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 3182 : CheckLogicalDecodingRequirements(void)
112 : {
113 3182 : CheckSlotRequirements();
114 :
115 : /*
116 : * NB: Adding a new requirement likely means that RestoreSlotFromDisk()
117 : * needs the same check.
118 : */
119 :
120 3182 : if (MyDatabaseId == InvalidOid)
121 2 : 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 3180 : if (RecoveryInProgress() && !IsLogicalDecodingEnabled())
130 4 : 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 3176 : }
135 :
136 : /*
137 : * Helper function for CreateInitDecodingContext() and
138 : * CreateDecodingContext() performing common tasks.
139 : */
140 : static LogicalDecodingContext *
141 2232 : 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 2232 : slot = MyReplicationSlot;
159 :
160 2232 : context = AllocSetContextCreate(CurrentMemoryContext,
161 : "Logical decoding context",
162 : ALLOCSET_DEFAULT_SIZES);
163 2232 : old_context = MemoryContextSwitchTo(context);
164 2232 : ctx = palloc0_object(LogicalDecodingContext);
165 :
166 2232 : ctx->context = context;
167 :
168 : /*
169 : * (re-)load output plugins, so we detect a bad (removed) output plugin
170 : * now.
171 : */
172 2232 : if (!fast_forward)
173 2186 : 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 2230 : if (!IsTransactionOrTransactionBlock())
188 : {
189 1110 : LWLockAcquire(ProcArrayLock, LW_EXCLUSIVE);
190 1110 : MyProc->statusFlags |= PROC_IN_LOGICAL_DECODING;
191 1110 : ProcGlobal->statusFlags[MyProc->pgxactoff] = MyProc->statusFlags;
192 1110 : LWLockRelease(ProcArrayLock);
193 : }
194 :
195 2230 : ctx->slot = slot;
196 :
197 2230 : ctx->reader = XLogReaderAllocate(wal_segment_size, NULL, xl_routine, ctx);
198 2230 : 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 2230 : ctx->reorder = ReorderBufferAllocate();
205 2230 : ctx->snapshot_builder =
206 2230 : AllocateSnapshotBuilder(ctx->reorder, xmin_horizon, start_lsn,
207 : need_full_snapshot, in_create, slot->data.two_phase_at);
208 :
209 2230 : ctx->reorder->private_data = ctx;
210 :
211 : /* wrap output plugin callbacks, so we can add error context information */
212 2230 : ctx->reorder->begin = begin_cb_wrapper;
213 2230 : ctx->reorder->apply_change = change_cb_wrapper;
214 2230 : ctx->reorder->apply_truncate = truncate_cb_wrapper;
215 2230 : ctx->reorder->commit = commit_cb_wrapper;
216 2230 : 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 4506 : ctx->streaming = (ctx->callbacks.stream_start_cb != NULL) ||
228 46 : (ctx->callbacks.stream_stop_cb != NULL) ||
229 46 : (ctx->callbacks.stream_abort_cb != NULL) ||
230 46 : (ctx->callbacks.stream_commit_cb != NULL) ||
231 46 : (ctx->callbacks.stream_change_cb != NULL) ||
232 2322 : (ctx->callbacks.stream_message_cb != NULL) ||
233 46 : (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 2230 : ctx->reorder->stream_start = stream_start_cb_wrapper;
244 2230 : ctx->reorder->stream_stop = stream_stop_cb_wrapper;
245 2230 : ctx->reorder->stream_abort = stream_abort_cb_wrapper;
246 2230 : ctx->reorder->stream_prepare = stream_prepare_cb_wrapper;
247 2230 : ctx->reorder->stream_commit = stream_commit_cb_wrapper;
248 2230 : ctx->reorder->stream_change = stream_change_cb_wrapper;
249 2230 : ctx->reorder->stream_message = stream_message_cb_wrapper;
250 2230 : 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 4506 : ctx->twophase = (ctx->callbacks.begin_prepare_cb != NULL) ||
263 46 : (ctx->callbacks.prepare_cb != NULL) ||
264 46 : (ctx->callbacks.commit_prepared_cb != NULL) ||
265 46 : (ctx->callbacks.rollback_prepared_cb != NULL) ||
266 2322 : (ctx->callbacks.stream_prepare_cb != NULL) ||
267 46 : (ctx->callbacks.filter_prepare_cb != NULL);
268 :
269 : /*
270 : * Callback to support decoding at prepare time.
271 : */
272 2230 : ctx->reorder->begin_prepare = begin_prepare_cb_wrapper;
273 2230 : ctx->reorder->prepare = prepare_cb_wrapper;
274 2230 : ctx->reorder->commit_prepared = commit_prepared_cb_wrapper;
275 2230 : 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 2230 : ctx->reorder->update_progress_txn = update_progress_txn_cb_wrapper;
282 :
283 2230 : ctx->out = makeStringInfo();
284 2230 : ctx->prepare_write = prepare_write;
285 2230 : ctx->write = do_write;
286 2230 : ctx->update_progress = update_progress;
287 :
288 2230 : ctx->output_plugin_options = output_plugin_options;
289 :
290 2230 : ctx->fast_forward = fast_forward;
291 :
292 2230 : MemoryContextSwitchTo(old_context);
293 :
294 2230 : return ctx;
295 : }
296 :
297 : /*
298 : * Create a new decoding context, for a new logical slot.
299 : *
300 : * plugin -- contains the name of the output plugin
301 : * output_plugin_options -- contains options passed to the output plugin
302 : * need_full_snapshot -- if true, must obtain a snapshot able to read all
303 : * tables; if false, one that can read only catalogs is acceptable.
304 : * restart_lsn -- if given as invalid, it's this routine's responsibility to
305 : * mark WAL as reserved by setting a convenient restart_lsn for the slot.
306 : * Otherwise, we set for decoding to start from the given LSN without
307 : * marking WAL reserved beforehand. In that scenario, it's up to the
308 : * caller to guarantee that WAL remains available.
309 : * xl_routine -- XLogReaderRoutine for underlying XLogReader
310 : * prepare_write, do_write, update_progress --
311 : * callbacks that perform the use-case dependent, actual, work.
312 : *
313 : * Needs to be called while in a memory context that's at least as long lived
314 : * as the decoding context because further memory contexts will be created
315 : * inside it.
316 : *
317 : * Returns an initialized decoding context after calling the output plugin's
318 : * startup function.
319 : */
320 : LogicalDecodingContext *
321 948 : CreateInitDecodingContext(const char *plugin,
322 : List *output_plugin_options,
323 : bool need_full_snapshot,
324 : XLogRecPtr restart_lsn,
325 : XLogReaderRoutine *xl_routine,
326 : LogicalOutputPluginWriterPrepareWrite prepare_write,
327 : LogicalOutputPluginWriterWrite do_write,
328 : LogicalOutputPluginWriterUpdateProgress update_progress)
329 : {
330 948 : TransactionId xmin_horizon = InvalidTransactionId;
331 : ReplicationSlot *slot;
332 : NameData plugin_name;
333 : LogicalDecodingContext *ctx;
334 : MemoryContext old_context;
335 :
336 : /*
337 : * On a standby, this check is also required while creating the slot.
338 : * Check the comments in the function.
339 : */
340 948 : CheckLogicalDecodingRequirements();
341 :
342 : /* shorter lines... */
343 948 : slot = MyReplicationSlot;
344 :
345 : /* first some sanity checks that are unlikely to be violated */
346 948 : if (slot == NULL)
347 0 : elog(ERROR, "cannot perform logical decoding without an acquired slot");
348 :
349 948 : if (plugin == NULL)
350 0 : elog(ERROR, "cannot initialize logical decoding without a specified plugin");
351 :
352 : /* Make sure the passed slot is suitable. These are user facing errors. */
353 948 : if (SlotIsPhysical(slot))
354 0 : ereport(ERROR,
355 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
356 : errmsg("cannot use physical replication slot for logical decoding")));
357 :
358 948 : if (slot->data.database != MyDatabaseId)
359 0 : ereport(ERROR,
360 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
361 : errmsg("replication slot \"%s\" was not created in this database",
362 : NameStr(slot->data.name))));
363 :
364 1610 : if (IsTransactionState() &&
365 662 : GetTopTransactionIdIfAny() != InvalidTransactionId)
366 4 : ereport(ERROR,
367 : (errcode(ERRCODE_ACTIVE_SQL_TRANSACTION),
368 : errmsg("cannot create logical replication slot in transaction that has performed writes")));
369 :
370 : /*
371 : * Register output plugin name with slot. We need the mutex to avoid
372 : * concurrent reading of a partially copied string. But we don't want any
373 : * complicated code while holding a spinlock, so do namestrcpy() outside.
374 : */
375 944 : namestrcpy(&plugin_name, plugin);
376 944 : SpinLockAcquire(&slot->mutex);
377 944 : slot->data.plugin = plugin_name;
378 944 : SpinLockRelease(&slot->mutex);
379 :
380 944 : if (!XLogRecPtrIsValid(restart_lsn))
381 930 : ReplicationSlotReserveWal();
382 : else
383 : {
384 14 : SpinLockAcquire(&slot->mutex);
385 14 : slot->data.restart_lsn = restart_lsn;
386 14 : SpinLockRelease(&slot->mutex);
387 : }
388 :
389 : /* ----
390 : * This is a bit tricky: We need to determine a safe xmin horizon to start
391 : * decoding from, to avoid starting from a running xacts record referring
392 : * to xids whose rows have been vacuumed or pruned
393 : * already. GetOldestSafeDecodingTransactionId() returns such a value, but
394 : * without further interlock its return value might immediately be out of
395 : * date.
396 : *
397 : * So we have to acquire both the ReplicationSlotControlLock and the
398 : * ProcArrayLock to prevent concurrent computation and update of new xmin
399 : * horizons by other backends, get the safe decoding xid, and inform the
400 : * slot machinery about the new limit. Once that's done both locks can be
401 : * released as the slot machinery now is protecting against vacuum.
402 : *
403 : * Note that, temporarily, the data, not just the catalog, xmin has to be
404 : * reserved if a data snapshot is to be exported. Otherwise the initial
405 : * data snapshot created here is not guaranteed to be valid. After that
406 : * the data xmin doesn't need to be managed anymore and the global xmin
407 : * should be recomputed. As we are fine with losing the pegged data xmin
408 : * after crash - no chance a snapshot would get exported anymore - we can
409 : * get away with just setting the slot's
410 : * effective_xmin. ReplicationSlotRelease will reset it again.
