Line data Source code
1 : /*-------------------------------------------------------------------------
2 : * logical.c
3 : * PostgreSQL logical decoding coordination
4 : *
5 : * Copyright (c) 2012-2025, 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 3012 : CheckLogicalDecodingRequirements(void)
112 : {
113 3012 : CheckSlotRequirements();
114 :
115 : /*
116 : * NB: Adding a new requirement likely means that RestoreSlotFromDisk()
117 : * needs the same check.
118 : */
119 :
120 3012 : if (wal_level < WAL_LEVEL_LOGICAL)
121 0 : ereport(ERROR,
122 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
123 : errmsg("logical decoding requires \"wal_level\" >= \"logical\"")));
124 :
125 3012 : if (MyDatabaseId == InvalidOid)
126 2 : ereport(ERROR,
127 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
128 : errmsg("logical decoding requires a database connection")));
129 :
130 3010 : if (RecoveryInProgress())
131 : {
132 : /*
133 : * This check may have race conditions, but whenever
134 : * XLOG_PARAMETER_CHANGE indicates that wal_level has changed, we
135 : * verify that there are no existing logical replication slots. And to
136 : * avoid races around creating a new slot,
137 : * CheckLogicalDecodingRequirements() is called once before creating
138 : * the slot, and once when logical decoding is initially starting up.
139 : */
140 138 : if (GetActiveWalLevelOnStandby() < WAL_LEVEL_LOGICAL)
141 2 : ereport(ERROR,
142 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
143 : errmsg("logical decoding on standby requires \"wal_level\" >= \"logical\" on the primary")));
144 : }
145 3008 : }
146 :
147 : /*
148 : * Helper function for CreateInitDecodingContext() and
149 : * CreateDecodingContext() performing common tasks.
150 : */
151 : static LogicalDecodingContext *
152 2108 : StartupDecodingContext(List *output_plugin_options,
153 : XLogRecPtr start_lsn,
154 : TransactionId xmin_horizon,
155 : bool need_full_snapshot,
156 : bool fast_forward,
157 : bool in_create,
158 : XLogReaderRoutine *xl_routine,
159 : LogicalOutputPluginWriterPrepareWrite prepare_write,
160 : LogicalOutputPluginWriterWrite do_write,
161 : LogicalOutputPluginWriterUpdateProgress update_progress)
162 : {
163 : ReplicationSlot *slot;
164 : MemoryContext context,
165 : old_context;
166 : LogicalDecodingContext *ctx;
167 :
168 : /* shorter lines... */
169 2108 : slot = MyReplicationSlot;
170 :
171 2108 : context = AllocSetContextCreate(CurrentMemoryContext,
172 : "Logical decoding context",
173 : ALLOCSET_DEFAULT_SIZES);
174 2108 : old_context = MemoryContextSwitchTo(context);
175 2108 : ctx = palloc0(sizeof(LogicalDecodingContext));
176 :
177 2108 : ctx->context = context;
178 :
179 : /*
180 : * (re-)load output plugins, so we detect a bad (removed) output plugin
181 : * now.
182 : */
183 2108 : if (!fast_forward)
184 2066 : LoadOutputPlugin(&ctx->callbacks, NameStr(slot->data.plugin));
185 :
186 : /*
187 : * Now that the slot's xmin has been set, we can announce ourselves as a
188 : * logical decoding backend which doesn't need to be checked individually
189 : * when computing the xmin horizon because the xmin is enforced via
190 : * replication slots.
191 : *
192 : * We can only do so if we're outside of a transaction (i.e. the case when
193 : * streaming changes via walsender), otherwise an already setup
194 : * snapshot/xid would end up being ignored. That's not a particularly
195 : * bothersome restriction since the SQL interface can't be used for
196 : * streaming anyway.
197 : */
198 2106 : if (!IsTransactionOrTransactionBlock())
199 : {
200 1024 : LWLockAcquire(ProcArrayLock, LW_EXCLUSIVE);
201 1024 : MyProc->statusFlags |= PROC_IN_LOGICAL_DECODING;
202 1024 : ProcGlobal->statusFlags[MyProc->pgxactoff] = MyProc->statusFlags;
203 1024 : LWLockRelease(ProcArrayLock);
204 : }
205 :
206 2106 : ctx->slot = slot;
207 :
208 2106 : ctx->reader = XLogReaderAllocate(wal_segment_size, NULL, xl_routine, ctx);
209 2106 : if (!ctx->reader)
210 0 : ereport(ERROR,
211 : (errcode(ERRCODE_OUT_OF_MEMORY),
212 : errmsg("out of memory"),
213 : errdetail("Failed while allocating a WAL reading processor.")));
214 :
215 2106 : ctx->reorder = ReorderBufferAllocate();
216 2106 : ctx->snapshot_builder =
217 2106 : AllocateSnapshotBuilder(ctx->reorder, xmin_horizon, start_lsn,
218 : need_full_snapshot, in_create, slot->data.two_phase_at);
219 :
220 2106 : ctx->reorder->private_data = ctx;
221 :
222 : /* wrap output plugin callbacks, so we can add error context information */
223 2106 : ctx->reorder->begin = begin_cb_wrapper;
224 2106 : ctx->reorder->apply_change = change_cb_wrapper;
225 2106 : ctx->reorder->apply_truncate = truncate_cb_wrapper;
226 2106 : ctx->reorder->commit = commit_cb_wrapper;
227 2106 : ctx->reorder->message = message_cb_wrapper;
228 :
229 : /*
230 : * To support streaming, we require start/stop/abort/commit/change
231 : * callbacks. The message and truncate callbacks are optional, similar to
232 : * regular output plugins. We however enable streaming when at least one
233 : * of the methods is enabled so that we can easily identify missing
234 : * methods.
235 : *
236 : * We decide it here, but only check it later in the wrappers.
237 : */
238 4254 : ctx->streaming = (ctx->callbacks.stream_start_cb != NULL) ||
239 42 : (ctx->callbacks.stream_stop_cb != NULL) ||
240 42 : (ctx->callbacks.stream_abort_cb != NULL) ||
241 42 : (ctx->callbacks.stream_commit_cb != NULL) ||
242 42 : (ctx->callbacks.stream_change_cb != NULL) ||
243 2190 : (ctx->callbacks.stream_message_cb != NULL) ||
244 42 : (ctx->callbacks.stream_truncate_cb != NULL);
245 :
246 : /*
247 : * streaming callbacks
248 : *
249 : * stream_message and stream_truncate callbacks are optional, so we do not
250 : * fail with ERROR when missing, but the wrappers simply do nothing. We
251 : * must set the ReorderBuffer callbacks to something, otherwise the calls
252 : * from there will crash (we don't want to move the checks there).
253 : */
254 2106 : ctx->reorder->stream_start = stream_start_cb_wrapper;
255 2106 : ctx->reorder->stream_stop = stream_stop_cb_wrapper;
256 2106 : ctx->reorder->stream_abort = stream_abort_cb_wrapper;
257 2106 : ctx->reorder->stream_prepare = stream_prepare_cb_wrapper;
258 2106 : ctx->reorder->stream_commit = stream_commit_cb_wrapper;
259 2106 : ctx->reorder->stream_change = stream_change_cb_wrapper;
260 2106 : ctx->reorder->stream_message = stream_message_cb_wrapper;
261 2106 : ctx->reorder->stream_truncate = stream_truncate_cb_wrapper;
262 :
263 :
264 : /*
265 : * To support two-phase logical decoding, we require
266 : * begin_prepare/prepare/commit-prepare/abort-prepare callbacks. The
267 : * filter_prepare callback is optional. We however enable two-phase
268 : * logical decoding when at least one of the methods is enabled so that we
269 : * can easily identify missing methods.
270 : *
271 : * We decide it here, but only check it later in the wrappers.
272 : */
273 4254 : ctx->twophase = (ctx->callbacks.begin_prepare_cb != NULL) ||
274 42 : (ctx->callbacks.prepare_cb != NULL) ||
275 42 : (ctx->callbacks.commit_prepared_cb != NULL) ||
276 42 : (ctx->callbacks.rollback_prepared_cb != NULL) ||
277 2190 : (ctx->callbacks.stream_prepare_cb != NULL) ||
278 42 : (ctx->callbacks.filter_prepare_cb != NULL);
279 :
280 : /*
281 : * Callback to support decoding at prepare time.
