
922 30 : ctx->accept_writes = true;
923 30 : ctx->write_xid = txn->xid;
924 30 : ctx->write_location = txn->first_lsn;
925 30 : ctx->end_xact = false;
926 :
927 : /*
928 : * If the plugin supports two-phase commits then begin prepare callback is
929 : * mandatory
930 : */
931 30 : if (ctx->callbacks.begin_prepare_cb == NULL)
932 0 : ereport(ERROR,
933 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
934 : errmsg("logical replication at prepare time requires a %s callback",
935 : "begin_prepare_cb")));
936 :
937 : /* do the actual work: call callback */
938 30 : ctx->callbacks.begin_prepare_cb(ctx, txn);
939 :
940 : /* Pop the error context stack */
941 30 : error_context_stack = errcallback.previous;
942 30 : }
943 :
944 : static void
945 30 : prepare_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
946 : XLogRecPtr prepare_lsn)
947 : {
948 30 : LogicalDecodingContext *ctx = cache->private_data;
949 : LogicalErrorCallbackState state;
950 : ErrorContextCallback errcallback;
951 :
952 : Assert(!ctx->fast_forward);
953 :
954 : /* We're only supposed to call this when two-phase commits are supported */
955 : Assert(ctx->twophase);
956 :
957 : /* Push callback + info on the error context stack */
958 30 : state.ctx = ctx;
959 30 : state.callback_name = "prepare";
960 30 : state.report_location = txn->final_lsn; /* beginning of prepare record */
961 30 : errcallback.callback = output_plugin_error_callback;
962 30 : errcallback.arg = &state;
963 30 : errcallback.previous = error_context_stack;
964 30 : error_context_stack = &errcallback;
965 :
966 : /* set output state */
967 30 : ctx->accept_writes = true;
968 30 : ctx->write_xid = txn->xid;
969 30 : ctx->write_location = txn->end_lsn; /* points to the end of the record */
970 30 : ctx->end_xact = true;
971 :
972 : /*
973 : * If the plugin supports two-phase commits then prepare callback is
974 : * mandatory
975 : */
976 30 : if (ctx->callbacks.prepare_cb == NULL)
977 0 : ereport(ERROR,
978 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
979 : errmsg("logical replication at prepare time requires a %s callback",
980 : "prepare_cb")));
981 :
982 : /* do the actual work: call callback */
983 30 : ctx->callbacks.prepare_cb(ctx, txn, prepare_lsn);
984 :
985 : /* Pop the error context stack */
986 30 : error_context_stack = errcallback.previous;
987 30 : }
988 :
989 : static void
990 34 : commit_prepared_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
991 : XLogRecPtr commit_lsn)
992 : {
993 34 : LogicalDecodingContext *ctx = cache->private_data;
994 : LogicalErrorCallbackState state;
995 : ErrorContextCallback errcallback;
996 :
997 : Assert(!ctx->fast_forward);
998 :
999 : /* We're only supposed to call this when two-phase commits are supported */
1000 : Assert(ctx->twophase);
1001 :
1002 : /* Push callback + info on the error context stack */
1003 34 : state.ctx = ctx;
1004 34 : state.callback_name = "commit_prepared";
1005 34 : state.report_location = txn->final_lsn; /* beginning of commit record */
1006 34 : errcallback.callback = output_plugin_error_callback;
1007 34 : errcallback.arg = &state;
1008 34 : errcallback.previous = error_context_stack;
1009 34 : error_context_stack = &errcallback;
1010 :
1011 : /* set output state */
1012 34 : ctx->accept_writes = true;
1013 34 : ctx->write_xid = txn->xid;
1014 34 : ctx->write_location = txn->end_lsn; /* points to the end of the record */
1015 34 : ctx->end_xact = true;
1016 :
1017 : /*
1018 : * If the plugin support two-phase commits then commit prepared callback
1019 : * is mandatory
1020 : */
1021 34 : if (ctx->callbacks.commit_prepared_cb == NULL)
1022 0 : ereport(ERROR,
1023 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1024 : errmsg("logical replication at prepare time requires a %s callback",
1025 : "commit_prepared_cb")));
1026 :
1027 : /* do the actual work: call callback */
1028 34 : ctx->callbacks.commit_prepared_cb(ctx, txn, commit_lsn);
1029 :
1030 : /* Pop the error context stack */
1031 34 : error_context_stack = errcallback.previous;
1032 34 : }
1033 :
1034 : static void
1035 11 : rollback_prepared_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1036 : XLogRecPtr prepare_end_lsn,
1037 : TimestampTz prepare_time)
1038 : {
1039 11 : LogicalDecodingContext *ctx = cache->private_data;
1040 : LogicalErrorCallbackState state;
1041 : ErrorContextCallback errcallback;
1042 :
1043 : Assert(!ctx->fast_forward);
1044 :
1045 : /* We're only supposed to call this when two-phase commits are supported */
1046 : Assert(ctx->twophase);
1047 :
1048 : /* Push callback + info on the error context stack */
1049 11 : state.ctx = ctx;
1050 11 : state.callback_name = "rollback_prepared";
1051 11 : state.report_location = txn->final_lsn; /* beginning of commit record */
1052 11 : errcallback.callback = output_plugin_error_callback;
1053 11 : errcallback.arg = &state;
1054 11 : errcallback.previous = error_context_stack;
1055 11 : error_context_stack = &errcallback;
1056 :
1057 : /* set output state */
1058 11 : ctx->accept_writes = true;
1059 11 : ctx->write_xid = txn->xid;
1060 11 : ctx->write_location = txn->end_lsn; /* points to the end of the record */
1061 11 : ctx->end_xact = true;
1062 :
1063 : /*
1064 : * If the plugin support two-phase commits then rollback prepared callback
1065 : * is mandatory
1066 : */
1067 11 : if (ctx->callbacks.rollback_prepared_cb == NULL)
1068 0 : ereport(ERROR,
1069 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1070 : errmsg("logical replication at prepare time requires a %s callback",
1071 : "rollback_prepared_cb")));
1072 :
1073 : /* do the actual work: call callback */
1074 11 : ctx->callbacks.rollback_prepared_cb(ctx, txn, prepare_end_lsn,
1075 : prepare_time);
1076 :
1077 : /* Pop the error context stack */
1078 11 : error_context_stack = errcallback.previous;
1079 11 : }
1080 :
1081 : static void
1082 179101 : change_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1083 : Relation relation, ReorderBufferChange *change)
1084 : {
1085 179101 : LogicalDecodingContext *ctx = cache->private_data;
1086 : LogicalErrorCallbackState state;
1087 : ErrorContextCallback errcallback;
1088 :
1089 : Assert(!ctx->fast_forward);
1090 :
1091 : /* Push callback + info on the error context stack */
1092 179101 : state.ctx = ctx;
1093 179101 : state.callback_name = "change";
1094 179101 : state.report_location = change->lsn;
1095 179101 : errcallback.callback = output_plugin_error_callback;
1096 179101 : errcallback.arg = &state;
1097 179101 : errcallback.previous = error_context_stack;
1098 179101 : error_context_stack = &errcallback;
1099 :
1100 : /* set output state */
1101 179101 : ctx->accept_writes = true;
1102 179101 : ctx->write_xid = txn->xid;
1103 :
1104 : /*
1105 : * Report this change's lsn so replies from clients can give an up-to-date
1106 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1107 : * receipt of this transaction, but it might allow another transaction's
1108 : * commit to be confirmed with one message.