411 : *
412 : * ----
413 : */
414 944 : LWLockAcquire(ReplicationSlotControlLock, LW_EXCLUSIVE);
415 944 : LWLockAcquire(ProcArrayLock, LW_EXCLUSIVE);
416 :
417 944 : xmin_horizon = GetOldestSafeDecodingTransactionId(!need_full_snapshot);
418 :
419 944 : SpinLockAcquire(&slot->mutex);
420 944 : slot->effective_catalog_xmin = xmin_horizon;
421 944 : slot->data.catalog_xmin = xmin_horizon;
422 944 : if (need_full_snapshot)
423 398 : slot->effective_xmin = xmin_horizon;
424 944 : SpinLockRelease(&slot->mutex);
425 :
426 944 : ReplicationSlotsComputeRequiredXmin(true);
427 :
428 944 : LWLockRelease(ProcArrayLock);
429 944 : LWLockRelease(ReplicationSlotControlLock);
430 :
431 944 : ReplicationSlotMarkDirty();
432 944 : ReplicationSlotSave();
433 :
434 944 : ctx = StartupDecodingContext(NIL, restart_lsn, xmin_horizon,
435 : need_full_snapshot, false, true,
436 : xl_routine, prepare_write, do_write,
437 : update_progress);
438 :
439 : /* call output plugin initialization callback */
440 942 : old_context = MemoryContextSwitchTo(ctx->context);
441 942 : if (ctx->callbacks.startup_cb != NULL)
442 942 : startup_cb_wrapper(ctx, &ctx->options, true);
443 942 : MemoryContextSwitchTo(old_context);
444 :
445 : /*
446 : * We allow decoding of prepared transactions when the two_phase is
447 : * enabled at the time of slot creation, or when the two_phase option is
448 : * given at the streaming start, provided the plugin supports all the
449 : * callbacks for two-phase.
450 : */
451 942 : ctx->twophase &= slot->data.two_phase;
452 :
453 942 : ctx->reorder->output_rewrites = ctx->options.receive_rewrites;
454 :
455 942 : return ctx;
456 : }
457 :
458 : /*
459 : * Create a new decoding context, for a logical slot that has previously been
460 : * used already.
461 : *
462 : * start_lsn
463 : * The LSN at which to start decoding. If InvalidXLogRecPtr, restart
464 : * from the slot's confirmed_flush; otherwise, start from the specified
465 : * location (but move it forwards to confirmed_flush if it's older than
466 : * that, see below).
467 : *
468 : * output_plugin_options
469 : * options passed to the output plugin.
470 : *
471 : * fast_forward
472 : * bypass the generation of logical changes.
473 : *
474 : * xl_routine
475 : * XLogReaderRoutine used by underlying xlogreader
476 : *
477 : * prepare_write, do_write, update_progress
478 : * callbacks that have to be filled to perform the use-case dependent,
479 : * actual work.
480 : *
481 : * Needs to be called while in a memory context that's at least as long lived
482 : * as the decoding context because further memory contexts will be created
483 : * inside it.
484 : *
485 : * Returns an initialized decoding context after calling the output plugin's
486 : * startup function.
487 : */
488 : LogicalDecodingContext *
489 1298 : CreateDecodingContext(XLogRecPtr start_lsn,
490 : List *output_plugin_options,
491 : bool fast_forward,
492 : XLogReaderRoutine *xl_routine,
493 : LogicalOutputPluginWriterPrepareWrite prepare_write,
494 : LogicalOutputPluginWriterWrite do_write,
495 : LogicalOutputPluginWriterUpdateProgress update_progress)
496 : {
497 : LogicalDecodingContext *ctx;
498 : ReplicationSlot *slot;
499 : MemoryContext old_context;
500 :
501 : /* shorter lines... */
502 1298 : slot = MyReplicationSlot;
503 :
504 : /* first some sanity checks that are unlikely to be violated */
505 1298 : if (slot == NULL)
506 0 : elog(ERROR, "cannot perform logical decoding without an acquired slot");
507 :
508 : /* make sure the passed slot is suitable, these are user facing errors */
509 1298 : if (SlotIsPhysical(slot))
510 2 : ereport(ERROR,
511 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
512 : errmsg("cannot use physical replication slot for logical decoding")));
513 :
514 : /*
515 : * We need to access the system tables during decoding to build the
516 : * logical changes unless we are in fast_forward mode where no changes are
517 : * generated.
518 : */
519 1296 : if (slot->data.database != MyDatabaseId && !fast_forward)
520 6 : ereport(ERROR,
521 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
522 : errmsg("replication slot \"%s\" was not created in this database",
523 : NameStr(slot->data.name))));
524 :
525 : /*
526 : * The slots being synced from the primary can't be used for decoding as
527 : * they are used after failover. However, we do allow advancing the LSNs
528 : * during the synchronization of slots. See update_local_synced_slot.
529 : */
530 1290 : if (RecoveryInProgress() && slot->data.synced && !IsSyncingReplicationSlots())
531 2 : ereport(ERROR,
532 : errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
533 : errmsg("cannot use replication slot \"%s\" for logical decoding",
534 : NameStr(slot->data.name)),
535 : errdetail("This replication slot is being synchronized from the primary server."),
536 : errhint("Specify another replication slot."));
537 :
538 : /* slot must be valid to allow decoding */
539 : Assert(slot->data.invalidated == RS_INVAL_NONE);
540 : Assert(XLogRecPtrIsValid(slot->data.restart_lsn));
541 :
542 1288 : if (!XLogRecPtrIsValid(start_lsn))
543 : {
544 : /* continue from last position */
545 750 : start_lsn = slot->data.confirmed_flush;
546 : }
547 538 : else if (start_lsn < slot->data.confirmed_flush)
548 : {
549 : /*
550 : * It might seem like we should error out in this case, but it's
551 : * pretty common for a client to acknowledge a LSN it doesn't have to
552 : * do anything for, and thus didn't store persistently, because the
553 : * xlog records didn't result in anything relevant for logical
554 : * decoding. Clients have to be able to do that to support synchronous
555 : * replication.
556 : *
557 : * Starting at a different LSN than requested might not catch certain
558 : * kinds of client errors; so the client may wish to check that
559 : * confirmed_flush_lsn matches its expectations.
560 : */
561 80 : elog(LOG, "%X/%08X has been already streamed, forwarding to %X/%08X",
562 : LSN_FORMAT_ARGS(start_lsn),
563 : LSN_FORMAT_ARGS(slot->data.confirmed_flush));
564 :
565 80 : start_lsn = slot->data.confirmed_flush;
566 : }
567 :
568 1288 : ctx = StartupDecodingContext(output_plugin_options,
569 : start_lsn, InvalidTransactionId, false,
570 : fast_forward, false, xl_routine, prepare_write,
571 : do_write, update_progress);
572 :
573 : /* call output plugin initialization callback */
574 1288 : old_context = MemoryContextSwitchTo(ctx->context);
575 1288 : if (ctx->callbacks.startup_cb != NULL)
576 1242 : startup_cb_wrapper(ctx, &ctx->options, false);
577 1282 : MemoryContextSwitchTo(old_context);
578 :
579 : /*
580 : * We allow decoding of prepared transactions when the two_phase is
581 : * enabled at the time of slot creation, or when the two_phase option is
582 : * given at the streaming start, provided the plugin supports all the
583 : * callbacks for two-phase.
584 : */
585 1282 : ctx->twophase &= (slot->data.two_phase || ctx->twophase_opt_given);
586 :
587 : /* Mark slot to allow two_phase decoding if not already marked */
588 1282 : if (ctx->twophase && !slot->data.two_phase)
589 : {
590 16 : SpinLockAcquire(&slot->mutex);
591 16 : slot->data.two_phase = true;
592 16 : slot->data.two_phase_at = start_lsn;
593 16 : SpinLockRelease(&slot->mutex);
594 16 : ReplicationSlotMarkDirty();
595 16 : ReplicationSlotSave();
596 16 : SnapBuildSetTwoPhaseAt(ctx->snapshot_builder, start_lsn);
597 : }
598 :
599 1282 : ctx->reorder->output_rewrites = ctx->options.receive_rewrites;
600 :
601 1282 : ereport(LOG,
602 : (errmsg("starting logical decoding for slot \"%s\"",
603 : NameStr(slot->data.name)),
604 : errdetail("Streaming transactions committing after %X/%08X, reading WAL from %X/%08X.",
605 : LSN_FORMAT_ARGS(slot->data.confirmed_flush),
606 : LSN_FORMAT_ARGS(slot->data.restart_lsn))));
607 :
608 1282 : return ctx;
609 : }
610 :
611 : /*
612 : * Returns true if a consistent initial decoding snapshot has been built.
613 : */
614 : bool
615 1030 : DecodingContextReady(LogicalDecodingContext *ctx)
616 : {
617 1030 : return SnapBuildCurrentState(ctx->snapshot_builder) == SNAPBUILD_CONSISTENT;
618 : }
619 :
620 : /*
621 : * Read from the decoding slot, until it is ready to start extracting changes.
622 : */
623 : void
624 928 : DecodingContextFindStartpoint(LogicalDecodingContext *ctx)
625 : {
626 928 : ReplicationSlot *slot = ctx->slot;
627 :
628 : /* Initialize from where to start reading WAL. */
629 928 : XLogBeginRead(ctx->reader, slot->data.restart_lsn);
630 :
631 928 : elog(DEBUG1, "searching for logical decoding starting point, starting at %X/%08X",
632 : LSN_FORMAT_ARGS(slot->data.restart_lsn));
633 :
634 : /* Wait for a consistent starting point */
635 : for (;;)
636 94 : {
637 : XLogRecord *record;
638 1022 : char *err = NULL;
639 :
640 : /* the read_page callback waits for new WAL */
641 1022 : record = XLogReadRecord(ctx->reader, &err);
642 1022 : if (err)
643 0 : elog(ERROR, "could not find logical decoding starting point: %s", err);
644 1022 : if (!record)
645 0 : elog(ERROR, "could not find logical decoding starting point");
646 :
647 1022 : LogicalDecodingProcessRecord(ctx, ctx->reader);
648 :
649 : /* only continue till we found a consistent spot */
650 1018 : if (DecodingContextReady(ctx))
651 924 : break;
652 :
653 94 : CHECK_FOR_INTERRUPTS();
654 : }
655 :
656 924 : SpinLockAcquire(&slot->mutex);
657 924 : slot->data.confirmed_flush = ctx->reader->EndRecPtr;
658 924 : if (slot->data.two_phase)
659 18 : slot->data.two_phase_at = ctx->reader->EndRecPtr;
660 924 : SpinLockRelease(&slot->mutex);
661 924 : }
662 :
663 : /*
664 : * Free a previously allocated decoding context, invoking the shutdown
665 : * callback if necessary.