282 : */
283 2106 : ctx->reorder->begin_prepare = begin_prepare_cb_wrapper;
284 2106 : ctx->reorder->prepare = prepare_cb_wrapper;
285 2106 : ctx->reorder->commit_prepared = commit_prepared_cb_wrapper;
286 2106 : ctx->reorder->rollback_prepared = rollback_prepared_cb_wrapper;
287 :
288 : /*
289 : * Callback to support updating progress during sending data of a
290 : * transaction (and its subtransactions) to the output plugin.
291 : */
292 2106 : ctx->reorder->update_progress_txn = update_progress_txn_cb_wrapper;
293 :
294 2106 : ctx->out = makeStringInfo();
295 2106 : ctx->prepare_write = prepare_write;
296 2106 : ctx->write = do_write;
297 2106 : ctx->update_progress = update_progress;
298 :
299 2106 : ctx->output_plugin_options = output_plugin_options;
300 :
301 2106 : ctx->fast_forward = fast_forward;
302 :
303 2106 : MemoryContextSwitchTo(old_context);
304 :
305 2106 : return ctx;
306 : }
307 :
308 : /*
309 : * Create a new decoding context, for a new logical slot.
310 : *
311 : * plugin -- contains the name of the output plugin
312 : * output_plugin_options -- contains options passed to the output plugin
313 : * need_full_snapshot -- if true, must obtain a snapshot able to read all
314 : * tables; if false, one that can read only catalogs is acceptable.
315 : * restart_lsn -- if given as invalid, it's this routine's responsibility to
316 : * mark WAL as reserved by setting a convenient restart_lsn for the slot.
317 : * Otherwise, we set for decoding to start from the given LSN without
318 : * marking WAL reserved beforehand. In that scenario, it's up to the
319 : * caller to guarantee that WAL remains available.
320 : * xl_routine -- XLogReaderRoutine for underlying XLogReader
321 : * prepare_write, do_write, update_progress --
322 : * callbacks that perform the use-case dependent, actual, work.
323 : *
324 : * Needs to be called while in a memory context that's at least as long lived
325 : * as the decoding context because further memory contexts will be created
326 : * inside it.
327 : *
328 : * Returns an initialized decoding context after calling the output plugin's
329 : * startup function.
330 : */
331 : LogicalDecodingContext *
332 902 : CreateInitDecodingContext(const char *plugin,
333 : List *output_plugin_options,
334 : bool need_full_snapshot,
335 : XLogRecPtr restart_lsn,
336 : XLogReaderRoutine *xl_routine,
337 : LogicalOutputPluginWriterPrepareWrite prepare_write,
338 : LogicalOutputPluginWriterWrite do_write,
339 : LogicalOutputPluginWriterUpdateProgress update_progress)
340 : {
341 902 : TransactionId xmin_horizon = InvalidTransactionId;
342 : ReplicationSlot *slot;
343 : NameData plugin_name;
344 : LogicalDecodingContext *ctx;
345 : MemoryContext old_context;
346 :
347 : /*
348 : * On a standby, this check is also required while creating the slot.
349 : * Check the comments in the function.
350 : */
351 902 : CheckLogicalDecodingRequirements();
352 :
353 : /* shorter lines... */
354 902 : slot = MyReplicationSlot;
355 :
356 : /* first some sanity checks that are unlikely to be violated */
357 902 : if (slot == NULL)
358 0 : elog(ERROR, "cannot perform logical decoding without an acquired slot");
359 :
360 902 : if (plugin == NULL)
361 0 : elog(ERROR, "cannot initialize logical decoding without a specified plugin");
362 :
363 : /* Make sure the passed slot is suitable. These are user facing errors. */
364 902 : if (SlotIsPhysical(slot))
365 0 : ereport(ERROR,
366 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
367 : errmsg("cannot use physical replication slot for logical decoding")));
368 :
369 902 : if (slot->data.database != MyDatabaseId)
370 0 : ereport(ERROR,
371 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
372 : errmsg("replication slot \"%s\" was not created in this database",
373 : NameStr(slot->data.name))));
374 :
375 1534 : if (IsTransactionState() &&
376 632 : GetTopTransactionIdIfAny() != InvalidTransactionId)
377 4 : ereport(ERROR,
378 : (errcode(ERRCODE_ACTIVE_SQL_TRANSACTION),
379 : errmsg("cannot create logical replication slot in transaction that has performed writes")));
380 :
381 : /*
382 : * Register output plugin name with slot. We need the mutex to avoid
383 : * concurrent reading of a partially copied string. But we don't want any
384 : * complicated code while holding a spinlock, so do namestrcpy() outside.
385 : */
386 898 : namestrcpy(&plugin_name, plugin);
387 898 : SpinLockAcquire(&slot->mutex);
388 898 : slot->data.plugin = plugin_name;
389 898 : SpinLockRelease(&slot->mutex);
390 :
391 898 : if (XLogRecPtrIsInvalid(restart_lsn))
392 886 : ReplicationSlotReserveWal();
393 : else
394 : {
395 12 : SpinLockAcquire(&slot->mutex);
396 12 : slot->data.restart_lsn = restart_lsn;
397 12 : SpinLockRelease(&slot->mutex);
398 : }
399 :
400 : /* ----
401 : * This is a bit tricky: We need to determine a safe xmin horizon to start
402 : * decoding from, to avoid starting from a running xacts record referring
403 : * to xids whose rows have been vacuumed or pruned
404 : * already. GetOldestSafeDecodingTransactionId() returns such a value, but
405 : * without further interlock its return value might immediately be out of
406 : * date.
407 : *
408 : * So we have to acquire the ProcArrayLock to prevent computation of new
409 : * xmin horizons by other backends, get the safe decoding xid, and inform
410 : * the slot machinery about the new limit. Once that's done the
411 : * ProcArrayLock can be released as the slot machinery now is
412 : * protecting against vacuum.
413 : *
414 : * Note that, temporarily, the data, not just the catalog, xmin has to be
415 : * reserved if a data snapshot is to be exported. Otherwise the initial
416 : * data snapshot created here is not guaranteed to be valid. After that
417 : * the data xmin doesn't need to be managed anymore and the global xmin
418 : * should be recomputed. As we are fine with losing the pegged data xmin
419 : * after crash - no chance a snapshot would get exported anymore - we can
420 : * get away with just setting the slot's
421 : * effective_xmin. ReplicationSlotRelease will reset it again.
422 : *
423 : * ----
424 : */
425 898 : LWLockAcquire(ProcArrayLock, LW_EXCLUSIVE);
426 :
427 898 : xmin_horizon = GetOldestSafeDecodingTransactionId(!need_full_snapshot);
428 :
429 898 : SpinLockAcquire(&slot->mutex);
430 898 : slot->effective_catalog_xmin = xmin_horizon;
431 898 : slot->data.catalog_xmin = xmin_horizon;
432 898 : if (need_full_snapshot)
433 388 : slot->effective_xmin = xmin_horizon;
434 898 : SpinLockRelease(&slot->mutex);
435 :
436 898 : ReplicationSlotsComputeRequiredXmin(true);
437 :
438 898 : LWLockRelease(ProcArrayLock);
439 :
440 898 : ReplicationSlotMarkDirty();
441 898 : ReplicationSlotSave();
442 :
443 898 : ctx = StartupDecodingContext(NIL, restart_lsn, xmin_horizon,
444 : need_full_snapshot, false, true,
445 : xl_routine, prepare_write, do_write,
446 : update_progress);
447 :
448 : /* call output plugin initialization callback */
449 896 : old_context = MemoryContextSwitchTo(ctx->context);
450 896 : if (ctx->callbacks.startup_cb != NULL)
451 896 : startup_cb_wrapper(ctx, &ctx->options, true);
452 896 : MemoryContextSwitchTo(old_context);
453 :
454 : /*
455 : * We allow decoding of prepared transactions when the two_phase is
456 : * enabled at the time of slot creation, or when the two_phase option is
457 : * given at the streaming start, provided the plugin supports all the
458 : * callbacks for two-phase.
459 : */
460 896 : ctx->twophase &= slot->data.two_phase;
461 :
462 896 : ctx->reorder->output_rewrites = ctx->options.receive_rewrites;
463 :
464 896 : return ctx;
465 : }
466 :
467 : /*
468 : * Create a new decoding context, for a logical slot that has previously been
469 : * used already.
470 : *
471 : * start_lsn
472 : * The LSN at which to start decoding. If InvalidXLogRecPtr, restart
473 : * from the slot's confirmed_flush; otherwise, start from the specified
474 : * location (but move it forwards to confirmed_flush if it's older than
475 : * that, see below).