1109 : */
1110 179101 : ctx->write_location = change->lsn;
1111 :
1112 179101 : ctx->end_xact = false;
1113 :
1114 179101 : ctx->callbacks.change_cb(ctx, txn, relation, change);
1115 :
1116 : /* Pop the error context stack */
1117 179096 : error_context_stack = errcallback.previous;
1118 179096 : }
1119 :
1120 : static void
1121 35 : truncate_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1122 : int nrelations, Relation relations[], ReorderBufferChange *change)
1123 : {
1124 35 : LogicalDecodingContext *ctx = cache->private_data;
1125 : LogicalErrorCallbackState state;
1126 : ErrorContextCallback errcallback;
1127 :
1128 : Assert(!ctx->fast_forward);
1129 :
1130 35 : if (!ctx->callbacks.truncate_cb)
1131 0 : return;
1132 :
1133 : /* Push callback + info on the error context stack */
1134 35 : state.ctx = ctx;
1135 35 : state.callback_name = "truncate";
1136 35 : state.report_location = change->lsn;
1137 35 : errcallback.callback = output_plugin_error_callback;
1138 35 : errcallback.arg = &state;
1139 35 : errcallback.previous = error_context_stack;
1140 35 : error_context_stack = &errcallback;
1141 :
1142 : /* set output state */
1143 35 : ctx->accept_writes = true;
1144 35 : ctx->write_xid = txn->xid;
1145 :
1146 : /*
1147 : * Report this change's lsn so replies from clients can give an up-to-date
1148 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1149 : * receipt of this transaction, but it might allow another transaction's
1150 : * commit to be confirmed with one message.
1151 : */
1152 35 : ctx->write_location = change->lsn;
1153 :
1154 35 : ctx->end_xact = false;
1155 :
1156 35 : ctx->callbacks.truncate_cb(ctx, txn, nrelations, relations, change);
1157 :
1158 : /* Pop the error context stack */
1159 35 : error_context_stack = errcallback.previous;
1160 : }
1161 :
1162 : bool
1163 148 : filter_prepare_cb_wrapper(LogicalDecodingContext *ctx, TransactionId xid,
1164 : const char *gid)
1165 : {
1166 : LogicalErrorCallbackState state;
1167 : ErrorContextCallback errcallback;
1168 : bool ret;
1169 :
1170 : Assert(!ctx->fast_forward);
1171 :
1172 : /* Push callback + info on the error context stack */
1173 148 : state.ctx = ctx;
1174 148 : state.callback_name = "filter_prepare";
1175 148 : state.report_location = InvalidXLogRecPtr;
1176 148 : errcallback.callback = output_plugin_error_callback;
1177 148 : errcallback.arg = &state;
1178 148 : errcallback.previous = error_context_stack;
1179 148 : error_context_stack = &errcallback;
1180 :
1181 : /* set output state */
1182 148 : ctx->accept_writes = false;
1183 148 : ctx->end_xact = false;
1184 :
1185 : /* do the actual work: call callback */
1186 148 : ret = ctx->callbacks.filter_prepare_cb(ctx, xid, gid);
1187 :
1188 : /* Pop the error context stack */
1189 148 : error_context_stack = errcallback.previous;
1190 :
1191 148 : return ret;
1192 : }
1193 :
1194 : bool
1195 1307402 : filter_by_origin_cb_wrapper(LogicalDecodingContext *ctx, ReplOriginId origin_id)
1196 : {
1197 : LogicalErrorCallbackState state;
1198 : ErrorContextCallback errcallback;
1199 : bool ret;
1200 :
1201 : Assert(!ctx->fast_forward);
1202 :
1203 : /* Push callback + info on the error context stack */
1204 1307402 : state.ctx = ctx;
1205 1307402 : state.callback_name = "filter_by_origin";
1206 1307402 : state.report_location = InvalidXLogRecPtr;
1207 1307402 : errcallback.callback = output_plugin_error_callback;
1208 1307402 : errcallback.arg = &state;
1209 1307402 : errcallback.previous = error_context_stack;
1210 1307402 : error_context_stack = &errcallback;
1211 :
1212 : /* set output state */
1213 1307402 : ctx->accept_writes = false;
1214 1307402 : ctx->end_xact = false;
1215 :
1216 : /* do the actual work: call callback */
1217 1307402 : ret = ctx->callbacks.filter_by_origin_cb(ctx, origin_id);
1218 :
1219 : /* Pop the error context stack */
1220 1307402 : error_context_stack = errcallback.previous;
1221 :
1222 1307402 : return ret;
1223 : }
1224 :
1225 : static void
1226 16 : message_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1227 : XLogRecPtr message_lsn, bool transactional,
1228 : const char *prefix, Size message_size, const char *message)
1229 : {
1230 16 : LogicalDecodingContext *ctx = cache->private_data;
1231 : LogicalErrorCallbackState state;
1232 : ErrorContextCallback errcallback;
1233 :
1234 : Assert(!ctx->fast_forward);
1235 :
1236 16 : if (ctx->callbacks.message_cb == NULL)
1237 0 : return;
1238 :
1239 : /* Push callback + info on the error context stack */
1240 16 : state.ctx = ctx;
1241 16 : state.callback_name = "message";
1242 16 : state.report_location = message_lsn;
1243 16 : errcallback.callback = output_plugin_error_callback;
1244 16 : errcallback.arg = &state;
1245 16 : errcallback.previous = error_context_stack;
1246 16 : error_context_stack = &errcallback;
1247 :
1248 : /* set output state */
1249 16 : ctx->accept_writes = true;
1250 16 : ctx->write_xid = txn != NULL ? txn->xid : InvalidTransactionId;
1251 16 : ctx->write_location = message_lsn;
1252 16 : ctx->end_xact = false;
1253 :
1254 : /* do the actual work: call callback */
1255 16 : ctx->callbacks.message_cb(ctx, txn, message_lsn, transactional, prefix,
1256 : message_size, message);
1257 :
1258 : /* Pop the error context stack */
1259 16 : error_context_stack = errcallback.previous;
1260 : }
1261 :
1262 : static void
1263 680 : stream_start_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1264 : XLogRecPtr first_lsn)
1265 : {
1266 680 : LogicalDecodingContext *ctx = cache->private_data;
1267 : LogicalErrorCallbackState state;
1268 : ErrorContextCallback errcallback;
1269 :
1270 : Assert(!ctx->fast_forward);
1271 :
1272 : /* We're only supposed to call this when streaming is supported. */
1273 : Assert(ctx->streaming);
1274 :
1275 : /* Push callback + info on the error context stack */
1276 680 : state.ctx = ctx;
1277 680 : state.callback_name = "stream_start";
1278 680 : state.report_location = first_lsn;
1279 680 : errcallback.callback = output_plugin_error_callback;
1280 680 : errcallback.arg = &state;
1281 680 : errcallback.previous = error_context_stack;
1282 680 : error_context_stack = &errcallback;
1283 :
1284 : /* set output state */
1285 680 : ctx->accept_writes = true;
1286 680 : ctx->write_xid = txn->xid;
1287 :
1288 : /*
1289 : * Report this message's lsn so replies from clients can give an
1290 : * up-to-date answer. This won't ever be enough (and shouldn't be!) to
1291 : * confirm receipt of this transaction, but it might allow another
1292 : * transaction's commit to be confirmed with one message.