666 : */
667 : void
668 1780 : FreeDecodingContext(LogicalDecodingContext *ctx)
669 : {
670 1780 : if (ctx->callbacks.shutdown_cb != NULL)
671 1734 : shutdown_cb_wrapper(ctx);
672 :
673 1780 : ReorderBufferFree(ctx->reorder);
674 1780 : FreeSnapshotBuilder(ctx->snapshot_builder);
675 1780 : XLogReaderFree(ctx->reader);
676 1780 : MemoryContextDelete(ctx->context);
677 1780 : }
678 :
679 : /*
680 : * Prepare a write using the context's output routine.
681 : */
682 : void
683 672414 : OutputPluginPrepareWrite(struct LogicalDecodingContext *ctx, bool last_write)
684 : {
685 672414 : if (!ctx->accept_writes)
686 0 : elog(ERROR, "writes are only accepted in commit, begin and change callbacks");
687 :
688 672414 : ctx->prepare_write(ctx, ctx->write_location, ctx->write_xid, last_write);
689 672414 : ctx->prepared_write = true;
690 672414 : }
691 :
692 : /*
693 : * Perform a write using the context's output routine.
694 : */
695 : void
696 672414 : OutputPluginWrite(struct LogicalDecodingContext *ctx, bool last_write)
697 : {
698 672414 : if (!ctx->prepared_write)
699 0 : elog(ERROR, "OutputPluginPrepareWrite needs to be called before OutputPluginWrite");
700 :
701 672414 : ctx->write(ctx, ctx->write_location, ctx->write_xid, last_write);
702 672378 : ctx->prepared_write = false;
703 672378 : }
704 :
705 : /*
706 : * Update progress tracking (if supported).
707 : */
708 : void
709 8314 : OutputPluginUpdateProgress(struct LogicalDecodingContext *ctx,
710 : bool skipped_xact)
711 : {
712 8314 : if (!ctx->update_progress)
713 3180 : return;
714 :
715 5134 : ctx->update_progress(ctx, ctx->write_location, ctx->write_xid,
716 : skipped_xact);
717 : }
718 :
719 : /*
720 : * Load the output plugin, lookup its output plugin init function, and check
721 : * that it provides the required callbacks.
722 : */
723 : static void
724 2186 : LoadOutputPlugin(OutputPluginCallbacks *callbacks, const char *plugin)
725 : {
726 : LogicalOutputPluginInit plugin_init;
727 :
728 2184 : plugin_init = (LogicalOutputPluginInit)
729 2186 : load_external_function(plugin, "_PG_output_plugin_init", false, NULL);
730 :
731 2184 : if (plugin_init == NULL)
732 0 : elog(ERROR, "output plugins have to declare the _PG_output_plugin_init symbol");
733 :
734 : /* ask the output plugin to fill the callback struct */
735 2184 : plugin_init(callbacks);
736 :
737 2184 : if (callbacks->begin_cb == NULL)
738 0 : elog(ERROR, "output plugins have to register a begin callback");
739 2184 : if (callbacks->change_cb == NULL)
740 0 : elog(ERROR, "output plugins have to register a change callback");
741 2184 : if (callbacks->commit_cb == NULL)
742 0 : elog(ERROR, "output plugins have to register a commit callback");
743 2184 : }
744 :
745 : static void
746 278 : output_plugin_error_callback(void *arg)
747 : {
748 278 : LogicalErrorCallbackState *state = (LogicalErrorCallbackState *) arg;
749 :
750 : /* not all callbacks have an associated LSN */
751 278 : if (XLogRecPtrIsValid(state->report_location))
752 272 : errcontext("slot \"%s\", output plugin \"%s\", in the %s callback, associated LSN %X/%08X",
753 272 : NameStr(state->ctx->slot->data.name),
754 272 : NameStr(state->ctx->slot->data.plugin),
755 : state->callback_name,
756 272 : LSN_FORMAT_ARGS(state->report_location));
757 : else
758 6 : errcontext("slot \"%s\", output plugin \"%s\", in the %s callback",
759 6 : NameStr(state->ctx->slot->data.name),
760 6 : NameStr(state->ctx->slot->data.plugin),
761 : state->callback_name);
762 278 : }
763 :
764 : static void
765 2184 : startup_cb_wrapper(LogicalDecodingContext *ctx, OutputPluginOptions *opt, bool is_init)
766 : {
767 : LogicalErrorCallbackState state;
768 : ErrorContextCallback errcallback;
769 :
770 : Assert(!ctx->fast_forward);
771 :
772 : /* Push callback + info on the error context stack */
773 2184 : state.ctx = ctx;
774 2184 : state.callback_name = "startup";
775 2184 : state.report_location = InvalidXLogRecPtr;
776 2184 : errcallback.callback = output_plugin_error_callback;
777 2184 : errcallback.arg = &state;
778 2184 : errcallback.previous = error_context_stack;
779 2184 : error_context_stack = &errcallback;
780 :
781 : /* set output state */
782 2184 : ctx->accept_writes = false;
783 2184 : ctx->end_xact = false;
784 :
785 : /* do the actual work: call callback */
786 2184 : ctx->callbacks.startup_cb(ctx, opt, is_init);
787 :
788 : /* Pop the error context stack */
789 2178 : error_context_stack = errcallback.previous;
790 2178 : }
791 :
792 : static void
793 1734 : shutdown_cb_wrapper(LogicalDecodingContext *ctx)
794 : {
795 : LogicalErrorCallbackState state;
796 : ErrorContextCallback errcallback;
797 :
798 : Assert(!ctx->fast_forward);
799 :
800 : /* Push callback + info on the error context stack */
801 1734 : state.ctx = ctx;
802 1734 : state.callback_name = "shutdown";
803 1734 : state.report_location = InvalidXLogRecPtr;
804 1734 : errcallback.callback = output_plugin_error_callback;
805 1734 : errcallback.arg = &state;
806 1734 : errcallback.previous = error_context_stack;
807 1734 : error_context_stack = &errcallback;
808 :
809 : /* set output state */
810 1734 : ctx->accept_writes = false;
811 1734 : ctx->end_xact = false;
812 :
813 : /* do the actual work: call callback */
814 1734 : ctx->callbacks.shutdown_cb(ctx);
815 :
816 : /* Pop the error context stack */
817 1734 : error_context_stack = errcallback.previous;
818 1734 : }
819 :
820 :
821 : /*
822 : * Callbacks for ReorderBuffer which add in some more information and then call
823 : * output_plugin.h plugins.
824 : */
825 : static void
826 2710 : begin_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn)
827 : {
828 2710 : LogicalDecodingContext *ctx = cache->private_data;
829 : LogicalErrorCallbackState state;
830 : ErrorContextCallback errcallback;
831 :
832 : Assert(!ctx->fast_forward);
833 :
834 : /* Push callback + info on the error context stack */
835 2710 : state.ctx = ctx;
836 2710 : state.callback_name = "begin";
837 2710 : state.report_location = txn->first_lsn;
838 2710 : errcallback.callback = output_plugin_error_callback;
839 2710 : errcallback.arg = &state;
840 2710 : errcallback.previous = error_context_stack;
841 2710 : error_context_stack = &errcallback;
842 :
843 : /* set output state */
844 2710 : ctx->accept_writes = true;
845 2710 : ctx->write_xid = txn->xid;
846 2710 : ctx->write_location = txn->first_lsn;
847 2710 : ctx->end_xact = false;
848 :
849 : /* do the actual work: call callback */
850 2710 : ctx->callbacks.begin_cb(ctx, txn);
851 :
852 : /* Pop the error context stack */
853 2710 : error_context_stack = errcallback.previous;
854 2710 : }
855 :
856 : static void
857 2704 : commit_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
858 : XLogRecPtr commit_lsn)
859 : {
860 2704 : LogicalDecodingContext *ctx = cache->private_data;
861 : LogicalErrorCallbackState state;
862 : ErrorContextCallback errcallback;
863 :
864 : Assert(!ctx->fast_forward);
865 :
866 : /* Push callback + info on the error context stack */
867 2704 : state.ctx = ctx;
868 2704 : state.callback_name = "commit";
869 2704 : state.report_location = txn->final_lsn; /* beginning of commit record */
870 2704 : errcallback.callback = output_plugin_error_callback;
871 2704 : errcallback.arg = &state;
872 2704 : errcallback.previous = error_context_stack;
873 2704 : error_context_stack = &errcallback;
874 :
875 : /* set output state */
876 2704 : ctx->accept_writes = true;
877 2704 : ctx->write_xid = txn->xid;
878 2704 : ctx->write_location = txn->end_lsn; /* points to the end of the record */
879 2704 : ctx->end_xact = true;
880 :
881 : /* do the actual work: call callback */
882 2704 : ctx->callbacks.commit_cb(ctx, txn, commit_lsn);
883 :
884 : /* Pop the error context stack */
885 2672 : error_context_stack = errcallback.previous;
886 2672 : }
887 :
888 : /*
889 : * The functionality of begin_prepare is quite similar to begin with the
890 : * exception that this will have gid (global transaction id) information which
891 : * can be used by plugin. Now, we thought about extending the existing begin
892 : * but that would break the replication protocol and additionally this looks
893 : * cleaner.