476 : *
477 : * output_plugin_options
478 : * options passed to the output plugin.
479 : *
480 : * fast_forward
481 : * bypass the generation of logical changes.
482 : *
483 : * xl_routine
484 : * XLogReaderRoutine used by underlying xlogreader
485 : *
486 : * prepare_write, do_write, update_progress
487 : * callbacks that have to be filled to perform the use-case dependent,
488 : * actual work.
489 : *
490 : * Needs to be called while in a memory context that's at least as long lived
491 : * as the decoding context because further memory contexts will be created
492 : * inside it.
493 : *
494 : * Returns an initialized decoding context after calling the output plugin's
495 : * startup function.
496 : */
497 : LogicalDecodingContext *
498 1220 : CreateDecodingContext(XLogRecPtr start_lsn,
499 : List *output_plugin_options,
500 : bool fast_forward,
501 : XLogReaderRoutine *xl_routine,
502 : LogicalOutputPluginWriterPrepareWrite prepare_write,
503 : LogicalOutputPluginWriterWrite do_write,
504 : LogicalOutputPluginWriterUpdateProgress update_progress)
505 : {
506 : LogicalDecodingContext *ctx;
507 : ReplicationSlot *slot;
508 : MemoryContext old_context;
509 :
510 : /* shorter lines... */
511 1220 : slot = MyReplicationSlot;
512 :
513 : /* first some sanity checks that are unlikely to be violated */
514 1220 : if (slot == NULL)
515 0 : elog(ERROR, "cannot perform logical decoding without an acquired slot");
516 :
517 : /* make sure the passed slot is suitable, these are user facing errors */
518 1220 : if (SlotIsPhysical(slot))
519 2 : ereport(ERROR,
520 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
521 : errmsg("cannot use physical replication slot for logical decoding")));
522 :
523 : /*
524 : * We need to access the system tables during decoding to build the
525 : * logical changes unless we are in fast_forward mode where no changes are
526 : * generated.
527 : */
528 1218 : if (slot->data.database != MyDatabaseId && !fast_forward)
529 6 : ereport(ERROR,
530 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
531 : errmsg("replication slot \"%s\" was not created in this database",
532 : NameStr(slot->data.name))));
533 :
534 : /*
535 : * The slots being synced from the primary can't be used for decoding as
536 : * they are used after failover. However, we do allow advancing the LSNs
537 : * during the synchronization of slots. See update_local_synced_slot.
538 : */
539 1212 : if (RecoveryInProgress() && slot->data.synced && !IsSyncingReplicationSlots())
540 2 : ereport(ERROR,
541 : errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
542 : errmsg("cannot use replication slot \"%s\" for logical decoding",
543 : NameStr(slot->data.name)),
544 : errdetail("This replication slot is being synchronized from the primary server."),
545 : errhint("Specify another replication slot."));
546 :
547 : /* slot must be valid to allow decoding */
548 : Assert(slot->data.invalidated == RS_INVAL_NONE);
549 : Assert(slot->data.restart_lsn != InvalidXLogRecPtr);
550 :
551 1210 : if (start_lsn == InvalidXLogRecPtr)
552 : {
553 : /* continue from last position */
554 726 : start_lsn = slot->data.confirmed_flush;
555 : }
556 484 : else if (start_lsn < slot->data.confirmed_flush)
557 : {
558 : /*
559 : * It might seem like we should error out in this case, but it's
560 : * pretty common for a client to acknowledge a LSN it doesn't have to
561 : * do anything for, and thus didn't store persistently, because the
562 : * xlog records didn't result in anything relevant for logical
563 : * decoding. Clients have to be able to do that to support synchronous
564 : * replication.
565 : *
566 : * Starting at a different LSN than requested might not catch certain
567 : * kinds of client errors; so the client may wish to check that
568 : * confirmed_flush_lsn matches its expectations.
569 : */
570 44 : elog(LOG, "%X/%X has been already streamed, forwarding to %X/%X",
571 : LSN_FORMAT_ARGS(start_lsn),
572 : LSN_FORMAT_ARGS(slot->data.confirmed_flush));
573 :
574 44 : start_lsn = slot->data.confirmed_flush;
575 : }
576 :
577 1210 : ctx = StartupDecodingContext(output_plugin_options,
578 : start_lsn, InvalidTransactionId, false,
579 : fast_forward, false, xl_routine, prepare_write,
580 : do_write, update_progress);
581 :
582 : /* call output plugin initialization callback */
583 1210 : old_context = MemoryContextSwitchTo(ctx->context);
584 1210 : if (ctx->callbacks.startup_cb != NULL)
585 1168 : startup_cb_wrapper(ctx, &ctx->options, false);
586 1204 : MemoryContextSwitchTo(old_context);
587 :
588 : /*
589 : * We allow decoding of prepared transactions when the two_phase is
590 : * enabled at the time of slot creation, or when the two_phase option is
591 : * given at the streaming start, provided the plugin supports all the
592 : * callbacks for two-phase.
593 : */
594 1204 : ctx->twophase &= (slot->data.two_phase || ctx->twophase_opt_given);
595 :
596 : /* Mark slot to allow two_phase decoding if not already marked */
597 1204 : if (ctx->twophase && !slot->data.two_phase)
598 : {
599 16 : SpinLockAcquire(&slot->mutex);
600 16 : slot->data.two_phase = true;
601 16 : slot->data.two_phase_at = start_lsn;
602 16 : SpinLockRelease(&slot->mutex);
603 16 : ReplicationSlotMarkDirty();
604 16 : ReplicationSlotSave();
605 16 : SnapBuildSetTwoPhaseAt(ctx->snapshot_builder, start_lsn);
606 : }
607 :
608 1204 : ctx->reorder->output_rewrites = ctx->options.receive_rewrites;
609 :
610 1204 : ereport(LOG,
611 : (errmsg("starting logical decoding for slot \"%s\"",
612 : NameStr(slot->data.name)),
613 : errdetail("Streaming transactions committing after %X/%X, reading WAL from %X/%X.",
614 : LSN_FORMAT_ARGS(slot->data.confirmed_flush),
615 : LSN_FORMAT_ARGS(slot->data.restart_lsn))));
616 :
617 1204 : return ctx;
618 : }
619 :
620 : /*
621 : * Returns true if a consistent initial decoding snapshot has been built.
622 : */
623 : bool
624 6974 : DecodingContextReady(LogicalDecodingContext *ctx)
625 : {
626 6974 : return SnapBuildCurrentState(ctx->snapshot_builder) == SNAPBUILD_CONSISTENT;
627 : }
628 :
629 : /*
630 : * Read from the decoding slot, until it is ready to start extracting changes.
631 : */
632 : void
633 884 : DecodingContextFindStartpoint(LogicalDecodingContext *ctx)
634 : {
635 884 : ReplicationSlot *slot = ctx->slot;
636 :
637 : /* Initialize from where to start reading WAL. */
638 884 : XLogBeginRead(ctx->reader, slot->data.restart_lsn);
639 :
640 884 : elog(DEBUG1, "searching for logical decoding starting point, starting at %X/%X",
641 : LSN_FORMAT_ARGS(slot->data.restart_lsn));
642 :
643 : /* Wait for a consistent starting point */
644 : for (;;)
645 6084 : {
646 : XLogRecord *record;
647 6968 : char *err = NULL;
648 :
649 : /* the read_page callback waits for new WAL */
650 6968 : record = XLogReadRecord(ctx->reader, &err);
651 6968 : if (err)
652 0 : elog(ERROR, "could not find logical decoding starting point: %s", err);
653 6968 : if (!record)
654 0 : elog(ERROR, "could not find logical decoding starting point");
655 :
656 6968 : LogicalDecodingProcessRecord(ctx, ctx->reader);
657 :
658 : /* only continue till we found a consistent spot */
659 6964 : if (DecodingContextReady(ctx))
660 880 : break;
661 :
662 6084 : CHECK_FOR_INTERRUPTS();
663 : }
664 :
665 880 : SpinLockAcquire(&slot->mutex);
666 880 : slot->data.confirmed_flush = ctx->reader->EndRecPtr;
667 880 : if (slot->data.two_phase)
668 18 : slot->data.two_phase_at = ctx->reader->EndRecPtr;
669 880 : SpinLockRelease(&slot->mutex);
670 880 : }
671 :
672 : /*
673 : * Free a previously allocated decoding context, invoking the shutdown
674 : * callback if necessary.