1293 : */
1294 680 : ctx->write_location = first_lsn;
1295 :
1296 680 : ctx->end_xact = false;
1297 :
1298 : /* in streaming mode, stream_start_cb is required */
1299 680 : if (ctx->callbacks.stream_start_cb == NULL)
1300 0 : ereport(ERROR,
1301 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1302 : errmsg("logical streaming requires a %s callback",
1303 : "stream_start_cb")));
1304 :
1305 680 : ctx->callbacks.stream_start_cb(ctx, txn);
1306 :
1307 : /* Pop the error context stack */
1308 680 : error_context_stack = errcallback.previous;
1309 680 : }
1310 :
1311 : static void
1312 680 : stream_stop_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1313 : XLogRecPtr last_lsn)
1314 : {
1315 680 : LogicalDecodingContext *ctx = cache->private_data;
1316 : LogicalErrorCallbackState state;
1317 : ErrorContextCallback errcallback;
1318 :
1319 : Assert(!ctx->fast_forward);
1320 :
1321 : /* We're only supposed to call this when streaming is supported. */
1322 : Assert(ctx->streaming);
1323 :
1324 : /* Push callback + info on the error context stack */
1325 680 : state.ctx = ctx;
1326 680 : state.callback_name = "stream_stop";
1327 680 : state.report_location = last_lsn;
1328 680 : errcallback.callback = output_plugin_error_callback;
1329 680 : errcallback.arg = &state;
1330 680 : errcallback.previous = error_context_stack;
1331 680 : error_context_stack = &errcallback;
1332 :
1333 : /* set output state */
1334 680 : ctx->accept_writes = true;
1335 680 : ctx->write_xid = txn->xid;
1336 :
1337 : /*
1338 : * Report this message's lsn so replies from clients can give an
1339 : * up-to-date answer. This won't ever be enough (and shouldn't be!) to
1340 : * confirm receipt of this transaction, but it might allow another
1341 : * transaction's commit to be confirmed with one message.
1342 : */
1343 680 : ctx->write_location = last_lsn;
1344 :
1345 680 : ctx->end_xact = false;
1346 :
1347 : /* in streaming mode, stream_stop_cb is required */
1348 680 : if (ctx->callbacks.stream_stop_cb == NULL)
1349 0 : ereport(ERROR,
1350 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1351 : errmsg("logical streaming requires a %s callback",
1352 : "stream_stop_cb")));
1353 :
1354 680 : ctx->callbacks.stream_stop_cb(ctx, txn);
1355 :
1356 : /* Pop the error context stack */
1357 680 : error_context_stack = errcallback.previous;
1358 680 : }
1359 :
1360 : static void
1361 30 : stream_abort_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1362 : XLogRecPtr abort_lsn)
1363 : {
1364 30 : LogicalDecodingContext *ctx = cache->private_data;
1365 : LogicalErrorCallbackState state;
1366 : ErrorContextCallback errcallback;
1367 :
1368 : Assert(!ctx->fast_forward);
1369 :
1370 : /* We're only supposed to call this when streaming is supported. */
1371 : Assert(ctx->streaming);
1372 :
1373 : /* Push callback + info on the error context stack */
1374 30 : state.ctx = ctx;
1375 30 : state.callback_name = "stream_abort";
1376 30 : state.report_location = abort_lsn;
1377 30 : errcallback.callback = output_plugin_error_callback;
1378 30 : errcallback.arg = &state;
1379 30 : errcallback.previous = error_context_stack;
1380 30 : error_context_stack = &errcallback;
1381 :
1382 : /* set output state */
1383 30 : ctx->accept_writes = true;
1384 30 : ctx->write_xid = txn->xid;
1385 30 : ctx->write_location = abort_lsn;
1386 30 : ctx->end_xact = true;
1387 :
1388 : /* in streaming mode, stream_abort_cb is required */
1389 30 : if (ctx->callbacks.stream_abort_cb == NULL)
1390 0 : ereport(ERROR,
1391 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1392 : errmsg("logical streaming requires a %s callback",
1393 : "stream_abort_cb")));
1394 :
1395 30 : ctx->callbacks.stream_abort_cb(ctx, txn, abort_lsn);
1396 :
1397 : /* Pop the error context stack */
1398 30 : error_context_stack = errcallback.previous;
1399 30 : }
1400 :
1401 : static void
1402 17 : stream_prepare_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1403 : XLogRecPtr prepare_lsn)
1404 : {
1405 17 : LogicalDecodingContext *ctx = cache->private_data;
1406 : LogicalErrorCallbackState state;
1407 : ErrorContextCallback errcallback;
1408 :
1409 : Assert(!ctx->fast_forward);
1410 :
1411 : /*
1412 : * We're only supposed to call this when streaming and two-phase commits
1413 : * are supported.