894 : */
895 : static void
896 60 : begin_prepare_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn)
897 : {
898 60 : LogicalDecodingContext *ctx = cache->private_data;
899 : LogicalErrorCallbackState state;
900 : ErrorContextCallback errcallback;
901 :
902 : Assert(!ctx->fast_forward);
903 :
904 : /* We're only supposed to call this when two-phase commits are supported */
905 : Assert(ctx->twophase);
906 :
907 : /* Push callback + info on the error context stack */
908 60 : state.ctx = ctx;
909 60 : state.callback_name = "begin_prepare";
910 60 : state.report_location = txn->first_lsn;
911 60 : errcallback.callback = output_plugin_error_callback;
912 60 : errcallback.arg = &state;
913 60 : errcallback.previous = error_context_stack;
914 60 : error_context_stack = &errcallback;
915 :
916 : /* set output state */
917 60 : ctx->accept_writes = true;
918 60 : ctx->write_xid = txn->xid;
919 60 : ctx->write_location = txn->first_lsn;
920 60 : ctx->end_xact = false;
921 :
922 : /*
923 : * If the plugin supports two-phase commits then begin prepare callback is
924 : * mandatory
925 : */
926 60 : if (ctx->callbacks.begin_prepare_cb == NULL)
927 0 : ereport(ERROR,
928 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
929 : errmsg("logical replication at prepare time requires a %s callback",
930 : "begin_prepare_cb")));
931 :
932 : /* do the actual work: call callback */
933 60 : ctx->callbacks.begin_prepare_cb(ctx, txn);
934 :
935 : /* Pop the error context stack */
936 60 : error_context_stack = errcallback.previous;
937 60 : }
938 :
939 : static void
940 60 : prepare_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
941 : XLogRecPtr prepare_lsn)
942 : {
943 60 : LogicalDecodingContext *ctx = cache->private_data;
944 : LogicalErrorCallbackState state;
945 : ErrorContextCallback errcallback;
946 :
947 : Assert(!ctx->fast_forward);
948 :
949 : /* We're only supposed to call this when two-phase commits are supported */
950 : Assert(ctx->twophase);
951 :
952 : /* Push callback + info on the error context stack */
953 60 : state.ctx = ctx;
954 60 : state.callback_name = "prepare";
955 60 : state.report_location = txn->final_lsn; /* beginning of prepare record */
956 60 : errcallback.callback = output_plugin_error_callback;
957 60 : errcallback.arg = &state;
958 60 : errcallback.previous = error_context_stack;
959 60 : error_context_stack = &errcallback;
960 :
961 : /* set output state */
962 60 : ctx->accept_writes = true;
963 60 : ctx->write_xid = txn->xid;
964 60 : ctx->write_location = txn->end_lsn; /* points to the end of the record */
965 60 : ctx->end_xact = true;
966 :
967 : /*
968 : * If the plugin supports two-phase commits then prepare callback is
969 : * mandatory
970 : */
971 60 : if (ctx->callbacks.prepare_cb == NULL)
972 0 : ereport(ERROR,
973 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
974 : errmsg("logical replication at prepare time requires a %s callback",
975 : "prepare_cb")));
976 :
977 : /* do the actual work: call callback */
978 60 : ctx->callbacks.prepare_cb(ctx, txn, prepare_lsn);
979 :
980 : /* Pop the error context stack */
981 60 : error_context_stack = errcallback.previous;
982 60 : }
983 :
984 : static void
985 66 : commit_prepared_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
986 : XLogRecPtr commit_lsn)
987 : {
988 66 : LogicalDecodingContext *ctx = cache->private_data;
989 : LogicalErrorCallbackState state;
990 : ErrorContextCallback errcallback;
991 :
992 : Assert(!ctx->fast_forward);
993 :
994 : /* We're only supposed to call this when two-phase commits are supported */
995 : Assert(ctx->twophase);
996 :
997 : /* Push callback + info on the error context stack */
998 66 : state.ctx = ctx;
999 66 : state.callback_name = "commit_prepared";
1000 66 : state.report_location = txn->final_lsn; /* beginning of commit record */
1001 66 : errcallback.callback = output_plugin_error_callback;
1002 66 : errcallback.arg = &state;
1003 66 : errcallback.previous = error_context_stack;
1004 66 : error_context_stack = &errcallback;
1005 :
1006 : /* set output state */
1007 66 : ctx->accept_writes = true;
1008 66 : ctx->write_xid = txn->xid;
1009 66 : ctx->write_location = txn->end_lsn; /* points to the end of the record */
1010 66 : ctx->end_xact = true;
1011 :
1012 : /*
1013 : * If the plugin support two-phase commits then commit prepared callback
1014 : * is mandatory
1015 : */
1016 66 : if (ctx->callbacks.commit_prepared_cb == NULL)
1017 0 : ereport(ERROR,
1018 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1019 : errmsg("logical replication at prepare time requires a %s callback",
1020 : "commit_prepared_cb")));
1021 :
1022 : /* do the actual work: call callback */
1023 66 : ctx->callbacks.commit_prepared_cb(ctx, txn, commit_lsn);
1024 :
1025 : /* Pop the error context stack */
1026 66 : error_context_stack = errcallback.previous;
1027 66 : }
1028 :
1029 : static void
1030 20 : rollback_prepared_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1031 : XLogRecPtr prepare_end_lsn,
1032 : TimestampTz prepare_time)
1033 : {
1034 20 : LogicalDecodingContext *ctx = cache->private_data;
1035 : LogicalErrorCallbackState state;
1036 : ErrorContextCallback errcallback;
1037 :
1038 : Assert(!ctx->fast_forward);
1039 :
1040 : /* We're only supposed to call this when two-phase commits are supported */
1041 : Assert(ctx->twophase);
1042 :
1043 : /* Push callback + info on the error context stack */
1044 20 : state.ctx = ctx;
1045 20 : state.callback_name = "rollback_prepared";
1046 20 : state.report_location = txn->final_lsn; /* beginning of commit record */
1047 20 : errcallback.callback = output_plugin_error_callback;
1048 20 : errcallback.arg = &state;
1049 20 : errcallback.previous = error_context_stack;
1050 20 : error_context_stack = &errcallback;
1051 :
1052 : /* set output state */
1053 20 : ctx->accept_writes = true;
1054 20 : ctx->write_xid = txn->xid;
1055 20 : ctx->write_location = txn->end_lsn; /* points to the end of the record */
1056 20 : ctx->end_xact = true;
1057 :
1058 : /*
1059 : * If the plugin support two-phase commits then rollback prepared callback
1060 : * is mandatory
1061 : */
1062 20 : if (ctx->callbacks.rollback_prepared_cb == NULL)
1063 0 : ereport(ERROR,
1064 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1065 : errmsg("logical replication at prepare time requires a %s callback",
1066 : "rollback_prepared_cb")));
1067 :
1068 : /* do the actual work: call callback */
1069 20 : ctx->callbacks.rollback_prepared_cb(ctx, txn, prepare_end_lsn,
1070 : prepare_time);
1071 :
1072 : /* Pop the error context stack */
1073 20 : error_context_stack = errcallback.previous;
1074 20 : }
1075 :
1076 : static void
1077 322142 : change_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1078 : Relation relation, ReorderBufferChange *change)
1079 : {
1080 322142 : LogicalDecodingContext *ctx = cache->private_data;
1081 : LogicalErrorCallbackState state;
1082 : ErrorContextCallback errcallback;
1083 :
1084 : Assert(!ctx->fast_forward);
1085 :
1086 : /* Push callback + info on the error context stack */
1087 322142 : state.ctx = ctx;
1088 322142 : state.callback_name = "change";
1089 322142 : state.report_location = change->lsn;
1090 322142 : errcallback.callback = output_plugin_error_callback;
1091 322142 : errcallback.arg = &state;
1092 322142 : errcallback.previous = error_context_stack;
1093 322142 : error_context_stack = &errcallback;
1094 :
1095 : /* set output state */
1096 322142 : ctx->accept_writes = true;
1097 322142 : ctx->write_xid = txn->xid;
1098 :
1099 : /*
1100 : * Report this change's lsn so replies from clients can give an up-to-date
1101 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1102 : * receipt of this transaction, but it might allow another transaction's
1103 : * commit to be confirmed with one message.
1104 : */
1105 322142 : ctx->write_location = change->lsn;
1106 :
1107 322142 : ctx->end_xact = false;
1108 :
1109 322142 : ctx->callbacks.change_cb(ctx, txn, relation, change);
1110 :
1111 : /* Pop the error context stack */
1112 322136 : error_context_stack = errcallback.previous;
1113 322136 : }
1114 :
1115 : static void
1116 52 : truncate_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1117 : int nrelations, Relation relations[], ReorderBufferChange *change)
1118 : {
1119 52 : LogicalDecodingContext *ctx = cache->private_data;
1120 : LogicalErrorCallbackState state;
1121 : ErrorContextCallback errcallback;
1122 :
1123 : Assert(!ctx->fast_forward);
1124 :
1125 52 : if (!ctx->callbacks.truncate_cb)
1126 0 : return;
1127 :
1128 : /* Push callback + info on the error context stack */
1129 52 : state.ctx = ctx;
1130 52 : state.callback_name = "truncate";
1131 52 : state.report_location = change->lsn;
1132 52 : errcallback.callback = output_plugin_error_callback;
1133 52 : errcallback.arg = &state;
1134 52 : errcallback.previous = error_context_stack;
1135 52 : error_context_stack = &errcallback;
1136 :
1137 : /* set output state */
1138 52 : ctx->accept_writes = true;
1139 52 : ctx->write_xid = txn->xid;
1140 :
1141 : /*
1142 : * Report this change's lsn so replies from clients can give an up-to-date
1143 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1144 : * receipt of this transaction, but it might allow another transaction's
1145 : * commit to be confirmed with one message.