675 : */
676 : void
677 1718 : FreeDecodingContext(LogicalDecodingContext *ctx)
678 : {
679 1718 : if (ctx->callbacks.shutdown_cb != NULL)
680 1676 : shutdown_cb_wrapper(ctx);
681 :
682 1718 : ReorderBufferFree(ctx->reorder);
683 1718 : FreeSnapshotBuilder(ctx->snapshot_builder);
684 1718 : XLogReaderFree(ctx->reader);
685 1718 : MemoryContextDelete(ctx->context);
686 1718 : }
687 :
688 : /*
689 : * Prepare a write using the context's output routine.
690 : */
691 : void
692 672062 : OutputPluginPrepareWrite(struct LogicalDecodingContext *ctx, bool last_write)
693 : {
694 672062 : if (!ctx->accept_writes)
695 0 : elog(ERROR, "writes are only accepted in commit, begin and change callbacks");
696 :
697 672062 : ctx->prepare_write(ctx, ctx->write_location, ctx->write_xid, last_write);
698 672062 : ctx->prepared_write = true;
699 672062 : }
700 :
701 : /*
702 : * Perform a write using the context's output routine.
703 : */
704 : void
705 672062 : OutputPluginWrite(struct LogicalDecodingContext *ctx, bool last_write)
706 : {
707 672062 : if (!ctx->prepared_write)
708 0 : elog(ERROR, "OutputPluginPrepareWrite needs to be called before OutputPluginWrite");
709 :
710 672062 : ctx->write(ctx, ctx->write_location, ctx->write_xid, last_write);
711 672040 : ctx->prepared_write = false;
712 672040 : }
713 :
714 : /*
715 : * Update progress tracking (if supported).
716 : */
717 : void
718 7962 : OutputPluginUpdateProgress(struct LogicalDecodingContext *ctx,
719 : bool skipped_xact)
720 : {
721 7962 : if (!ctx->update_progress)
722 3180 : return;
723 :
724 4782 : ctx->update_progress(ctx, ctx->write_location, ctx->write_xid,
725 : skipped_xact);
726 : }
727 :
728 : /*
729 : * Load the output plugin, lookup its output plugin init function, and check
730 : * that it provides the required callbacks.
731 : */
732 : static void
733 2066 : LoadOutputPlugin(OutputPluginCallbacks *callbacks, const char *plugin)
734 : {
735 : LogicalOutputPluginInit plugin_init;
736 :
737 2064 : plugin_init = (LogicalOutputPluginInit)
738 2066 : load_external_function(plugin, "_PG_output_plugin_init", false, NULL);
739 :
740 2064 : if (plugin_init == NULL)
741 0 : elog(ERROR, "output plugins have to declare the _PG_output_plugin_init symbol");
742 :
743 : /* ask the output plugin to fill the callback struct */
744 2064 : plugin_init(callbacks);
745 :
746 2064 : if (callbacks->begin_cb == NULL)
747 0 : elog(ERROR, "output plugins have to register a begin callback");
748 2064 : if (callbacks->change_cb == NULL)
749 0 : elog(ERROR, "output plugins have to register a change callback");
750 2064 : if (callbacks->commit_cb == NULL)
751 0 : elog(ERROR, "output plugins have to register a commit callback");
752 2064 : }
753 :
754 : static void
755 36 : output_plugin_error_callback(void *arg)
756 : {
757 36 : LogicalErrorCallbackState *state = (LogicalErrorCallbackState *) arg;
758 :
759 : /* not all callbacks have an associated LSN */
760 36 : if (state->report_location != InvalidXLogRecPtr)
761 30 : errcontext("slot \"%s\", output plugin \"%s\", in the %s callback, associated LSN %X/%X",
762 30 : NameStr(state->ctx->slot->data.name),
763 30 : NameStr(state->ctx->slot->data.plugin),
764 : state->callback_name,
765 30 : LSN_FORMAT_ARGS(state->report_location));
766 : else
767 6 : errcontext("slot \"%s\", output plugin \"%s\", in the %s callback",
768 6 : NameStr(state->ctx->slot->data.name),
769 6 : NameStr(state->ctx->slot->data.plugin),
770 : state->callback_name);
771 36 : }
772 :
773 : static void
774 2064 : startup_cb_wrapper(LogicalDecodingContext *ctx, OutputPluginOptions *opt, bool is_init)
775 : {
776 : LogicalErrorCallbackState state;
777 : ErrorContextCallback errcallback;
778 :
779 : Assert(!ctx->fast_forward);
780 :
781 : /* Push callback + info on the error context stack */
782 2064 : state.ctx = ctx;
783 2064 : state.callback_name = "startup";
784 2064 : state.report_location = InvalidXLogRecPtr;
785 2064 : errcallback.callback = output_plugin_error_callback;
786 2064 : errcallback.arg = &state;
787 2064 : errcallback.previous = error_context_stack;
788 2064 : error_context_stack = &errcallback;
789 :
790 : /* set output state */
791 2064 : ctx->accept_writes = false;
792 2064 : ctx->end_xact = false;
793 :
794 : /* do the actual work: call callback */
795 2064 : ctx->callbacks.startup_cb(ctx, opt, is_init);
796 :
797 : /* Pop the error context stack */
798 2058 : error_context_stack = errcallback.previous;
799 2058 : }
800 :
801 : static void
802 1676 : shutdown_cb_wrapper(LogicalDecodingContext *ctx)
803 : {
804 : LogicalErrorCallbackState state;
805 : ErrorContextCallback errcallback;
806 :
807 : Assert(!ctx->fast_forward);
808 :
809 : /* Push callback + info on the error context stack */
810 1676 : state.ctx = ctx;
811 1676 : state.callback_name = "shutdown";
812 1676 : state.report_location = InvalidXLogRecPtr;
813 1676 : errcallback.callback = output_plugin_error_callback;
814 1676 : errcallback.arg = &state;
815 1676 : errcallback.previous = error_context_stack;
816 1676 : error_context_stack = &errcallback;
817 :
818 : /* set output state */
819 1676 : ctx->accept_writes = false;
820 1676 : ctx->end_xact = false;
821 :
822 : /* do the actual work: call callback */
823 1676 : ctx->callbacks.shutdown_cb(ctx);
824 :
825 : /* Pop the error context stack */
826 1676 : error_context_stack = errcallback.previous;
827 1676 : }
828 :
829 :
830 : /*
831 : * Callbacks for ReorderBuffer which add in some more information and then call
832 : * output_plugin.h plugins.
833 : */
834 : static void
835 2404 : begin_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn)
836 : {
837 2404 : LogicalDecodingContext *ctx = cache->private_data;
838 : LogicalErrorCallbackState state;
839 : ErrorContextCallback errcallback;
840 :
841 : Assert(!ctx->fast_forward);
842 :
843 : /* Push callback + info on the error context stack */
844 2404 : state.ctx = ctx;
845 2404 : state.callback_name = "begin";
846 2404 : state.report_location = txn->first_lsn;
847 2404 : errcallback.callback = output_plugin_error_callback;
848 2404 : errcallback.arg = &state;
849 2404 : errcallback.previous = error_context_stack;
850 2404 : error_context_stack = &errcallback;
851 :
852 : /* set output state */
853 2404 : ctx->accept_writes = true;
854 2404 : ctx->write_xid = txn->xid;
855 2404 : ctx->write_location = txn->first_lsn;
856 2404 : ctx->end_xact = false;
857 :
858 : /* do the actual work: call callback */
859 2404 : ctx->callbacks.begin_cb(ctx, txn);
860 :
861 : /* Pop the error context stack */
862 2404 : error_context_stack = errcallback.previous;
863 2404 : }
864 :
865 : static void
866 2400 : commit_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
867 : XLogRecPtr commit_lsn)
868 : {
869 2400 : LogicalDecodingContext *ctx = cache->private_data;
870 : LogicalErrorCallbackState state;
871 : ErrorContextCallback errcallback;
872 :
873 : Assert(!ctx->fast_forward);
874 :
875 : /* Push callback + info on the error context stack */
876 2400 : state.ctx = ctx;
877 2400 : state.callback_name = "commit";
878 2400 : state.report_location = txn->final_lsn; /* beginning of commit record */
879 2400 : errcallback.callback = output_plugin_error_callback;
880 2400 : errcallback.arg = &state;
881 2400 : errcallback.previous = error_context_stack;
882 2400 : error_context_stack = &errcallback;
883 :
884 : /* set output state */
885 2400 : ctx->accept_writes = true;
886 2400 : ctx->write_xid = txn->xid;
887 2400 : ctx->write_location = txn->end_lsn; /* points to the end of the record */
888 2400 : ctx->end_xact = true;
889 :
890 : /* do the actual work: call callback */
891 2400 : ctx->callbacks.commit_cb(ctx, txn, commit_lsn);
892 :
893 : /* Pop the error context stack */
894 2380 : error_context_stack = errcallback.previous;
895 2380 : }
896 :
897 : /*
898 : * The functionality of begin_prepare is quite similar to begin with the
899 : * exception that this will have gid (global transaction id) information which
900 : * can be used by plugin. Now, we thought about extending the existing begin
901 : * but that would break the replication protocol and additionally this looks
902 : * cleaner.