1414 : */
1415 : Assert(ctx->streaming);
1416 : Assert(ctx->twophase);
1417 :
1418 : /* Push callback + info on the error context stack */
1419 17 : state.ctx = ctx;
1420 17 : state.callback_name = "stream_prepare";
1421 17 : state.report_location = txn->final_lsn;
1422 17 : errcallback.callback = output_plugin_error_callback;
1423 17 : errcallback.arg = &state;
1424 17 : errcallback.previous = error_context_stack;
1425 17 : error_context_stack = &errcallback;
1426 :
1427 : /* set output state */
1428 17 : ctx->accept_writes = true;
1429 17 : ctx->write_xid = txn->xid;
1430 17 : ctx->write_location = txn->end_lsn;
1431 17 : ctx->end_xact = true;
1432 :
1433 : /* in streaming mode with two-phase commits, stream_prepare_cb is required */
1434 17 : if (ctx->callbacks.stream_prepare_cb == NULL)
1435 0 : ereport(ERROR,
1436 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1437 : errmsg("logical streaming at prepare time requires a %s callback",
1438 : "stream_prepare_cb")));
1439 :
1440 17 : ctx->callbacks.stream_prepare_cb(ctx, txn, prepare_lsn);
1441 :
1442 : /* Pop the error context stack */
1443 17 : error_context_stack = errcallback.previous;
1444 17 : }
1445 :
1446 : static void
1447 51 : stream_commit_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1448 : XLogRecPtr commit_lsn)
1449 : {
1450 51 : LogicalDecodingContext *ctx = cache->private_data;
1451 : LogicalErrorCallbackState state;
1452 : ErrorContextCallback errcallback;
1453 :
1454 : Assert(!ctx->fast_forward);
1455 :
1456 : /* We're only supposed to call this when streaming is supported. */
1457 : Assert(ctx->streaming);
1458 :
1459 : /* Push callback + info on the error context stack */
1460 51 : state.ctx = ctx;
1461 51 : state.callback_name = "stream_commit";
1462 51 : state.report_location = txn->final_lsn;
1463 51 : errcallback.callback = output_plugin_error_callback;
1464 51 : errcallback.arg = &state;
1465 51 : errcallback.previous = error_context_stack;
1466 51 : error_context_stack = &errcallback;
1467 :
1468 : /* set output state */
1469 51 : ctx->accept_writes = true;
1470 51 : ctx->write_xid = txn->xid;
1471 51 : ctx->write_location = txn->end_lsn;
1472 51 : ctx->end_xact = true;
1473 :
1474 : /* in streaming mode, stream_commit_cb is required */
1475 51 : if (ctx->callbacks.stream_commit_cb == NULL)
1476 0 : ereport(ERROR,
1477 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1478 : errmsg("logical streaming requires a %s callback",
1479 : "stream_commit_cb")));
1480 :
1481 51 : ctx->callbacks.stream_commit_cb(ctx, txn, commit_lsn);
1482 :
1483 : /* Pop the error context stack */
1484 51 : error_context_stack = errcallback.previous;
1485 51 : }
1486 :
1487 : static void
1488 180978 : stream_change_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1489 : Relation relation, ReorderBufferChange *change)
1490 : {
1491 180978 : LogicalDecodingContext *ctx = cache->private_data;
1492 : LogicalErrorCallbackState state;
1493 : ErrorContextCallback errcallback;
1494 :
1495 : Assert(!ctx->fast_forward);
1496 :
1497 : /* We're only supposed to call this when streaming is supported. */
1498 : Assert(ctx->streaming);
1499 :
1500 : /* Push callback + info on the error context stack */
1501 180978 : state.ctx = ctx;
1502 180978 : state.callback_name = "stream_change";
1503 180978 : state.report_location = change->lsn;
1504 180978 : errcallback.callback = output_plugin_error_callback;
1505 180978 : errcallback.arg = &state;
1506 180978 : errcallback.previous = error_context_stack;
1507 180978 : error_context_stack = &errcallback;
1508 :
1509 : /* set output state */
1510 180978 : ctx->accept_writes = true;
1511 180978 : ctx->write_xid = txn->xid;
1512 :
1513 : /*
1514 : * Report this change's lsn so replies from clients can give an up-to-date
1515 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1516 : * receipt of this transaction, but it might allow another transaction's
1517 : * commit to be confirmed with one message.
1518 : */
1519 180978 : ctx->write_location = change->lsn;
1520 :
1521 180978 : ctx->end_xact = false;
1522 :
1523 : /* in streaming mode, stream_change_cb is required */
1524 180978 : if (ctx->callbacks.stream_change_cb == NULL)
1525 0 : ereport(ERROR,
1526 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1527 : errmsg("logical streaming requires a %s callback",
1528 : "stream_change_cb")));
1529 :
1530 180978 : ctx->callbacks.stream_change_cb(ctx, txn, relation, change);
1531 :
1532 : /* Pop the error context stack */
1533 180978 : error_context_stack = errcallback.previous;
1534 180978 : }
1535 :
1536 : static void
1537 3 : stream_message_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1538 : XLogRecPtr message_lsn, bool transactional,
1539 : const char *prefix, Size message_size, const char *message)
1540 : {
1541 3 : LogicalDecodingContext *ctx = cache->private_data;
1542 : LogicalErrorCallbackState state;
1543 : ErrorContextCallback errcallback;
1544 :
1545 : Assert(!ctx->fast_forward);
1546 :
1547 : /* We're only supposed to call this when streaming is supported. */
1548 : Assert(ctx->streaming);
1549 :
1550 : /* this callback is optional */
1551 3 : if (ctx->callbacks.stream_message_cb == NULL)
1552 0 : return;
1553 :
1554 : /* Push callback + info on the error context stack */
1555 3 : state.ctx = ctx;
1556 3 : state.callback_name = "stream_message";
1557 3 : state.report_location = message_lsn;
1558 3 : errcallback.callback = output_plugin_error_callback;
1559 3 : errcallback.arg = &state;
1560 3 : errcallback.previous = error_context_stack;
1561 3 : error_context_stack = &errcallback;
1562 :
1563 : /* set output state */
1564 3 : ctx->accept_writes = true;
1565 3 : ctx->write_xid = txn != NULL ? txn->xid : InvalidTransactionId;
1566 3 : ctx->write_location = message_lsn;
1567 3 : ctx->end_xact = false;
1568 :
1569 : /* do the actual work: call callback */
1570 3 : ctx->callbacks.stream_message_cb(ctx, txn, message_lsn, transactional, prefix,
1571 : message_size, message);
1572 :
1573 : /* Pop the error context stack */
1574 3 : error_context_stack = errcallback.previous;
1575 : }
1576 :
1577 : static void
1578 0 : stream_truncate_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1579 : int nrelations, Relation relations[],
1580 : ReorderBufferChange *change)
1581 : {
1582 0 : LogicalDecodingContext *ctx = cache->private_data;
1583 : LogicalErrorCallbackState state;
1584 : ErrorContextCallback errcallback;
1585 :
1586 : Assert(!ctx->fast_forward);
1587 :
1588 : /* We're only supposed to call this when streaming is supported. */
1589 : Assert(ctx->streaming);
1590 :
1591 : /* this callback is optional */
1592 0 : if (!ctx->callbacks.stream_truncate_cb)
1593 0 : return;
1594 :
1595 : /* Push callback + info on the error context stack */
1596 0 : state.ctx = ctx;
1597 0 : state.callback_name = "stream_truncate";
1598 0 : state.report_location = change->lsn;
1599 0 : errcallback.callback = output_plugin_error_callback;
1600 0 : errcallback.arg = &state;
1601 0 : errcallback.previous = error_context_stack;
1602 0 : error_context_stack = &errcallback;
1603 :
1604 : /* set output state */
1605 0 : ctx->accept_writes = true;
1606 0 : ctx->write_xid = txn->xid;
1607 :
1608 : /*
1609 : * Report this change's lsn so replies from clients can give an up-to-date
1610 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1611 : * receipt of this transaction, but it might allow another transaction's
1612 : * commit to be confirmed with one message.