1146 : */
1147 52 : ctx->write_location = change->lsn;
1148 :
1149 52 : ctx->end_xact = false;
1150 :
1151 52 : ctx->callbacks.truncate_cb(ctx, txn, nrelations, relations, change);
1152 :
1153 : /* Pop the error context stack */
1154 52 : error_context_stack = errcallback.previous;
1155 : }
1156 :
1157 : bool
1158 296 : filter_prepare_cb_wrapper(LogicalDecodingContext *ctx, TransactionId xid,
1159 : const char *gid)
1160 : {
1161 : LogicalErrorCallbackState state;
1162 : ErrorContextCallback errcallback;
1163 : bool ret;
1164 :
1165 : Assert(!ctx->fast_forward);
1166 :
1167 : /* Push callback + info on the error context stack */
1168 296 : state.ctx = ctx;
1169 296 : state.callback_name = "filter_prepare";
1170 296 : state.report_location = InvalidXLogRecPtr;
1171 296 : errcallback.callback = output_plugin_error_callback;
1172 296 : errcallback.arg = &state;
1173 296 : errcallback.previous = error_context_stack;
1174 296 : error_context_stack = &errcallback;
1175 :
1176 : /* set output state */
1177 296 : ctx->accept_writes = false;
1178 296 : ctx->end_xact = false;
1179 :
1180 : /* do the actual work: call callback */
1181 296 : ret = ctx->callbacks.filter_prepare_cb(ctx, xid, gid);
1182 :
1183 : /* Pop the error context stack */
1184 296 : error_context_stack = errcallback.previous;
1185 :
1186 296 : return ret;
1187 : }
1188 :
1189 : bool
1190 3207182 : filter_by_origin_cb_wrapper(LogicalDecodingContext *ctx, RepOriginId origin_id)
1191 : {
1192 : LogicalErrorCallbackState state;
1193 : ErrorContextCallback errcallback;
1194 : bool ret;
1195 :
1196 : Assert(!ctx->fast_forward);
1197 :
1198 : /* Push callback + info on the error context stack */
1199 3207182 : state.ctx = ctx;
1200 3207182 : state.callback_name = "filter_by_origin";
1201 3207182 : state.report_location = InvalidXLogRecPtr;
1202 3207182 : errcallback.callback = output_plugin_error_callback;
1203 3207182 : errcallback.arg = &state;
1204 3207182 : errcallback.previous = error_context_stack;
1205 3207182 : error_context_stack = &errcallback;
1206 :
1207 : /* set output state */
1208 3207182 : ctx->accept_writes = false;
1209 3207182 : ctx->end_xact = false;
1210 :
1211 : /* do the actual work: call callback */
1212 3207182 : ret = ctx->callbacks.filter_by_origin_cb(ctx, origin_id);
1213 :
1214 : /* Pop the error context stack */
1215 3207182 : error_context_stack = errcallback.previous;
1216 :
1217 3207182 : return ret;
1218 : }
1219 :
1220 : static void
1221 32 : message_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1222 : XLogRecPtr message_lsn, bool transactional,
1223 : const char *prefix, Size message_size, const char *message)
1224 : {
1225 32 : LogicalDecodingContext *ctx = cache->private_data;
1226 : LogicalErrorCallbackState state;
1227 : ErrorContextCallback errcallback;
1228 :
1229 : Assert(!ctx->fast_forward);
1230 :
1231 32 : if (ctx->callbacks.message_cb == NULL)
1232 0 : return;
1233 :
1234 : /* Push callback + info on the error context stack */
1235 32 : state.ctx = ctx;
1236 32 : state.callback_name = "message";
1237 32 : state.report_location = message_lsn;
1238 32 : errcallback.callback = output_plugin_error_callback;
1239 32 : errcallback.arg = &state;
1240 32 : errcallback.previous = error_context_stack;
1241 32 : error_context_stack = &errcallback;
1242 :
1243 : /* set output state */
1244 32 : ctx->accept_writes = true;
1245 32 : ctx->write_xid = txn != NULL ? txn->xid : InvalidTransactionId;
1246 32 : ctx->write_location = message_lsn;
1247 32 : ctx->end_xact = false;
1248 :
1249 : /* do the actual work: call callback */
1250 32 : ctx->callbacks.message_cb(ctx, txn, message_lsn, transactional, prefix,
1251 : message_size, message);
1252 :
1253 : /* Pop the error context stack */
1254 32 : error_context_stack = errcallback.previous;
1255 : }
1256 :
1257 : static void
1258 1336 : stream_start_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1259 : XLogRecPtr first_lsn)
1260 : {
1261 1336 : LogicalDecodingContext *ctx = cache->private_data;
1262 : LogicalErrorCallbackState state;
1263 : ErrorContextCallback errcallback;
1264 :
1265 : Assert(!ctx->fast_forward);
1266 :
1267 : /* We're only supposed to call this when streaming is supported. */
1268 : Assert(ctx->streaming);
1269 :
1270 : /* Push callback + info on the error context stack */
1271 1336 : state.ctx = ctx;
1272 1336 : state.callback_name = "stream_start";
1273 1336 : state.report_location = first_lsn;
1274 1336 : errcallback.callback = output_plugin_error_callback;
1275 1336 : errcallback.arg = &state;
1276 1336 : errcallback.previous = error_context_stack;
1277 1336 : error_context_stack = &errcallback;
1278 :
1279 : /* set output state */
1280 1336 : ctx->accept_writes = true;
1281 1336 : ctx->write_xid = txn->xid;
1282 :
1283 : /*
1284 : * Report this message's lsn so replies from clients can give an
1285 : * up-to-date answer. This won't ever be enough (and shouldn't be!) to
1286 : * confirm receipt of this transaction, but it might allow another
1287 : * transaction's commit to be confirmed with one message.
1288 : */
1289 1336 : ctx->write_location = first_lsn;
1290 :
1291 1336 : ctx->end_xact = false;
1292 :
1293 : /* in streaming mode, stream_start_cb is required */
1294 1336 : if (ctx->callbacks.stream_start_cb == NULL)
1295 0 : ereport(ERROR,
1296 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1297 : errmsg("logical streaming requires a %s callback",
1298 : "stream_start_cb")));
1299 :
1300 1336 : ctx->callbacks.stream_start_cb(ctx, txn);
1301 :
1302 : /* Pop the error context stack */
1303 1336 : error_context_stack = errcallback.previous;
1304 1336 : }
1305 :
1306 : static void
1307 1336 : stream_stop_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1308 : XLogRecPtr last_lsn)
1309 : {
1310 1336 : LogicalDecodingContext *ctx = cache->private_data;
1311 : LogicalErrorCallbackState state;
1312 : ErrorContextCallback errcallback;
1313 :
1314 : Assert(!ctx->fast_forward);
1315 :
1316 : /* We're only supposed to call this when streaming is supported. */
1317 : Assert(ctx->streaming);
1318 :
1319 : /* Push callback + info on the error context stack */
1320 1336 : state.ctx = ctx;
1321 1336 : state.callback_name = "stream_stop";
1322 1336 : state.report_location = last_lsn;
1323 1336 : errcallback.callback = output_plugin_error_callback;
1324 1336 : errcallback.arg = &state;
1325 1336 : errcallback.previous = error_context_stack;
1326 1336 : error_context_stack = &errcallback;
1327 :
1328 : /* set output state */
1329 1336 : ctx->accept_writes = true;
1330 1336 : ctx->write_xid = txn->xid;
1331 :
1332 : /*
1333 : * Report this message's lsn so replies from clients can give an
1334 : * up-to-date answer. This won't ever be enough (and shouldn't be!) to
1335 : * confirm receipt of this transaction, but it might allow another
1336 : * transaction's commit to be confirmed with one message.
1337 : */
1338 1336 : ctx->write_location = last_lsn;
1339 :
1340 1336 : ctx->end_xact = false;
1341 :
1342 : /* in streaming mode, stream_stop_cb is required */
1343 1336 : if (ctx->callbacks.stream_stop_cb == NULL)
1344 0 : ereport(ERROR,
1345 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1346 : errmsg("logical streaming requires a %s callback",
1347 : "stream_stop_cb")));
1348 :
1349 1336 : ctx->callbacks.stream_stop_cb(ctx, txn);
1350 :
1351 : /* Pop the error context stack */
1352 1336 : error_context_stack = errcallback.previous;
1353 1336 : }
1354 :
1355 : static void
1356 60 : stream_abort_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1357 : XLogRecPtr abort_lsn)
1358 : {
1359 60 : LogicalDecodingContext *ctx = cache->private_data;
1360 : LogicalErrorCallbackState state;
1361 : ErrorContextCallback errcallback;
1362 :
1363 : Assert(!ctx->fast_forward);
1364 :
1365 : /* We're only supposed to call this when streaming is supported. */
1366 : Assert(ctx->streaming);
1367 :
1368 : /* Push callback + info on the error context stack */
1369 60 : state.ctx = ctx;
1370 60 : state.callback_name = "stream_abort";
1371 60 : state.report_location = abort_lsn;
1372 60 : errcallback.callback = output_plugin_error_callback;
1373 60 : errcallback.arg = &state;
1374 60 : errcallback.previous = error_context_stack;
1375 60 : error_context_stack = &errcallback;
1376 :
1377 : /* set output state */
1378 60 : ctx->accept_writes = true;
1379 60 : ctx->write_xid = txn->xid;
1380 60 : ctx->write_location = abort_lsn;
1381 60 : ctx->end_xact = true;
1382 :
1383 : /* in streaming mode, stream_abort_cb is required */
1384 60 : if (ctx->callbacks.stream_abort_cb == NULL)
1385 0 : ereport(ERROR,
1386 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1387 : errmsg("logical streaming requires a %s callback",
1388 : "stream_abort_cb")));
1389 :
1390 60 : ctx->callbacks.stream_abort_cb(ctx, txn, abort_lsn);
1391 :
1392 : /* Pop the error context stack */
1393 60 : error_context_stack = errcallback.previous;
1394 60 : }
1395 :
1396 : static void
1397 28 : stream_prepare_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1398 : XLogRecPtr prepare_lsn)
1399 : {
1400 28 : LogicalDecodingContext *ctx = cache->private_data;
1401 : LogicalErrorCallbackState state;
1402 : ErrorContextCallback errcallback;
1403 :
1404 : Assert(!ctx->fast_forward);
1405 :
1406 : /*
1407 : * We're only supposed to call this when streaming and two-phase commits
1408 : * are supported.