903 : */
904 : static void
905 60 : begin_prepare_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn)
906 : {
907 60 : LogicalDecodingContext *ctx = cache->private_data;
908 : LogicalErrorCallbackState state;
909 : ErrorContextCallback errcallback;
910 :
911 : Assert(!ctx->fast_forward);
912 :
913 : /* We're only supposed to call this when two-phase commits are supported */
914 : Assert(ctx->twophase);
915 :
916 : /* Push callback + info on the error context stack */
917 60 : state.ctx = ctx;
918 60 : state.callback_name = "begin_prepare";
919 60 : state.report_location = txn->first_lsn;
920 60 : errcallback.callback = output_plugin_error_callback;
921 60 : errcallback.arg = &state;
922 60 : errcallback.previous = error_context_stack;
923 60 : error_context_stack = &errcallback;
924 :
925 : /* set output state */
926 60 : ctx->accept_writes = true;
927 60 : ctx->write_xid = txn->xid;
928 60 : ctx->write_location = txn->first_lsn;
929 60 : ctx->end_xact = false;
930 :
931 : /*
932 : * If the plugin supports two-phase commits then begin prepare callback is
933 : * mandatory
934 : */
935 60 : if (ctx->callbacks.begin_prepare_cb == NULL)
936 0 : ereport(ERROR,
937 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
938 : errmsg("logical replication at prepare time requires a %s callback",
939 : "begin_prepare_cb")));
940 :
941 : /* do the actual work: call callback */
942 60 : ctx->callbacks.begin_prepare_cb(ctx, txn);
943 :
944 : /* Pop the error context stack */
945 60 : error_context_stack = errcallback.previous;
946 60 : }
947 :
948 : static void
949 60 : prepare_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
950 : XLogRecPtr prepare_lsn)
951 : {
952 60 : LogicalDecodingContext *ctx = cache->private_data;
953 : LogicalErrorCallbackState state;
954 : ErrorContextCallback errcallback;
955 :
956 : Assert(!ctx->fast_forward);
957 :
958 : /* We're only supposed to call this when two-phase commits are supported */
959 : Assert(ctx->twophase);
960 :
961 : /* Push callback + info on the error context stack */
962 60 : state.ctx = ctx;
963 60 : state.callback_name = "prepare";
964 60 : state.report_location = txn->final_lsn; /* beginning of prepare record */
965 60 : errcallback.callback = output_plugin_error_callback;
966 60 : errcallback.arg = &state;
967 60 : errcallback.previous = error_context_stack;
968 60 : error_context_stack = &errcallback;
969 :
970 : /* set output state */
971 60 : ctx->accept_writes = true;
972 60 : ctx->write_xid = txn->xid;
973 60 : ctx->write_location = txn->end_lsn; /* points to the end of the record */
974 60 : ctx->end_xact = true;
975 :
976 : /*
977 : * If the plugin supports two-phase commits then prepare callback is
978 : * mandatory
979 : */
980 60 : if (ctx->callbacks.prepare_cb == NULL)
981 0 : ereport(ERROR,
982 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
983 : errmsg("logical replication at prepare time requires a %s callback",
984 : "prepare_cb")));
985 :
986 : /* do the actual work: call callback */
987 60 : ctx->callbacks.prepare_cb(ctx, txn, prepare_lsn);
988 :
989 : /* Pop the error context stack */
990 60 : error_context_stack = errcallback.previous;
991 60 : }
992 :
993 : static void
994 64 : commit_prepared_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
995 : XLogRecPtr commit_lsn)
996 : {
997 64 : LogicalDecodingContext *ctx = cache->private_data;
998 : LogicalErrorCallbackState state;
999 : ErrorContextCallback errcallback;
1000 :
1001 : Assert(!ctx->fast_forward);
1002 :
1003 : /* We're only supposed to call this when two-phase commits are supported */
1004 : Assert(ctx->twophase);
1005 :
1006 : /* Push callback + info on the error context stack */
1007 64 : state.ctx = ctx;
1008 64 : state.callback_name = "commit_prepared";
1009 64 : state.report_location = txn->final_lsn; /* beginning of commit record */
1010 64 : errcallback.callback = output_plugin_error_callback;
1011 64 : errcallback.arg = &state;
1012 64 : errcallback.previous = error_context_stack;
1013 64 : error_context_stack = &errcallback;
1014 :
1015 : /* set output state */
1016 64 : ctx->accept_writes = true;
1017 64 : ctx->write_xid = txn->xid;
1018 64 : ctx->write_location = txn->end_lsn; /* points to the end of the record */
1019 64 : ctx->end_xact = true;
1020 :
1021 : /*
1022 : * If the plugin support two-phase commits then commit prepared callback
1023 : * is mandatory
1024 : */
1025 64 : if (ctx->callbacks.commit_prepared_cb == NULL)
1026 0 : ereport(ERROR,
1027 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1028 : errmsg("logical replication at prepare time requires a %s callback",
1029 : "commit_prepared_cb")));
1030 :
1031 : /* do the actual work: call callback */
1032 64 : ctx->callbacks.commit_prepared_cb(ctx, txn, commit_lsn);
1033 :
1034 : /* Pop the error context stack */
1035 64 : error_context_stack = errcallback.previous;
1036 64 : }
1037 :
1038 : static void
1039 20 : rollback_prepared_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1040 : XLogRecPtr prepare_end_lsn,
1041 : TimestampTz prepare_time)
1042 : {
1043 20 : LogicalDecodingContext *ctx = cache->private_data;
1044 : LogicalErrorCallbackState state;
1045 : ErrorContextCallback errcallback;
1046 :
1047 : Assert(!ctx->fast_forward);
1048 :
1049 : /* We're only supposed to call this when two-phase commits are supported */
1050 : Assert(ctx->twophase);
1051 :
1052 : /* Push callback + info on the error context stack */
1053 20 : state.ctx = ctx;
1054 20 : state.callback_name = "rollback_prepared";
1055 20 : state.report_location = txn->final_lsn; /* beginning of commit record */
1056 20 : errcallback.callback = output_plugin_error_callback;
1057 20 : errcallback.arg = &state;
1058 20 : errcallback.previous = error_context_stack;
1059 20 : error_context_stack = &errcallback;
1060 :
1061 : /* set output state */
1062 20 : ctx->accept_writes = true;
1063 20 : ctx->write_xid = txn->xid;
1064 20 : ctx->write_location = txn->end_lsn; /* points to the end of the record */
1065 20 : ctx->end_xact = true;
1066 :
1067 : /*
1068 : * If the plugin support two-phase commits then rollback prepared callback
1069 : * is mandatory
1070 : */
1071 20 : if (ctx->callbacks.rollback_prepared_cb == NULL)
1072 0 : ereport(ERROR,
1073 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1074 : errmsg("logical replication at prepare time requires a %s callback",
1075 : "rollback_prepared_cb")));
1076 :
1077 : /* do the actual work: call callback */
1078 20 : ctx->callbacks.rollback_prepared_cb(ctx, txn, prepare_end_lsn,
1079 : prepare_time);
1080 :
1081 : /* Pop the error context stack */
1082 20 : error_context_stack = errcallback.previous;
1083 20 : }
1084 :
1085 : static void
1086 315998 : change_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1087 : Relation relation, ReorderBufferChange *change)
1088 : {
1089 315998 : LogicalDecodingContext *ctx = cache->private_data;
1090 : LogicalErrorCallbackState state;
1091 : ErrorContextCallback errcallback;
1092 :
1093 : Assert(!ctx->fast_forward);
1094 :
1095 : /* Push callback + info on the error context stack */
1096 315998 : state.ctx = ctx;
1097 315998 : state.callback_name = "change";
1098 315998 : state.report_location = change->lsn;
1099 315998 : errcallback.callback = output_plugin_error_callback;
1100 315998 : errcallback.arg = &state;
1101 315998 : errcallback.previous = error_context_stack;
1102 315998 : error_context_stack = &errcallback;
1103 :
1104 : /* set output state */
1105 315998 : ctx->accept_writes = true;
1106 315998 : ctx->write_xid = txn->xid;
1107 :
1108 : /*
1109 : * Report this change's lsn so replies from clients can give an up-to-date
1110 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1111 : * receipt of this transaction, but it might allow another transaction's
1112 : * commit to be confirmed with one message.