1613 : */
1614 0 : ctx->write_location = change->lsn;
1615 :
1616 0 : ctx->end_xact = false;
1617 :
1618 0 : ctx->callbacks.stream_truncate_cb(ctx, txn, nrelations, relations, change);
1619 :
1620 : /* Pop the error context stack */
1621 0 : error_context_stack = errcallback.previous;
1622 : }
1623 :
1624 : static void
1625 3362 : update_progress_txn_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1626 : XLogRecPtr lsn)
1627 : {
1628 3362 : LogicalDecodingContext *ctx = cache->private_data;
1629 : LogicalErrorCallbackState state;
1630 : ErrorContextCallback errcallback;
1631 :
1632 : Assert(!ctx->fast_forward);
1633 :
1634 : /* Push callback + info on the error context stack */
1635 3362 : state.ctx = ctx;
1636 3362 : state.callback_name = "update_progress_txn";
1637 3362 : state.report_location = lsn;
1638 3362 : errcallback.callback = output_plugin_error_callback;
1639 3362 : errcallback.arg = &state;
1640 3362 : errcallback.previous = error_context_stack;
1641 3362 : error_context_stack = &errcallback;
1642 :
1643 : /* set output state */
1644 3362 : ctx->accept_writes = false;
1645 3362 : ctx->write_xid = txn->xid;
1646 :
1647 : /*
1648 : * Report this change's lsn so replies from clients can give an up-to-date
1649 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1650 : * receipt of this transaction, but it might allow another transaction's
1651 : * commit to be confirmed with one message.
1652 : */
1653 3362 : ctx->write_location = lsn;
1654 :
1655 3362 : ctx->end_xact = false;
1656 :
1657 3362 : OutputPluginUpdateProgress(ctx, false);
1658 :
1659 : /* Pop the error context stack */
1660 3362 : error_context_stack = errcallback.previous;
1661 3362 : }
1662 :
1663 : /*
1664 : * Set the required catalog xmin horizon for historic snapshots in the current
1665 : * replication slot.
1666 : *
1667 : * Note that in the most cases, we won't be able to immediately use the xmin
1668 : * to increase the xmin horizon: we need to wait till the client has confirmed
1669 : * receiving current_lsn with LogicalConfirmReceivedLocation().
1670 : */
1671 : void
1672 541 : LogicalIncreaseXminForSlot(XLogRecPtr current_lsn, TransactionId xmin)
1673 : {
1674 541 : bool updated_xmin = false;
1675 : ReplicationSlot *slot;
1676 541 : bool got_new_xmin = false;
1677 :
1678 541 : slot = MyReplicationSlot;
1679 :
1680 : Assert(slot != NULL);
1681 :
1682 541 : SpinLockAcquire(&slot->mutex);
1683 :
1684 : /*
1685 : * don't overwrite if we already have a newer xmin. This can happen if we
1686 : * restart decoding in a slot.
1687 : */
1688 541 : if (TransactionIdPrecedesOrEquals(xmin, slot->data.catalog_xmin))
1689 : {
1690 : }
1691 :
1692 : /*
1693 : * If the client has already confirmed up to this lsn, we directly can
1694 : * mark this as accepted. This can happen if we restart decoding in a
1695 : * slot.
1696 : */
1697 193 : else if (current_lsn <= slot->data.confirmed_flush)
1698 : {
1699 67 : slot->candidate_catalog_xmin = xmin;
1700 67 : slot->candidate_xmin_lsn = current_lsn;
1701 :
1702 : /* our candidate can directly be used */
1703 67 : updated_xmin = true;
1704 : }
1705 :
1706 : /*
1707 : * Only increase if the previous values have been applied, otherwise we
1708 : * might never end up updating if the receiver acks too slowly.
1709 : */
1710 126 : else if (!XLogRecPtrIsValid(slot->candidate_xmin_lsn))
1711 : {
1712 23 : slot->candidate_catalog_xmin = xmin;
1713 23 : slot->candidate_xmin_lsn = current_lsn;
1714 :
1715 : /*
1716 : * Log new xmin at an appropriate log level after releasing the
1717 : * spinlock.
1718 : */
1719 23 : got_new_xmin = true;
1720 : }
1721 541 : SpinLockRelease(&slot->mutex);
1722 :
1723 541 : if (got_new_xmin)
1724 23 : elog(DEBUG1, "got new catalog xmin %u at %X/%08X", xmin,
1725 : LSN_FORMAT_ARGS(current_lsn));
1726 :
1727 : /* candidate already valid with the current flush position, apply */
1728 541 : if (updated_xmin)
1729 67 : LogicalConfirmReceivedLocation(slot->data.confirmed_flush);
1730 541 : }
1731 :
1732 : /*
1733 : * Mark the minimal LSN (restart_lsn) we need to read to replay all
1734 : * transactions that have not yet committed at current_lsn.
1735 : *
1736 : * Just like LogicalIncreaseXminForSlot this only takes effect when the
1737 : * client has confirmed to have received current_lsn.
1738 : */
1739 : void
1740 490 : LogicalIncreaseRestartDecodingForSlot(XLogRecPtr current_lsn, XLogRecPtr restart_lsn)
1741 : {
1742 490 : bool updated_lsn = false;
1743 : ReplicationSlot *slot;
1744 :
1745 490 : slot = MyReplicationSlot;
1746 :
1747 : Assert(slot != NULL);
1748 : Assert(XLogRecPtrIsValid(restart_lsn));
1749 : Assert(XLogRecPtrIsValid(current_lsn));
1750 :
1751 490 : SpinLockAcquire(&slot->mutex);
1752 :
1753 : /* don't overwrite if have a newer restart lsn */
1754 490 : if (restart_lsn <= slot->data.restart_lsn)
1755 : {
1756 10 : SpinLockRelease(&slot->mutex);
1757 : }
1758 :
1759 : /*
1760 : * We might have already flushed far enough to directly accept this lsn,
1761 : * in this case there is no need to check for existing candidate LSNs
1762 : */
1763 480 : else if (current_lsn <= slot->data.confirmed_flush)
1764 : {
1765 324 : slot->candidate_restart_valid = current_lsn;
1766 324 : slot->candidate_restart_lsn = restart_lsn;
1767 324 : SpinLockRelease(&slot->mutex);
1768 :
1769 : /* our candidate can directly be used */
1770 324 : updated_lsn = true;
1771 : }
1772 :
1773 : /*
1774 : * Only increase if the previous values have been applied, otherwise we
1775 : * might never end up updating if the receiver acks too slowly. A missed
1776 : * value here will just cause some extra effort after reconnecting.