1409 : */
1410 : Assert(ctx->streaming);
1411 : Assert(ctx->twophase);
1412 :
1413 : /* Push callback + info on the error context stack */
1414 28 : state.ctx = ctx;
1415 28 : state.callback_name = "stream_prepare";
1416 28 : state.report_location = txn->final_lsn;
1417 28 : errcallback.callback = output_plugin_error_callback;
1418 28 : errcallback.arg = &state;
1419 28 : errcallback.previous = error_context_stack;
1420 28 : error_context_stack = &errcallback;
1421 :
1422 : /* set output state */
1423 28 : ctx->accept_writes = true;
1424 28 : ctx->write_xid = txn->xid;
1425 28 : ctx->write_location = txn->end_lsn;
1426 28 : ctx->end_xact = true;
1427 :
1428 : /* in streaming mode with two-phase commits, stream_prepare_cb is required */
1429 28 : if (ctx->callbacks.stream_prepare_cb == NULL)
1430 0 : ereport(ERROR,
1431 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1432 : errmsg("logical streaming at prepare time requires a %s callback",
1433 : "stream_prepare_cb")));
1434 :
1435 28 : ctx->callbacks.stream_prepare_cb(ctx, txn, prepare_lsn);
1436 :
1437 : /* Pop the error context stack */
1438 28 : error_context_stack = errcallback.previous;
1439 28 : }
1440 :
1441 : static void
1442 100 : stream_commit_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1443 : XLogRecPtr commit_lsn)
1444 : {
1445 100 : LogicalDecodingContext *ctx = cache->private_data;
1446 : LogicalErrorCallbackState state;
1447 : ErrorContextCallback errcallback;
1448 :
1449 : Assert(!ctx->fast_forward);
1450 :
1451 : /* We're only supposed to call this when streaming is supported. */
1452 : Assert(ctx->streaming);
1453 :
1454 : /* Push callback + info on the error context stack */
1455 100 : state.ctx = ctx;
1456 100 : state.callback_name = "stream_commit";
1457 100 : state.report_location = txn->final_lsn;
1458 100 : errcallback.callback = output_plugin_error_callback;
1459 100 : errcallback.arg = &state;
1460 100 : errcallback.previous = error_context_stack;
1461 100 : error_context_stack = &errcallback;
1462 :
1463 : /* set output state */
1464 100 : ctx->accept_writes = true;
1465 100 : ctx->write_xid = txn->xid;
1466 100 : ctx->write_location = txn->end_lsn;
1467 100 : ctx->end_xact = true;
1468 :
1469 : /* in streaming mode, stream_commit_cb is required */
1470 100 : if (ctx->callbacks.stream_commit_cb == NULL)
1471 0 : ereport(ERROR,
1472 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1473 : errmsg("logical streaming requires a %s callback",
1474 : "stream_commit_cb")));
1475 :
1476 100 : ctx->callbacks.stream_commit_cb(ctx, txn, commit_lsn);
1477 :
1478 : /* Pop the error context stack */
1479 100 : error_context_stack = errcallback.previous;
1480 100 : }
1481 :
1482 : static void
1483 351976 : stream_change_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1484 : Relation relation, ReorderBufferChange *change)
1485 : {
1486 351976 : LogicalDecodingContext *ctx = cache->private_data;
1487 : LogicalErrorCallbackState state;
1488 : ErrorContextCallback errcallback;
1489 :
1490 : Assert(!ctx->fast_forward);
1491 :
1492 : /* We're only supposed to call this when streaming is supported. */
1493 : Assert(ctx->streaming);
1494 :
1495 : /* Push callback + info on the error context stack */
1496 351976 : state.ctx = ctx;
1497 351976 : state.callback_name = "stream_change";
1498 351976 : state.report_location = change->lsn;
1499 351976 : errcallback.callback = output_plugin_error_callback;
1500 351976 : errcallback.arg = &state;
1501 351976 : errcallback.previous = error_context_stack;
1502 351976 : error_context_stack = &errcallback;
1503 :
1504 : /* set output state */
1505 351976 : ctx->accept_writes = true;
1506 351976 : ctx->write_xid = txn->xid;
1507 :
1508 : /*
1509 : * Report this change's lsn so replies from clients can give an up-to-date
1510 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1511 : * receipt of this transaction, but it might allow another transaction's
1512 : * commit to be confirmed with one message.
1513 : */
1514 351976 : ctx->write_location = change->lsn;
1515 :
1516 351976 : ctx->end_xact = false;
1517 :
1518 : /* in streaming mode, stream_change_cb is required */
1519 351976 : if (ctx->callbacks.stream_change_cb == NULL)
1520 0 : ereport(ERROR,
1521 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1522 : errmsg("logical streaming requires a %s callback",
1523 : "stream_change_cb")));
1524 :
1525 351976 : ctx->callbacks.stream_change_cb(ctx, txn, relation, change);
1526 :
1527 : /* Pop the error context stack */
1528 351976 : error_context_stack = errcallback.previous;
1529 351976 : }
1530 :
1531 : static void
1532 6 : stream_message_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1533 : XLogRecPtr message_lsn, bool transactional,
1534 : const char *prefix, Size message_size, const char *message)
1535 : {
1536 6 : LogicalDecodingContext *ctx = cache->private_data;
1537 : LogicalErrorCallbackState state;
1538 : ErrorContextCallback errcallback;
1539 :
1540 : Assert(!ctx->fast_forward);
1541 :
1542 : /* We're only supposed to call this when streaming is supported. */
1543 : Assert(ctx->streaming);
1544 :
1545 : /* this callback is optional */
1546 6 : if (ctx->callbacks.stream_message_cb == NULL)
1547 0 : return;
1548 :
1549 : /* Push callback + info on the error context stack */
1550 6 : state.ctx = ctx;
1551 6 : state.callback_name = "stream_message";
1552 6 : state.report_location = message_lsn;
1553 6 : errcallback.callback = output_plugin_error_callback;
1554 6 : errcallback.arg = &state;
1555 6 : errcallback.previous = error_context_stack;
1556 6 : error_context_stack = &errcallback;
1557 :
1558 : /* set output state */
1559 6 : ctx->accept_writes = true;
1560 6 : ctx->write_xid = txn != NULL ? txn->xid : InvalidTransactionId;
1561 6 : ctx->write_location = message_lsn;
1562 6 : ctx->end_xact = false;
1563 :
1564 : /* do the actual work: call callback */
1565 6 : ctx->callbacks.stream_message_cb(ctx, txn, message_lsn, transactional, prefix,
1566 : message_size, message);
1567 :
1568 : /* Pop the error context stack */
1569 6 : error_context_stack = errcallback.previous;
1570 : }
1571 :
1572 : static void
1573 0 : stream_truncate_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1574 : int nrelations, Relation relations[],
1575 : ReorderBufferChange *change)
1576 : {
1577 0 : LogicalDecodingContext *ctx = cache->private_data;
1578 : LogicalErrorCallbackState state;
1579 : ErrorContextCallback errcallback;
1580 :
1581 : Assert(!ctx->fast_forward);
1582 :
1583 : /* We're only supposed to call this when streaming is supported. */
1584 : Assert(ctx->streaming);
1585 :
1586 : /* this callback is optional */
1587 0 : if (!ctx->callbacks.stream_truncate_cb)
1588 0 : return;
1589 :
1590 : /* Push callback + info on the error context stack */
1591 0 : state.ctx = ctx;
1592 0 : state.callback_name = "stream_truncate";
1593 0 : state.report_location = change->lsn;
1594 0 : errcallback.callback = output_plugin_error_callback;
1595 0 : errcallback.arg = &state;
1596 0 : errcallback.previous = error_context_stack;
1597 0 : error_context_stack = &errcallback;
1598 :
1599 : /* set output state */
1600 0 : ctx->accept_writes = true;
1601 0 : ctx->write_xid = txn->xid;
1602 :
1603 : /*
1604 : * Report this change's lsn so replies from clients can give an up-to-date
1605 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1606 : * receipt of this transaction, but it might allow another transaction's
1607 : * commit to be confirmed with one message.
1608 : */
1609 0 : ctx->write_location = change->lsn;
1610 :
1611 0 : ctx->end_xact = false;
1612 :
1613 0 : ctx->callbacks.stream_truncate_cb(ctx, txn, nrelations, relations, change);
1614 :
1615 : /* Pop the error context stack */
1616 0 : error_context_stack = errcallback.previous;
1617 : }
1618 :
1619 : static void
1620 6258 : update_progress_txn_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1621 : XLogRecPtr lsn)
1622 : {
1623 6258 : LogicalDecodingContext *ctx = cache->private_data;
1624 : LogicalErrorCallbackState state;
1625 : ErrorContextCallback errcallback;
1626 :
1627 : Assert(!ctx->fast_forward);
1628 :
1629 : /* Push callback + info on the error context stack */
1630 6258 : state.ctx = ctx;
1631 6258 : state.callback_name = "update_progress_txn";
1632 6258 : state.report_location = lsn;
1633 6258 : errcallback.callback = output_plugin_error_callback;
1634 6258 : errcallback.arg = &state;
1635 6258 : errcallback.previous = error_context_stack;
1636 6258 : error_context_stack = &errcallback;
1637 :
1638 : /* set output state */
1639 6258 : ctx->accept_writes = false;
1640 6258 : ctx->write_xid = txn->xid;
1641 :
1642 : /*
1643 : * Report this change's lsn so replies from clients can give an up-to-date
1644 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1645 : * receipt of this transaction, but it might allow another transaction's
1646 : * commit to be confirmed with one message.
1647 : */
1648 6258 : ctx->write_location = lsn;
1649 :
1650 6258 : ctx->end_xact = false;
1651 :
1652 6258 : OutputPluginUpdateProgress(ctx, false);
1653 :
1654 : /* Pop the error context stack */
1655 6258 : error_context_stack = errcallback.previous;
1656 6258 : }
1657 :
1658 : /*
1659 : * Set the required catalog xmin horizon for historic snapshots in the current
1660 : * replication slot.
1661 : *
1662 : * Note that in the most cases, we won't be able to immediately use the xmin
1663 : * to increase the xmin horizon: we need to wait till the client has confirmed
1664 : * receiving current_lsn with LogicalConfirmReceivedLocation().
1665 : */
1666 : void
1667 848 : LogicalIncreaseXminForSlot(XLogRecPtr current_lsn, TransactionId xmin)
1668 : {
1669 848 : bool updated_xmin = false;
1670 : ReplicationSlot *slot;
1671 848 : bool got_new_xmin = false;
1672 :
1673 848 : slot = MyReplicationSlot;
1674 :
1675 : Assert(slot != NULL);
1676 :
1677 848 : SpinLockAcquire(&slot->mutex);
1678 :
1679 : /*
1680 : * don't overwrite if we already have a newer xmin. This can happen if we
1681 : * restart decoding in a slot.
1682 : */
1683 848 : if (TransactionIdPrecedesOrEquals(xmin, slot->data.catalog_xmin))
1684 : {
1685 : }
1686 :
1687 : /*
1688 : * If the client has already confirmed up to this lsn, we directly can
1689 : * mark this as accepted. This can happen if we restart decoding in a
1690 : * slot.
1691 : */
1692 214 : else if (current_lsn <= slot->data.confirmed_flush)
1693 : {
1694 108 : slot->candidate_catalog_xmin = xmin;
1695 108 : slot->candidate_xmin_lsn = current_lsn;
1696 :
1697 : /* our candidate can directly be used */
1698 108 : updated_xmin = true;
1699 : }
1700 :
1701 : /*
1702 : * Only increase if the previous values have been applied, otherwise we
1703 : * might never end up updating if the receiver acks too slowly.
1704 : */
1705 106 : else if (!XLogRecPtrIsValid(slot->candidate_xmin_lsn))
1706 : {
1707 44 : slot->candidate_catalog_xmin = xmin;
1708 44 : slot->candidate_xmin_lsn = current_lsn;
1709 :
1710 : /*
1711 : * Log new xmin at an appropriate log level after releasing the
1712 : * spinlock.