1113 : */
1114 315998 : ctx->write_location = change->lsn;
1115 :
1116 315998 : ctx->end_xact = false;
1117 :
1118 315998 : ctx->callbacks.change_cb(ctx, txn, relation, change);
1119 :
1120 : /* Pop the error context stack */
1121 315994 : error_context_stack = errcallback.previous;
1122 315994 : }
1123 :
1124 : static void
1125 44 : truncate_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1126 : int nrelations, Relation relations[], ReorderBufferChange *change)
1127 : {
1128 44 : LogicalDecodingContext *ctx = cache->private_data;
1129 : LogicalErrorCallbackState state;
1130 : ErrorContextCallback errcallback;
1131 :
1132 : Assert(!ctx->fast_forward);
1133 :
1134 44 : if (!ctx->callbacks.truncate_cb)
1135 0 : return;
1136 :
1137 : /* Push callback + info on the error context stack */
1138 44 : state.ctx = ctx;
1139 44 : state.callback_name = "truncate";
1140 44 : state.report_location = change->lsn;
1141 44 : errcallback.callback = output_plugin_error_callback;
1142 44 : errcallback.arg = &state;
1143 44 : errcallback.previous = error_context_stack;
1144 44 : error_context_stack = &errcallback;
1145 :
1146 : /* set output state */
1147 44 : ctx->accept_writes = true;
1148 44 : ctx->write_xid = txn->xid;
1149 :
1150 : /*
1151 : * Report this change's lsn so replies from clients can give an up-to-date
1152 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1153 : * receipt of this transaction, but it might allow another transaction's
1154 : * commit to be confirmed with one message.
1155 : */
1156 44 : ctx->write_location = change->lsn;
1157 :
1158 44 : ctx->end_xact = false;
1159 :
1160 44 : ctx->callbacks.truncate_cb(ctx, txn, nrelations, relations, change);
1161 :
1162 : /* Pop the error context stack */
1163 44 : error_context_stack = errcallback.previous;
1164 : }
1165 :
1166 : bool
1167 296 : filter_prepare_cb_wrapper(LogicalDecodingContext *ctx, TransactionId xid,
1168 : const char *gid)
1169 : {
1170 : LogicalErrorCallbackState state;
1171 : ErrorContextCallback errcallback;
1172 : bool ret;
1173 :
1174 : Assert(!ctx->fast_forward);
1175 :
1176 : /* Push callback + info on the error context stack */
1177 296 : state.ctx = ctx;
1178 296 : state.callback_name = "filter_prepare";
1179 296 : state.report_location = InvalidXLogRecPtr;
1180 296 : errcallback.callback = output_plugin_error_callback;
1181 296 : errcallback.arg = &state;
1182 296 : errcallback.previous = error_context_stack;
1183 296 : error_context_stack = &errcallback;
1184 :
1185 : /* set output state */
1186 296 : ctx->accept_writes = false;
1187 296 : ctx->end_xact = false;
1188 :
1189 : /* do the actual work: call callback */
1190 296 : ret = ctx->callbacks.filter_prepare_cb(ctx, xid, gid);
1191 :
1192 : /* Pop the error context stack */
1193 296 : error_context_stack = errcallback.previous;
1194 :
1195 296 : return ret;
1196 : }
1197 :
1198 : bool
1199 2469842 : filter_by_origin_cb_wrapper(LogicalDecodingContext *ctx, RepOriginId origin_id)
1200 : {
1201 : LogicalErrorCallbackState state;
1202 : ErrorContextCallback errcallback;
1203 : bool ret;
1204 :
1205 : Assert(!ctx->fast_forward);
1206 :
1207 : /* Push callback + info on the error context stack */
1208 2469842 : state.ctx = ctx;
1209 2469842 : state.callback_name = "filter_by_origin";
1210 2469842 : state.report_location = InvalidXLogRecPtr;
1211 2469842 : errcallback.callback = output_plugin_error_callback;
1212 2469842 : errcallback.arg = &state;
1213 2469842 : errcallback.previous = error_context_stack;
1214 2469842 : error_context_stack = &errcallback;
1215 :
1216 : /* set output state */
1217 2469842 : ctx->accept_writes = false;
1218 2469842 : ctx->end_xact = false;
1219 :
1220 : /* do the actual work: call callback */
1221 2469842 : ret = ctx->callbacks.filter_by_origin_cb(ctx, origin_id);
1222 :
1223 : /* Pop the error context stack */
1224 2469842 : error_context_stack = errcallback.previous;
1225 :
1226 2469842 : return ret;
1227 : }
1228 :
1229 : static void
1230 32 : message_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1231 : XLogRecPtr message_lsn, bool transactional,
1232 : const char *prefix, Size message_size, const char *message)
1233 : {
1234 32 : LogicalDecodingContext *ctx = cache->private_data;
1235 : LogicalErrorCallbackState state;
1236 : ErrorContextCallback errcallback;
1237 :
1238 : Assert(!ctx->fast_forward);
1239 :
1240 32 : if (ctx->callbacks.message_cb == NULL)
1241 0 : return;
1242 :
1243 : /* Push callback + info on the error context stack */
1244 32 : state.ctx = ctx;
1245 32 : state.callback_name = "message";
1246 32 : state.report_location = message_lsn;
1247 32 : errcallback.callback = output_plugin_error_callback;
1248 32 : errcallback.arg = &state;
1249 32 : errcallback.previous = error_context_stack;
1250 32 : error_context_stack = &errcallback;
1251 :
1252 : /* set output state */
1253 32 : ctx->accept_writes = true;
1254 32 : ctx->write_xid = txn != NULL ? txn->xid : InvalidTransactionId;
1255 32 : ctx->write_location = message_lsn;
1256 32 : ctx->end_xact = false;
1257 :
1258 : /* do the actual work: call callback */
1259 32 : ctx->callbacks.message_cb(ctx, txn, message_lsn, transactional, prefix,
1260 : message_size, message);
1261 :
1262 : /* Pop the error context stack */
1263 32 : error_context_stack = errcallback.previous;
1264 : }
1265 :
1266 : static void
1267 1316 : stream_start_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1268 : XLogRecPtr first_lsn)
1269 : {
1270 1316 : LogicalDecodingContext *ctx = cache->private_data;
1271 : LogicalErrorCallbackState state;
1272 : ErrorContextCallback errcallback;
1273 :
1274 : Assert(!ctx->fast_forward);
1275 :
1276 : /* We're only supposed to call this when streaming is supported. */
1277 : Assert(ctx->streaming);
1278 :
1279 : /* Push callback + info on the error context stack */
1280 1316 : state.ctx = ctx;
1281 1316 : state.callback_name = "stream_start";
1282 1316 : state.report_location = first_lsn;
1283 1316 : errcallback.callback = output_plugin_error_callback;
1284 1316 : errcallback.arg = &state;
1285 1316 : errcallback.previous = error_context_stack;
1286 1316 : error_context_stack = &errcallback;
1287 :
1288 : /* set output state */
1289 1316 : ctx->accept_writes = true;
1290 1316 : ctx->write_xid = txn->xid;
1291 :
1292 : /*
1293 : * Report this message's lsn so replies from clients can give an
1294 : * up-to-date answer. This won't ever be enough (and shouldn't be!) to
1295 : * confirm receipt of this transaction, but it might allow another
1296 : * transaction's commit to be confirmed with one message.