1777 : */
1778 156 : else if (!XLogRecPtrIsValid(slot->candidate_restart_valid))
1779 : {
1780 44 : slot->candidate_restart_valid = current_lsn;
1781 44 : slot->candidate_restart_lsn = restart_lsn;
1782 44 : SpinLockRelease(&slot->mutex);
1783 :
1784 44 : elog(DEBUG1, "got new restart lsn %X/%08X at %X/%08X",
1785 : LSN_FORMAT_ARGS(restart_lsn),
1786 : LSN_FORMAT_ARGS(current_lsn));
1787 : }
1788 : else
1789 : {
1790 : XLogRecPtr candidate_restart_lsn;
1791 : XLogRecPtr candidate_restart_valid;
1792 : XLogRecPtr confirmed_flush;
1793 :
1794 112 : candidate_restart_lsn = slot->candidate_restart_lsn;
1795 112 : candidate_restart_valid = slot->candidate_restart_valid;
1796 112 : confirmed_flush = slot->data.confirmed_flush;
1797 112 : SpinLockRelease(&slot->mutex);
1798 :
1799 112 : elog(DEBUG1, "failed to increase restart lsn: proposed %X/%08X, after %X/%08X, current candidate %X/%08X, current after %X/%08X, flushed up to %X/%08X",
1800 : LSN_FORMAT_ARGS(restart_lsn),
1801 : LSN_FORMAT_ARGS(current_lsn),
1802 : LSN_FORMAT_ARGS(candidate_restart_lsn),
1803 : LSN_FORMAT_ARGS(candidate_restart_valid),
1804 : LSN_FORMAT_ARGS(confirmed_flush));
1805 : }
1806 :
1807 : /* candidates are already valid with the current flush position, apply */
1808 490 : if (updated_lsn)
1809 324 : LogicalConfirmReceivedLocation(slot->data.confirmed_flush);
1810 490 : }
1811 :
1812 : /*
1813 : * Handle a consumer's confirmation having received all changes up to lsn.
1814 : */
1815 : void
1816 19539 : LogicalConfirmReceivedLocation(XLogRecPtr lsn)
1817 : {
1818 : Assert(XLogRecPtrIsValid(lsn));
1819 :
1820 : /* Do an unlocked check for candidate_lsn first. */
1821 19539 : if (XLogRecPtrIsValid(MyReplicationSlot->candidate_xmin_lsn) ||
1822 19418 : XLogRecPtrIsValid(MyReplicationSlot->candidate_restart_valid))
1823 467 : {
1824 467 : bool updated_xmin = false;
1825 467 : bool updated_restart = false;
1826 : XLogRecPtr restart_lsn pg_attribute_unused();
1827 :
1828 467 : SpinLockAcquire(&MyReplicationSlot->mutex);
1829 :
1830 : /* remember the old restart lsn */
1831 467 : restart_lsn = MyReplicationSlot->data.restart_lsn;
1832 :
1833 : /*
1834 : * Prevent moving the confirmed_flush backwards, as this could lead to
1835 : * data duplication issues caused by replicating already replicated
1836 : * changes.
1837 : *
1838 : * This can happen when a client acknowledges an LSN it doesn't have
1839 : * to do anything for, and thus didn't store persistently. After a
1840 : * restart, the client can send the prior LSN that it stored
1841 : * persistently as an acknowledgement, but we need to ignore such an
1842 : * LSN. See similar case handling in CreateDecodingContext.
1843 : */
1844 467 : if (lsn > MyReplicationSlot->data.confirmed_flush)
1845 39 : MyReplicationSlot->data.confirmed_flush = lsn;
1846 :
1847 : /* if we're past the location required for bumping xmin, do so */
1848 467 : if (XLogRecPtrIsValid(MyReplicationSlot->candidate_xmin_lsn) &&
1849 121 : MyReplicationSlot->candidate_xmin_lsn <= lsn)
1850 : {
1851 : /*
1852 : * We have to write the changed xmin to disk *before* we change
1853 : * the in-memory value, otherwise after a crash we wouldn't know
1854 : * that some catalog tuples might have been removed already.
1855 : *
1856 : * Ensure that by first writing to ->xmin and only update
1857 : * ->effective_xmin once the new state is synced to disk. After a
1858 : * crash ->effective_xmin is set to ->xmin.
1859 : */
1860 84 : if (TransactionIdIsValid(MyReplicationSlot->candidate_catalog_xmin) &&
1861 84 : MyReplicationSlot->data.catalog_xmin != MyReplicationSlot->candidate_catalog_xmin)
1862 : {
1863 84 : MyReplicationSlot->data.catalog_xmin = MyReplicationSlot->candidate_catalog_xmin;
1864 84 : MyReplicationSlot->candidate_catalog_xmin = InvalidTransactionId;
1865 84 : MyReplicationSlot->candidate_xmin_lsn = InvalidXLogRecPtr;
1866 84 : updated_xmin = true;
1867 : }
1868 : }
1869 :
1870 467 : if (XLogRecPtrIsValid(MyReplicationSlot->candidate_restart_valid) &&
1871 400 : MyReplicationSlot->candidate_restart_valid <= lsn)
1872 : {
1873 : Assert(XLogRecPtrIsValid(MyReplicationSlot->candidate_restart_lsn));
1874 :
1875 362 : MyReplicationSlot->data.restart_lsn = MyReplicationSlot->candidate_restart_lsn;
1876 362 : MyReplicationSlot->candidate_restart_lsn = InvalidXLogRecPtr;
1877 362 : MyReplicationSlot->candidate_restart_valid = InvalidXLogRecPtr;
1878 362 : updated_restart = true;
1879 : }
1880 :
1881 467 : SpinLockRelease(&MyReplicationSlot->mutex);
1882 :
1883 : /* first write new xmin to disk, so we know what's up after a crash */
1884 467 : if (updated_xmin || updated_restart)
1885 : {
1886 : #ifdef USE_INJECTION_POINTS
1887 : XLogSegNo seg1,
1888 : seg2;
1889 :
1890 429 : XLByteToSeg(restart_lsn, seg1, wal_segment_size);
1891 429 : XLByteToSeg(MyReplicationSlot->data.restart_lsn, seg2, wal_segment_size);
1892 :
1893 : /* trigger injection point, but only if segment changes */
1894 429 : if (seg1 != seg2)
1895 18 : INJECTION_POINT("logical-replication-slot-advance-segment", NULL);
1896 : #endif
1897 :
1898 429 : ReplicationSlotMarkDirty();
1899 429 : ReplicationSlotSave();
1900 429 : elog(DEBUG1, "updated xmin: %u restart: %u", updated_xmin, updated_restart);
1901 : }
1902 :
1903 : /*
1904 : * Now the new xmin is safely on disk, we can let the global value
1905 : * advance. We do not take ProcArrayLock or similar since we only
1906 : * advance xmin here and there's not much harm done by a concurrent
1907 : * computation missing that.