1713 : */
1714 44 : got_new_xmin = true;
1715 : }
1716 848 : SpinLockRelease(&slot->mutex);
1717 :
1718 848 : if (got_new_xmin)
1719 44 : elog(DEBUG1, "got new catalog xmin %u at %X/%08X", xmin,
1720 : LSN_FORMAT_ARGS(current_lsn));
1721 :
1722 : /* candidate already valid with the current flush position, apply */
1723 848 : if (updated_xmin)
1724 108 : LogicalConfirmReceivedLocation(slot->data.confirmed_flush);
1725 848 : }
1726 :
1727 : /*
1728 : * Mark the minimal LSN (restart_lsn) we need to read to replay all
1729 : * transactions that have not yet committed at current_lsn.
1730 : *
1731 : * Just like LogicalIncreaseXminForSlot this only takes effect when the
1732 : * client has confirmed to have received current_lsn.
1733 : */
1734 : void
1735 744 : LogicalIncreaseRestartDecodingForSlot(XLogRecPtr current_lsn, XLogRecPtr restart_lsn)
1736 : {
1737 744 : bool updated_lsn = false;
1738 : ReplicationSlot *slot;
1739 :
1740 744 : slot = MyReplicationSlot;
1741 :
1742 : Assert(slot != NULL);
1743 : Assert(XLogRecPtrIsValid(restart_lsn));
1744 : Assert(XLogRecPtrIsValid(current_lsn));
1745 :
1746 744 : SpinLockAcquire(&slot->mutex);
1747 :
1748 : /* don't overwrite if have a newer restart lsn */
1749 744 : if (restart_lsn <= slot->data.restart_lsn)
1750 : {
1751 20 : SpinLockRelease(&slot->mutex);
1752 : }
1753 :
1754 : /*
1755 : * We might have already flushed far enough to directly accept this lsn,
1756 : * in this case there is no need to check for existing candidate LSNs
1757 : */
1758 724 : else if (current_lsn <= slot->data.confirmed_flush)
1759 : {
1760 562 : slot->candidate_restart_valid = current_lsn;
1761 562 : slot->candidate_restart_lsn = restart_lsn;
1762 562 : SpinLockRelease(&slot->mutex);
1763 :
1764 : /* our candidate can directly be used */
1765 562 : updated_lsn = true;
1766 : }
1767 :
1768 : /*
1769 : * Only increase if the previous values have been applied, otherwise we
1770 : * might never end up updating if the receiver acks too slowly. A missed
1771 : * value here will just cause some extra effort after reconnecting.
1772 : */
1773 162 : else if (!XLogRecPtrIsValid(slot->candidate_restart_valid))
1774 : {
1775 80 : slot->candidate_restart_valid = current_lsn;
1776 80 : slot->candidate_restart_lsn = restart_lsn;
1777 80 : SpinLockRelease(&slot->mutex);
1778 :
1779 80 : elog(DEBUG1, "got new restart lsn %X/%08X at %X/%08X",
1780 : LSN_FORMAT_ARGS(restart_lsn),
1781 : LSN_FORMAT_ARGS(current_lsn));
1782 : }
1783 : else
1784 : {
1785 : XLogRecPtr candidate_restart_lsn;
1786 : XLogRecPtr candidate_restart_valid;
1787 : XLogRecPtr confirmed_flush;
1788 :
1789 82 : candidate_restart_lsn = slot->candidate_restart_lsn;
1790 82 : candidate_restart_valid = slot->candidate_restart_valid;
1791 82 : confirmed_flush = slot->data.confirmed_flush;
1792 82 : SpinLockRelease(&slot->mutex);
1793 :
1794 82 : 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",
1795 : LSN_FORMAT_ARGS(restart_lsn),
1796 : LSN_FORMAT_ARGS(current_lsn),
1797 : LSN_FORMAT_ARGS(candidate_restart_lsn),
1798 : LSN_FORMAT_ARGS(candidate_restart_valid),
1799 : LSN_FORMAT_ARGS(confirmed_flush));
1800 : }
1801 :
1802 : /* candidates are already valid with the current flush position, apply */
1803 744 : if (updated_lsn)
1804 562 : LogicalConfirmReceivedLocation(slot->data.confirmed_flush);
1805 744 : }
1806 :
1807 : /*
1808 : * Handle a consumer's confirmation having received all changes up to lsn.
1809 : */
1810 : void
1811 103900 : LogicalConfirmReceivedLocation(XLogRecPtr lsn)
1812 : {
1813 : Assert(XLogRecPtrIsValid(lsn));
1814 :
1815 : /* Do an unlocked check for candidate_lsn first. */
1816 103900 : if (XLogRecPtrIsValid(MyReplicationSlot->candidate_xmin_lsn) ||
1817 103702 : XLogRecPtrIsValid(MyReplicationSlot->candidate_restart_valid))
1818 798 : {
1819 798 : bool updated_xmin = false;
1820 798 : bool updated_restart = false;
1821 : XLogRecPtr restart_lsn pg_attribute_unused();
1822 :
1823 798 : SpinLockAcquire(&MyReplicationSlot->mutex);
1824 :
1825 : /* remember the old restart lsn */
1826 798 : restart_lsn = MyReplicationSlot->data.restart_lsn;
1827 :
1828 : /*
1829 : * Prevent moving the confirmed_flush backwards, as this could lead to
1830 : * data duplication issues caused by replicating already replicated
1831 : * changes.
1832 : *
1833 : * This can happen when a client acknowledges an LSN it doesn't have
1834 : * to do anything for, and thus didn't store persistently. After a
1835 : * restart, the client can send the prior LSN that it stored
1836 : * persistently as an acknowledgement, but we need to ignore such an
1837 : * LSN. See similar case handling in CreateDecodingContext.
1838 : */
1839 798 : if (lsn > MyReplicationSlot->data.confirmed_flush)
1840 72 : MyReplicationSlot->data.confirmed_flush = lsn;
1841 :
1842 : /* if we're past the location required for bumping xmin, do so */
1843 798 : if (XLogRecPtrIsValid(MyReplicationSlot->candidate_xmin_lsn) &&
1844 198 : MyReplicationSlot->candidate_xmin_lsn <= lsn)
1845 : {
1846 : /*
1847 : * We have to write the changed xmin to disk *before* we change
1848 : * the in-memory value, otherwise after a crash we wouldn't know
1849 : * that some catalog tuples might have been removed already.
1850 : *
1851 : * Ensure that by first writing to ->xmin and only update
1852 : * ->effective_xmin once the new state is synced to disk. After a
1853 : * crash ->effective_xmin is set to ->xmin.
1854 : */
1855 140 : if (TransactionIdIsValid(MyReplicationSlot->candidate_catalog_xmin) &&
1856 140 : MyReplicationSlot->data.catalog_xmin != MyReplicationSlot->candidate_catalog_xmin)
1857 : {
1858 140 : MyReplicationSlot->data.catalog_xmin = MyReplicationSlot->candidate_catalog_xmin;
1859 140 : MyReplicationSlot->candidate_catalog_xmin = InvalidTransactionId;
1860 140 : MyReplicationSlot->candidate_xmin_lsn = InvalidXLogRecPtr;
1861 140 : updated_xmin = true;
1862 : }
1863 : }
1864 :
1865 798 : if (XLogRecPtrIsValid(MyReplicationSlot->candidate_restart_valid) &&
1866 690 : MyReplicationSlot->candidate_restart_valid <= lsn)
1867 : {
1868 : Assert(XLogRecPtrIsValid(MyReplicationSlot->candidate_restart_lsn));
1869 :
1870 632 : MyReplicationSlot->data.restart_lsn = MyReplicationSlot->candidate_restart_lsn;
1871 632 : MyReplicationSlot->candidate_restart_lsn = InvalidXLogRecPtr;
1872 632 : MyReplicationSlot->candidate_restart_valid = InvalidXLogRecPtr;
1873 632 : updated_restart = true;
1874 : }
1875 :
1876 798 : SpinLockRelease(&MyReplicationSlot->mutex);
1877 :
1878 : /* first write new xmin to disk, so we know what's up after a crash */
1879 798 : if (updated_xmin || updated_restart)
1880 : {
1881 : #ifdef USE_INJECTION_POINTS
1882 : XLogSegNo seg1,
1883 : seg2;
1884 :
1885 740 : XLByteToSeg(restart_lsn, seg1, wal_segment_size);
1886 740 : XLByteToSeg(MyReplicationSlot->data.restart_lsn, seg2, wal_segment_size);
1887 :
1888 : /* trigger injection point, but only if segment changes */
1889 740 : if (seg1 != seg2)
1890 12 : INJECTION_POINT("logical-replication-slot-advance-segment", NULL);
1891 : #endif
1892 :
1893 740 : ReplicationSlotMarkDirty();
1894 740 : ReplicationSlotSave();
1895 740 : elog(DEBUG1, "updated xmin: %u restart: %u", updated_xmin, updated_restart);
1896 : }
1897 :
1898 : /*
1899 : * Now the new xmin is safely on disk, we can let the global value
1900 : * advance. We do not take ProcArrayLock or similar since we only
1901 : * advance xmin here and there's not much harm done by a concurrent
1902 : * computation missing that.
1903 : */
1904 798 : if (updated_xmin)
1905 : {
1906 140 : SpinLockAcquire(&MyReplicationSlot->mutex);
1907 140 : MyReplicationSlot->effective_catalog_xmin = MyReplicationSlot->data.catalog_xmin;
1908 140 : SpinLockRelease(&MyReplicationSlot->mutex);
1909 :
1910 140 : ReplicationSlotsComputeRequiredXmin(false);
1911 140 : ReplicationSlotsComputeRequiredLSN();
1912 : }
1913 : }
1914 : else
1915 : {
1916 103102 : SpinLockAcquire(&MyReplicationSlot->mutex);
1917 :
1918 : /*
1919 : * Prevent moving the confirmed_flush backwards. See comments above
1920 : * for the details.
1921 : */
1922 103102 : if (lsn > MyReplicationSlot->data.confirmed_flush)
1923 101568 : MyReplicationSlot->data.confirmed_flush = lsn;
1924 :
1925 103102 : SpinLockRelease(&MyReplicationSlot->mutex);
1926 : }
1927 103900 : }
1928 :
1929 : /*
1930 : * Clear logical streaming state during (sub)transaction abort.
1931 : */
1932 : void
1933 61554 : ResetLogicalStreamingState(void)
1934 : {
1935 61554 : CheckXidAlive = InvalidTransactionId;
1936 61554 : bsysscan = false;
1937 61554 : }
1938 :
1939 : /*
1940 : * Report stats for a slot.