1297 : */
1298 1316 : ctx->write_location = first_lsn;
1299 :
1300 1316 : ctx->end_xact = false;
1301 :
1302 : /* in streaming mode, stream_start_cb is required */
1303 1316 : if (ctx->callbacks.stream_start_cb == NULL)
1304 0 : ereport(ERROR,
1305 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1306 : errmsg("logical streaming requires a %s callback",
1307 : "stream_start_cb")));
1308 :
1309 1316 : ctx->callbacks.stream_start_cb(ctx, txn);
1310 :
1311 : /* Pop the error context stack */
1312 1316 : error_context_stack = errcallback.previous;
1313 1316 : }
1314 :
1315 : static void
1316 1316 : stream_stop_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1317 : XLogRecPtr last_lsn)
1318 : {
1319 1316 : LogicalDecodingContext *ctx = cache->private_data;
1320 : LogicalErrorCallbackState state;
1321 : ErrorContextCallback errcallback;
1322 :
1323 : Assert(!ctx->fast_forward);
1324 :
1325 : /* We're only supposed to call this when streaming is supported. */
1326 : Assert(ctx->streaming);
1327 :
1328 : /* Push callback + info on the error context stack */
1329 1316 : state.ctx = ctx;
1330 1316 : state.callback_name = "stream_stop";
1331 1316 : state.report_location = last_lsn;
1332 1316 : errcallback.callback = output_plugin_error_callback;
1333 1316 : errcallback.arg = &state;
1334 1316 : errcallback.previous = error_context_stack;
1335 1316 : error_context_stack = &errcallback;
1336 :
1337 : /* set output state */
1338 1316 : ctx->accept_writes = true;
1339 1316 : ctx->write_xid = txn->xid;
1340 :
1341 : /*
1342 : * Report this message's lsn so replies from clients can give an
1343 : * up-to-date answer. This won't ever be enough (and shouldn't be!) to
1344 : * confirm receipt of this transaction, but it might allow another
1345 : * transaction's commit to be confirmed with one message.
1346 : */
1347 1316 : ctx->write_location = last_lsn;
1348 :
1349 1316 : ctx->end_xact = false;
1350 :
1351 : /* in streaming mode, stream_stop_cb is required */
1352 1316 : if (ctx->callbacks.stream_stop_cb == NULL)
1353 0 : ereport(ERROR,
1354 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1355 : errmsg("logical streaming requires a %s callback",
1356 : "stream_stop_cb")));
1357 :
1358 1316 : ctx->callbacks.stream_stop_cb(ctx, txn);
1359 :
1360 : /* Pop the error context stack */
1361 1316 : error_context_stack = errcallback.previous;
1362 1316 : }
1363 :
1364 : static void
1365 60 : stream_abort_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1366 : XLogRecPtr abort_lsn)
1367 : {
1368 60 : LogicalDecodingContext *ctx = cache->private_data;
1369 : LogicalErrorCallbackState state;
1370 : ErrorContextCallback errcallback;
1371 :
1372 : Assert(!ctx->fast_forward);
1373 :
1374 : /* We're only supposed to call this when streaming is supported. */
1375 : Assert(ctx->streaming);
1376 :
1377 : /* Push callback + info on the error context stack */
1378 60 : state.ctx = ctx;
1379 60 : state.callback_name = "stream_abort";
1380 60 : state.report_location = abort_lsn;
1381 60 : errcallback.callback = output_plugin_error_callback;
1382 60 : errcallback.arg = &state;
1383 60 : errcallback.previous = error_context_stack;
1384 60 : error_context_stack = &errcallback;
1385 :
1386 : /* set output state */
1387 60 : ctx->accept_writes = true;
1388 60 : ctx->write_xid = txn->xid;
1389 60 : ctx->write_location = abort_lsn;
1390 60 : ctx->end_xact = true;
1391 :
1392 : /* in streaming mode, stream_abort_cb is required */
1393 60 : if (ctx->callbacks.stream_abort_cb == NULL)
1394 0 : ereport(ERROR,
1395 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1396 : errmsg("logical streaming requires a %s callback",
1397 : "stream_abort_cb")));
1398 :
1399 60 : ctx->callbacks.stream_abort_cb(ctx, txn, abort_lsn);
1400 :
1401 : /* Pop the error context stack */
1402 60 : error_context_stack = errcallback.previous;
1403 60 : }
1404 :
1405 : static void
1406 24 : stream_prepare_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1407 : XLogRecPtr prepare_lsn)
1408 : {
1409 24 : LogicalDecodingContext *ctx = cache->private_data;
1410 : LogicalErrorCallbackState state;
1411 : ErrorContextCallback errcallback;
1412 :
1413 : Assert(!ctx->fast_forward);
1414 :
1415 : /*
1416 : * We're only supposed to call this when streaming and two-phase commits
1417 : * are supported.
1418 : */
1419 : Assert(ctx->streaming);
1420 : Assert(ctx->twophase);
1421 :
1422 : /* Push callback + info on the error context stack */
1423 24 : state.ctx = ctx;
1424 24 : state.callback_name = "stream_prepare";
1425 24 : state.report_location = txn->final_lsn;
1426 24 : errcallback.callback = output_plugin_error_callback;
1427 24 : errcallback.arg = &state;
1428 24 : errcallback.previous = error_context_stack;
1429 24 : error_context_stack = &errcallback;
1430 :
1431 : /* set output state */
1432 24 : ctx->accept_writes = true;
1433 24 : ctx->write_xid = txn->xid;
1434 24 : ctx->write_location = txn->end_lsn;
1435 24 : ctx->end_xact = true;
1436 :
1437 : /* in streaming mode with two-phase commits, stream_prepare_cb is required */
1438 24 : if (ctx->callbacks.stream_prepare_cb == NULL)
1439 0 : ereport(ERROR,
1440 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1441 : errmsg("logical streaming at prepare time requires a %s callback",
1442 : "stream_prepare_cb")));
1443 :
1444 24 : ctx->callbacks.stream_prepare_cb(ctx, txn, prepare_lsn);
1445 :
1446 : /* Pop the error context stack */
1447 24 : error_context_stack = errcallback.previous;
1448 24 : }
1449 :
1450 : static void
1451 100 : stream_commit_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1452 : XLogRecPtr commit_lsn)
1453 : {
1454 100 : LogicalDecodingContext *ctx = cache->private_data;
1455 : LogicalErrorCallbackState state;
1456 : ErrorContextCallback errcallback;
1457 :
1458 : Assert(!ctx->fast_forward);
1459 :
1460 : /* We're only supposed to call this when streaming is supported. */
1461 : Assert(ctx->streaming);
1462 :
1463 : /* Push callback + info on the error context stack */
1464 100 : state.ctx = ctx;
1465 100 : state.callback_name = "stream_commit";
1466 100 : state.report_location = txn->final_lsn;
1467 100 : errcallback.callback = output_plugin_error_callback;
1468 100 : errcallback.arg = &state;
1469 100 : errcallback.previous = error_context_stack;
1470 100 : error_context_stack = &errcallback;
1471 :
1472 : /* set output state */
1473 100 : ctx->accept_writes = true;
1474 100 : ctx->write_xid = txn->xid;
1475 100 : ctx->write_location = txn->end_lsn;
1476 100 : ctx->end_xact = true;
1477 :
1478 : /* in streaming mode, stream_commit_cb is required */
1479 100 : if (ctx->callbacks.stream_commit_cb == NULL)
1480 0 : ereport(ERROR,
1481 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1482 : errmsg("logical streaming requires a %s callback",
1483 : "stream_commit_cb")));
1484 :
1485 100 : ctx->callbacks.stream_commit_cb(ctx, txn, commit_lsn);
1486 :
1487 : /* Pop the error context stack */
1488 100 : error_context_stack = errcallback.previous;
1489 100 : }
1490 :
1491 : static void
1492 351972 : stream_change_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1493 : Relation relation, ReorderBufferChange *change)
1494 : {
1495 351972 : LogicalDecodingContext *ctx = cache->private_data;
1496 : LogicalErrorCallbackState state;
1497 : ErrorContextCallback errcallback;
1498 :
1499 : Assert(!ctx->fast_forward);
1500 :
1501 : /* We're only supposed to call this when streaming is supported. */
1502 : Assert(ctx->streaming);
1503 :
1504 : /* Push callback + info on the error context stack */
1505 351972 : state.ctx = ctx;
1506 351972 : state.callback_name = "stream_change";
1507 351972 : state.report_location = change->lsn;
1508 351972 : errcallback.callback = output_plugin_error_callback;
1509 351972 : errcallback.arg = &state;
1510 351972 : errcallback.previous = error_context_stack;
1511 351972 : error_context_stack = &errcallback;
1512 :
1513 : /* set output state */
1514 351972 : ctx->accept_writes = true;
1515 351972 : ctx->write_xid = txn->xid;
1516 :
1517 : /*
1518 : * Report this change's lsn so replies from clients can give an up-to-date
1519 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1520 : * receipt of this transaction, but it might allow another transaction's
1521 : * commit to be confirmed with one message.