1908 : */
1909 467 : if (updated_xmin)
1910 : {
1911 84 : SpinLockAcquire(&MyReplicationSlot->mutex);
1912 84 : MyReplicationSlot->effective_catalog_xmin = MyReplicationSlot->data.catalog_xmin;
1913 84 : SpinLockRelease(&MyReplicationSlot->mutex);
1914 :
1915 84 : ReplicationSlotsComputeRequiredXmin(false);
1916 84 : ReplicationSlotsComputeRequiredLSN();
1917 : }
1918 : }
1919 : else
1920 : {
1921 19072 : SpinLockAcquire(&MyReplicationSlot->mutex);
1922 :
1923 : /*
1924 : * Prevent moving the confirmed_flush backwards. See comments above
1925 : * for the details.
1926 : */
1927 19072 : if (lsn > MyReplicationSlot->data.confirmed_flush)
1928 17270 : MyReplicationSlot->data.confirmed_flush = lsn;
1929 :
1930 19072 : SpinLockRelease(&MyReplicationSlot->mutex);
1931 : }
1932 19539 : }
1933 :
1934 : /*
1935 : * Clear logical streaming state during (sub)transaction abort.
1936 : */
1937 : void
1938 40836 : ResetLogicalStreamingState(void)
1939 : {
1940 40836 : CheckXidAlive = InvalidTransactionId;
1941 40836 : bsysscan = false;
1942 40836 : }
1943 :
1944 : /*
1945 : * Report stats for a slot.
1946 : */
1947 : void
1948 6327 : UpdateDecodingStats(LogicalDecodingContext *ctx)
1949 : {
1950 6327 : ReorderBuffer *rb = ctx->reorder;
1951 : PgStat_StatReplSlotEntry repSlotStat;
1952 :
1953 : /* Nothing to do if we don't have any replication stats to be sent. */
1954 6327 : if (rb->spillBytes <= 0 && rb->streamBytes <= 0 && rb->totalBytes <= 0 &&
1955 465 : rb->memExceededCount <= 0)
1956 456 : return;
1957 :
1958 5871 : elog(DEBUG2, "UpdateDecodingStats: updating stats %p %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64,
1959 : rb,
1960 : rb->spillTxns,
1961 : rb->spillCount,
1962 : rb->spillBytes,
1963 : rb->streamTxns,
1964 : rb->streamCount,
1965 : rb->streamBytes,
1966 : rb->memExceededCount,
1967 : rb->totalTxns,
1968 : rb->totalBytes);
1969 :
1970 5871 : repSlotStat.spill_txns = rb->spillTxns;
1971 5871 : repSlotStat.spill_count = rb->spillCount;
1972 5871 : repSlotStat.spill_bytes = rb->spillBytes;
1973 5871 : repSlotStat.stream_txns = rb->streamTxns;
1974 5871 : repSlotStat.stream_count = rb->streamCount;
1975 5871 : repSlotStat.stream_bytes = rb->streamBytes;
1976 5871 : repSlotStat.mem_exceeded_count = rb->memExceededCount;
1977 5871 : repSlotStat.total_txns = rb->totalTxns;
1978 5871 : repSlotStat.total_bytes = rb->totalBytes;
1979 :
1980 5871 : pgstat_report_replslot(ctx->slot, &repSlotStat);
1981 :
1982 5871 : rb->spillTxns = 0;
1983 5871 : rb->spillCount = 0;
1984 5871 : rb->spillBytes = 0;
1985 5871 : rb->streamTxns = 0;
1986 5871 : rb->streamCount = 0;
1987 5871 : rb->streamBytes = 0;
1988 5871 : rb->memExceededCount = 0;
1989 5871 : rb->totalTxns = 0;
1990 5871 : rb->totalBytes = 0;
1991 : }
1992 :
1993 : /*
1994 : * Read up to the end of WAL starting from the decoding slot's restart_lsn
1995 : * to end_of_wal in order to check if any meaningful/decodable WAL records
1996 : * are encountered. scan_cutoff_lsn is the LSN, where we can terminate the
1997 : * WAL scan early if we find a decodable WAL record after this LSN.
1998 : *
1999 : * Returns the last LSN decodable WAL record's LSN if found, otherwise
2000 : * returns InvalidXLogRecPtr.
2001 : */
2002 : XLogRecPtr
2003 3 : LogicalReplicationSlotCheckPendingWal(XLogRecPtr end_of_wal,
2004 : XLogRecPtr scan_cutoff_lsn)
2005 : {
2006 3 : XLogRecPtr last_pending_wal = InvalidXLogRecPtr;
2007 :
2008 : Assert(MyReplicationSlot);
2009 : Assert(end_of_wal >= scan_cutoff_lsn);
2010 :
2011 3 : PG_TRY();
2012 : {
2013 : LogicalDecodingContext *ctx;
2014 :
2015 : /*
2016 : * Create our decoding context in fast_forward mode, passing start_lsn
2017 : * as InvalidXLogRecPtr, so that we start processing from the slot's
2018 : * confirmed_flush.
2019 : */
2020 6 : ctx = CreateDecodingContext(InvalidXLogRecPtr,
2021 : NIL,
2022 : true, /* fast_forward */
2023 3 : XL_ROUTINE(.page_read = read_local_xlog_page,
2024 : .segment_open = wal_segment_open,
2025 : .segment_close = wal_segment_close),
2026 : NULL, NULL, NULL);
2027 :
2028 : /*
2029 : * Start reading at the slot's restart_lsn, which we know points to a
2030 : * valid record.
2031 : */
2032 3 : XLogBeginRead(ctx->reader, MyReplicationSlot->data.restart_lsn);
2033 :
2034 : /* Invalidate non-timetravel entries */
2035 3 : InvalidateSystemCaches();
2036 :
2037 81 : while (ctx->reader->EndRecPtr < end_of_wal)
2038 : {
2039 : XLogRecord *record;
2040 78 : char *errm = NULL;
2041 :
2042 78 : record = XLogReadRecord(ctx->reader, &errm);
2043 :
2044 78 : if (errm)
2045 0 : elog(ERROR, "could not find record for logical decoding: %s", errm);
2046 :
2047 78 : if (record != NULL)
2048 78 : LogicalDecodingProcessRecord(ctx, ctx->reader);
2049 :
2050 78 : if (ctx->processing_required)
2051 : {
2052 2 : last_pending_wal = ctx->reader->ReadRecPtr;
2053 :
2054 : /*
2055 : * If we find a decodable WAL after the scan_cutoff_lsn point,
2056 : * we can terminate the scan early.