1941 : */
1942 : void
1943 13014 : UpdateDecodingStats(LogicalDecodingContext *ctx)
1944 : {
1945 13014 : ReorderBuffer *rb = ctx->reorder;
1946 : PgStat_StatReplSlotEntry repSlotStat;
1947 :
1948 : /* Nothing to do if we don't have any replication stats to be sent. */
1949 13014 : if (rb->spillBytes <= 0 && rb->streamBytes <= 0 && rb->totalBytes <= 0 &&
1950 692 : rb->memExceededCount <= 0)
1951 674 : return;
1952 :
1953 12340 : elog(DEBUG2, "UpdateDecodingStats: updating stats %p %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64,
1954 : rb,
1955 : rb->spillTxns,
1956 : rb->spillCount,
1957 : rb->spillBytes,
1958 : rb->streamTxns,
1959 : rb->streamCount,
1960 : rb->streamBytes,
1961 : rb->memExceededCount,
1962 : rb->totalTxns,
1963 : rb->totalBytes);
1964 :
1965 12340 : repSlotStat.spill_txns = rb->spillTxns;
1966 12340 : repSlotStat.spill_count = rb->spillCount;
1967 12340 : repSlotStat.spill_bytes = rb->spillBytes;
1968 12340 : repSlotStat.stream_txns = rb->streamTxns;
1969 12340 : repSlotStat.stream_count = rb->streamCount;
1970 12340 : repSlotStat.stream_bytes = rb->streamBytes;
1971 12340 : repSlotStat.mem_exceeded_count = rb->memExceededCount;
1972 12340 : repSlotStat.total_txns = rb->totalTxns;
1973 12340 : repSlotStat.total_bytes = rb->totalBytes;
1974 :
1975 12340 : pgstat_report_replslot(ctx->slot, &repSlotStat);
1976 :
1977 12340 : rb->spillTxns = 0;
1978 12340 : rb->spillCount = 0;
1979 12340 : rb->spillBytes = 0;
1980 12340 : rb->streamTxns = 0;
1981 12340 : rb->streamCount = 0;
1982 12340 : rb->streamBytes = 0;
1983 12340 : rb->memExceededCount = 0;
1984 12340 : rb->totalTxns = 0;
1985 12340 : rb->totalBytes = 0;
1986 : }
1987 :
1988 : /*
1989 : * Read up to the end of WAL starting from the decoding slot's restart_lsn.
1990 : * Return true if any meaningful/decodable WAL records are encountered,
1991 : * otherwise false.
1992 : */
1993 : bool
1994 10 : LogicalReplicationSlotHasPendingWal(XLogRecPtr end_of_wal)
1995 : {
1996 10 : bool has_pending_wal = false;
1997 :
1998 : Assert(MyReplicationSlot);
1999 :
2000 10 : PG_TRY();
2001 : {
2002 : LogicalDecodingContext *ctx;
2003 :
2004 : /*
2005 : * Create our decoding context in fast_forward mode, passing start_lsn
2006 : * as InvalidXLogRecPtr, so that we start processing from the slot's
2007 : * confirmed_flush.
2008 : */
2009 20 : ctx = CreateDecodingContext(InvalidXLogRecPtr,
2010 : NIL,
2011 : true, /* fast_forward */
2012 10 : XL_ROUTINE(.page_read = read_local_xlog_page,
2013 : .segment_open = wal_segment_open,
2014 : .segment_close = wal_segment_close),
2015 : NULL, NULL, NULL);
2016 :
2017 : /*
2018 : * Start reading at the slot's restart_lsn, which we know points to a
2019 : * valid record.
2020 : */
2021 10 : XLogBeginRead(ctx->reader, MyReplicationSlot->data.restart_lsn);
2022 :
2023 : /* Invalidate non-timetravel entries */
2024 10 : InvalidateSystemCaches();
2025 :
2026 : /* Loop until the end of WAL or some changes are processed */
2027 296 : while (!has_pending_wal && ctx->reader->EndRecPtr < end_of_wal)
2028 : {
2029 : XLogRecord *record;
2030 286 : char *errm = NULL;
2031 :
2032 286 : record = XLogReadRecord(ctx->reader, &errm);
2033 :
2034 286 : if (errm)
2035 0 : elog(ERROR, "could not find record for logical decoding: %s", errm);
2036 :
2037 286 : if (record != NULL)
2038 286 : LogicalDecodingProcessRecord(ctx, ctx->reader);
2039 :
2040 286 : has_pending_wal = ctx->processing_required;
2041 :
2042 286 : CHECK_FOR_INTERRUPTS();
2043 : }
2044 :
2045 : /* Clean up */
2046 10 : FreeDecodingContext(ctx);
2047 10 : InvalidateSystemCaches();
2048 : }
2049 0 : PG_CATCH();
2050 : {
2051 : /* clear all timetravel entries */
2052 0 : InvalidateSystemCaches();
2053 :
2054 0 : PG_RE_THROW();
2055 : }
2056 10 : PG_END_TRY();
2057 :
2058 10 : return has_pending_wal;
2059 : }
2060 :
2061 : /*
2062 : * Helper function for advancing our logical replication slot forward.
2063 : *
2064 : * The slot's restart_lsn is used as start point for reading records, while
2065 : * confirmed_flush is used as base point for the decoding context.
2066 : *
2067 : * We cannot just do LogicalConfirmReceivedLocation to update confirmed_flush,
2068 : * because we need to digest WAL to advance restart_lsn allowing to recycle
2069 : * WAL and removal of old catalog tuples. As decoding is done in fast_forward
2070 : * mode, no changes are generated anyway.
2071 : *
2072 : * *found_consistent_snapshot will be true if the initial decoding snapshot has
2073 : * been built; Otherwise, it will be false.
2074 : */
2075 : XLogRecPtr
2076 36 : LogicalSlotAdvanceAndCheckSnapState(XLogRecPtr moveto,
2077 : bool *found_consistent_snapshot)
2078 : {
2079 : LogicalDecodingContext *ctx;
2080 36 : ResourceOwner old_resowner PG_USED_FOR_ASSERTS_ONLY = CurrentResourceOwner;
2081 : XLogRecPtr retlsn;
2082 :
2083 : Assert(XLogRecPtrIsValid(moveto));
2084 :
2085 36 : if (found_consistent_snapshot)
2086 12 : *found_consistent_snapshot = false;
2087 :
2088 36 : PG_TRY();
2089 : {
2090 : /*
2091 : * Create our decoding context in fast_forward mode, passing start_lsn
2092 : * as InvalidXLogRecPtr, so that we start processing from my slot's
2093 : * confirmed_flush.
2094 : */
2095 72 : ctx = CreateDecodingContext(InvalidXLogRecPtr,
2096 : NIL,
2097 : true, /* fast_forward */
2098 36 : XL_ROUTINE(.page_read = read_local_xlog_page,
2099 : .segment_open = wal_segment_open,
2100 : .segment_close = wal_segment_close),
2101 : NULL, NULL, NULL);
2102 :
2103 : /*
2104 : * Wait for specified streaming replication standby servers (if any)
2105 : * to confirm receipt of WAL up to moveto lsn.
2106 : */
2107 36 : WaitForStandbyConfirmation(moveto);
2108 :
2109 : /*
2110 : * Start reading at the slot's restart_lsn, which we know to point to
2111 : * a valid record.
2112 : */
2113 36 : XLogBeginRead(ctx->reader, MyReplicationSlot->data.restart_lsn);
2114 :
2115 : /* invalidate non-timetravel entries */
2116 36 : InvalidateSystemCaches();
2117 :
2118 : /* Decode records until we reach the requested target */
2119 4274 : while (ctx->reader->EndRecPtr < moveto)
2120 : {
2121 4238 : char *errm = NULL;
2122 : XLogRecord *record;
2123 :
2124 : /*
2125 : * Read records. No changes are generated in fast_forward mode,
2126 : * but snapbuilder/slot statuses are updated properly.
2127 : */
2128 4238 : record = XLogReadRecord(ctx->reader, &errm);
2129 4238 : if (errm)
2130 0 : elog(ERROR, "could not find record while advancing replication slot: %s",
2131 : errm);
2132 :
2133 : /*
2134 : * Process the record. Storage-level changes are ignored in
2135 : * fast_forward mode, but other modules (such as snapbuilder)
2136 : * might still have critical updates to do.
2137 : */
2138 4238 : if (record)
2139 : {
2140 4238 : LogicalDecodingProcessRecord(ctx, ctx->reader);
2141 :
2142 : /*
2143 : * We used to have bugs where logical decoding would fail to
2144 : * preserve the resource owner. That's important here, so
2145 : * verify that that doesn't happen anymore. XXX this could be
2146 : * removed once it's been battle-tested.
2147 : */
2148 : Assert(CurrentResourceOwner == old_resowner);
2149 : }
2150 :
2151 4238 : CHECK_FOR_INTERRUPTS();
2152 : }
2153 :
2154 36 : if (found_consistent_snapshot && DecodingContextReady(ctx))
2155 12 : *found_consistent_snapshot = true;
2156 :
2157 36 : if (XLogRecPtrIsValid(ctx->reader->EndRecPtr))
2158 : {
2159 36 : LogicalConfirmReceivedLocation(moveto);
2160 :
2161 : /*
2162 : * If only the confirmed_flush LSN has changed the slot won't get
2163 : * marked as dirty by the above. Callers on the walsender
2164 : * interface are expected to keep track of their own progress and
2165 : * don't need it written out. But SQL-interface users cannot
2166 : * specify their own start positions and it's harder for them to
2167 : * keep track of their progress, so we should make more of an
2168 : * effort to save it for them.
2169 : *
2170 : * Dirty the slot so it is written out at the next checkpoint. The
2171 : * LSN position advanced to may still be lost on a crash but this
2172 : * makes the data consistent after a clean shutdown.
2173 : */
2174 36 : ReplicationSlotMarkDirty();
2175 : }
2176 :
2177 36 : retlsn = MyReplicationSlot->data.confirmed_flush;
2178 :
2179 : /* free context, call shutdown callback */
2180 36 : FreeDecodingContext(ctx);
2181 :
2182 36 : InvalidateSystemCaches();
2183 : }
2184 0 : PG_CATCH();
2185 : {
2186 : /* clear all timetravel entries */
2187 0 : InvalidateSystemCaches();
2188 :
2189 0 : PG_RE_THROW();
2190 : }
2191 36 : PG_END_TRY();
2192 :
2193 36 : return retlsn;
2194 : }
|