2057 : */
2058 2 : if (last_pending_wal >= scan_cutoff_lsn)
2059 0 : break;
2060 :
2061 : /* Reset the flag and continue checking */
2062 2 : ctx->processing_required = false;
2063 : }
2064 :
2065 78 : CHECK_FOR_INTERRUPTS();
2066 : }
2067 :
2068 : /* Clean up */
2069 3 : FreeDecodingContext(ctx);
2070 3 : InvalidateSystemCaches();
2071 : }
2072 0 : PG_CATCH();
2073 : {
2074 : /* clear all timetravel entries */
2075 0 : InvalidateSystemCaches();
2076 :
2077 0 : PG_RE_THROW();
2078 : }
2079 3 : PG_END_TRY();
2080 :
2081 3 : return last_pending_wal;
2082 : }
2083 :
2084 : /*
2085 : * Helper function for advancing our logical replication slot forward.
2086 : *
2087 : * The slot's restart_lsn is used as start point for reading records, while
2088 : * confirmed_flush is used as base point for the decoding context.
2089 : *
2090 : * We cannot just do LogicalConfirmReceivedLocation to update confirmed_flush,
2091 : * because we need to digest WAL to advance restart_lsn allowing to recycle
2092 : * WAL and removal of old catalog tuples. As decoding is done in fast_forward
2093 : * mode, no changes are generated anyway.
2094 : *
2095 : * *found_consistent_snapshot will be true if the initial decoding snapshot has
2096 : * been built; Otherwise, it will be false.
2097 : */
2098 : XLogRecPtr
2099 22 : LogicalSlotAdvanceAndCheckSnapState(XLogRecPtr moveto,
2100 : bool *found_consistent_snapshot)
2101 : {
2102 : LogicalDecodingContext *ctx;
2103 22 : ResourceOwner old_resowner PG_USED_FOR_ASSERTS_ONLY = CurrentResourceOwner;
2104 : XLogRecPtr retlsn;
2105 :
2106 : Assert(XLogRecPtrIsValid(moveto));
2107 :
2108 22 : if (found_consistent_snapshot)
2109 11 : *found_consistent_snapshot = false;
2110 :
2111 22 : PG_TRY();
2112 : {
2113 : /*
2114 : * Create our decoding context in fast_forward mode, passing start_lsn
2115 : * as InvalidXLogRecPtr, so that we start processing from my slot's
2116 : * confirmed_flush.
2117 : */
2118 44 : ctx = CreateDecodingContext(InvalidXLogRecPtr,
2119 : NIL,
2120 : true, /* fast_forward */
2121 22 : XL_ROUTINE(.page_read = read_local_xlog_page,
2122 : .segment_open = wal_segment_open,
2123 : .segment_close = wal_segment_close),
2124 : NULL, NULL, NULL);
2125 :
2126 : /*
2127 : * Wait for specified streaming replication standby servers (if any)
2128 : * to confirm receipt of WAL up to moveto lsn.
2129 : */
2130 22 : WaitForStandbyConfirmation(moveto);
2131 :
2132 : /*
2133 : * Start reading at the slot's restart_lsn, which we know to point to
2134 : * a valid record.
2135 : */
2136 22 : XLogBeginRead(ctx->reader, MyReplicationSlot->data.restart_lsn);
2137 :
2138 : /* invalidate non-timetravel entries */
2139 22 : InvalidateSystemCaches();
2140 :
2141 : /* Decode records until we reach the requested target */
2142 2961 : while (ctx->reader->EndRecPtr < moveto)
2143 : {
2144 2939 : char *errm = NULL;
2145 : XLogRecord *record;
2146 :
2147 : /*
2148 : * Read records. No changes are generated in fast_forward mode,
2149 : * but snapbuilder/slot statuses are updated properly.
2150 : */
2151 2939 : record = XLogReadRecord(ctx->reader, &errm);
2152 2939 : if (errm)
2153 0 : elog(ERROR, "could not find record while advancing replication slot: %s",
2154 : errm);
2155 :
2156 : /*
2157 : * Process the record. Storage-level changes are ignored in
2158 : * fast_forward mode, but other modules (such as snapbuilder)
2159 : * might still have critical updates to do.
2160 : */
2161 2939 : if (record)
2162 : {
2163 2939 : LogicalDecodingProcessRecord(ctx, ctx->reader);
2164 :
2165 : /*
2166 : * We used to have bugs where logical decoding would fail to
2167 : * preserve the resource owner. That's important here, so
2168 : * verify that that doesn't happen anymore. XXX this could be
2169 : * removed once it's been battle-tested.
2170 : */
2171 : Assert(CurrentResourceOwner == old_resowner);
2172 : }
2173 :
2174 2939 : CHECK_FOR_INTERRUPTS();
2175 : }
2176 :
2177 22 : if (found_consistent_snapshot && DecodingContextReady(ctx))
2178 11 : *found_consistent_snapshot = true;
2179 :
2180 22 : if (XLogRecPtrIsValid(ctx->reader->EndRecPtr))
2181 : {
2182 22 : LogicalConfirmReceivedLocation(moveto);
2183 :
2184 : /*
2185 : * If only the confirmed_flush LSN has changed the slot won't get
2186 : * marked as dirty by the above. Callers on the walsender
2187 : * interface are expected to keep track of their own progress and
2188 : * don't need it written out. But SQL-interface users cannot
2189 : * specify their own start positions and it's harder for them to
2190 : * keep track of their progress, so we should make more of an
2191 : * effort to save it for them.
2192 : *
2193 : * Dirty the slot so it is written out at the next checkpoint. The
2194 : * LSN position advanced to may still be lost on a crash but this
2195 : * makes the data consistent after a clean shutdown.
2196 : */
2197 22 : ReplicationSlotMarkDirty();
2198 : }
2199 :
2200 22 : retlsn = MyReplicationSlot->data.confirmed_flush;
2201 :
2202 : /* free context, call shutdown callback */
2203 22 : FreeDecodingContext(ctx);
2204 :
2205 22 : InvalidateSystemCaches();
2206 : }
2207 0 : PG_CATCH();
2208 : {
2209 : /* clear all timetravel entries */
2210 0 : InvalidateSystemCaches();
2211 :
2212 0 : PG_RE_THROW();
2213 : }
2214 22 : PG_END_TRY();
2215 :
2216 22 : return retlsn;
2217 : }
|