1522 : */
1523 351972 : ctx->write_location = change->lsn;
1524 :
1525 351972 : ctx->end_xact = false;
1526 :
1527 : /* in streaming mode, stream_change_cb is required */
1528 351972 : if (ctx->callbacks.stream_change_cb == NULL)
1529 0 : ereport(ERROR,
1530 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1531 : errmsg("logical streaming requires a %s callback",
1532 : "stream_change_cb")));
1533 :
1534 351972 : ctx->callbacks.stream_change_cb(ctx, txn, relation, change);
1535 :
1536 : /* Pop the error context stack */
1537 351972 : error_context_stack = errcallback.previous;
1538 351972 : }
1539 :
1540 : static void
1541 6 : stream_message_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1542 : XLogRecPtr message_lsn, bool transactional,
1543 : const char *prefix, Size message_size, const char *message)
1544 : {
1545 6 : LogicalDecodingContext *ctx = cache->private_data;
1546 : LogicalErrorCallbackState state;
1547 : ErrorContextCallback errcallback;
1548 :
1549 : Assert(!ctx->fast_forward);
1550 :
1551 : /* We're only supposed to call this when streaming is supported. */
1552 : Assert(ctx->streaming);
1553 :
1554 : /* this callback is optional */
1555 6 : if (ctx->callbacks.stream_message_cb == NULL)
1556 0 : return;
1557 :
1558 : /* Push callback + info on the error context stack */
1559 6 : state.ctx = ctx;
1560 6 : state.callback_name = "stream_message";
1561 6 : state.report_location = message_lsn;
1562 6 : errcallback.callback = output_plugin_error_callback;
1563 6 : errcallback.arg = &state;
1564 6 : errcallback.previous = error_context_stack;
1565 6 : error_context_stack = &errcallback;
1566 :
1567 : /* set output state */
1568 6 : ctx->accept_writes = true;
1569 6 : ctx->write_xid = txn != NULL ? txn->xid : InvalidTransactionId;
1570 6 : ctx->write_location = message_lsn;
1571 6 : ctx->end_xact = false;
1572 :
1573 : /* do the actual work: call callback */
1574 6 : ctx->callbacks.stream_message_cb(ctx, txn, message_lsn, transactional, prefix,
1575 : message_size, message);
1576 :
1577 : /* Pop the error context stack */
1578 6 : error_context_stack = errcallback.previous;
1579 : }
1580 :
1581 : static void
1582 0 : stream_truncate_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1583 : int nrelations, Relation relations[],
1584 : ReorderBufferChange *change)
1585 : {
1586 0 : LogicalDecodingContext *ctx = cache->private_data;
1587 : LogicalErrorCallbackState state;
1588 : ErrorContextCallback errcallback;
1589 :
1590 : Assert(!ctx->fast_forward);
1591 :
1592 : /* We're only supposed to call this when streaming is supported. */
1593 : Assert(ctx->streaming);
1594 :
1595 : /* this callback is optional */
1596 0 : if (!ctx->callbacks.stream_truncate_cb)
1597 0 : return;
1598 :
1599 : /* Push callback + info on the error context stack */
1600 0 : state.ctx = ctx;
1601 0 : state.callback_name = "stream_truncate";
1602 0 : state.report_location = change->lsn;
1603 0 : errcallback.callback = output_plugin_error_callback;
1604 0 : errcallback.arg = &state;
1605 0 : errcallback.previous = error_context_stack;
1606 0 : error_context_stack = &errcallback;
1607 :
1608 : /* set output state */
1609 0 : ctx->accept_writes = true;
1610 0 : ctx->write_xid = txn->xid;
1611 :
1612 : /*
1613 : * Report this change's lsn so replies from clients can give an up-to-date
1614 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1615 : * receipt of this transaction, but it might allow another transaction's
1616 : * commit to be confirmed with one message.
1617 : */
1618 0 : ctx->write_location = change->lsn;
1619 :
1620 0 : ctx->end_xact = false;
1621 :
1622 0 : ctx->callbacks.stream_truncate_cb(ctx, txn, nrelations, relations, change);
1623 :
1624 : /* Pop the error context stack */
1625 0 : error_context_stack = errcallback.previous;
1626 : }
1627 :
1628 : static void
1629 6214 : update_progress_txn_cb_wrapper(ReorderBuffer *cache, ReorderBufferTXN *txn,
1630 : XLogRecPtr lsn)
1631 : {
1632 6214 : LogicalDecodingContext *ctx = cache->private_data;
1633 : LogicalErrorCallbackState state;
1634 : ErrorContextCallback errcallback;
1635 :
1636 : Assert(!ctx->fast_forward);
1637 :
1638 : /* Push callback + info on the error context stack */
1639 6214 : state.ctx = ctx;
1640 6214 : state.callback_name = "update_progress_txn";
1641 6214 : state.report_location = lsn;
1642 6214 : errcallback.callback = output_plugin_error_callback;
1643 6214 : errcallback.arg = &state;
1644 6214 : errcallback.previous = error_context_stack;
1645 6214 : error_context_stack = &errcallback;
1646 :
1647 : /* set output state */
1648 6214 : ctx->accept_writes = false;
1649 6214 : ctx->write_xid = txn->xid;
1650 :
1651 : /*
1652 : * Report this change's lsn so replies from clients can give an up-to-date
1653 : * answer. This won't ever be enough (and shouldn't be!) to confirm
1654 : * receipt of this transaction, but it might allow another transaction's
1655 : * commit to be confirmed with one message.
1656 : */
1657 6214 : ctx->write_location = lsn;
1658 :
1659 6214 : ctx->end_xact = false;
1660 :
1661 6214 : OutputPluginUpdateProgress(ctx, false);
1662 :
1663 : /* Pop the error context stack */
1664 6214 : error_context_stack = errcallback.previous;
1665 6214 : }
1666 :
1667 : /*
1668 : * Set the required catalog xmin horizon for historic snapshots in the current
1669 : * replication slot.
1670 : *
1671 : * Note that in the most cases, we won't be able to immediately use the xmin
1672 : * to increase the xmin horizon: we need to wait till the client has confirmed
1673 : * receiving current_lsn with LogicalConfirmReceivedLocation().
1674 : */
1675 : void
1676 794 : LogicalIncreaseXminForSlot(XLogRecPtr current_lsn, TransactionId xmin)
1677 : {
1678 794 : bool updated_xmin = false;
1679 : ReplicationSlot *slot;
1680 794 : bool got_new_xmin = false;
1681 :
1682 794 : slot = MyReplicationSlot;
1683 :
1684 : Assert(slot != NULL);
1685 :
1686 794 : SpinLockAcquire(&slot->mutex);
1687 :
1688 : /*
1689 : * don't overwrite if we already have a newer xmin. This can happen if we
1690 : * restart decoding in a slot.
1691 : */
1692 794 : if (TransactionIdPrecedesOrEquals(xmin, slot->data.catalog_xmin))
1693 : {
1694 : }
1695 :
1696 : /*
1697 : * If the client has already confirmed up to this lsn, we directly can
1698 : * mark this as accepted. This can happen if we restart decoding in a
1699 : * slot.
1700 : */
1701 198 : else if (current_lsn <= slot->data.confirmed_flush)
1702 : {
1703 102 : slot->candidate_catalog_xmin = xmin;
1704 102 : slot->candidate_xmin_lsn = current_lsn;
1705 :
1706 : /* our candidate can directly be used */
1707 102 : updated_xmin = true;
1708 : }
1709 :
1710 : /*
1711 : * Only increase if the previous values have been applied, otherwise we
1712 : * might never end up updating if the receiver acks too slowly.
1713 : */
1714 96 : else if (slot->candidate_xmin_lsn == InvalidXLogRecPtr)
1715 : {
1716 46 : slot->candidate_catalog_xmin = xmin;
1717 46 : slot->candidate_xmin_lsn = current_lsn;
1718 :
1719 : /*
1720 : * Log new xmin at an appropriate log level after releasing the
1721 : * spinlock.
1722 : */
1723 46 : got_new_xmin = true;
1724 : }
1725 794 : SpinLockRelease(&slot->mutex);
1726 :
1727 794 : if (got_new_xmin)
1728 46 : elog(DEBUG1, "got new catalog xmin %u at %X/%X", xmin,
1729 : LSN_FORMAT_ARGS(current_lsn));
1730 :
1731 : /* candidate already valid with the current flush position, apply */
1732 794 : if (updated_xmin)
1733 102 : LogicalConfirmReceivedLocation(slot->data.confirmed_flush);
1734 794 : }
1735 :
1736 : /*
1737 : * Mark the minimal LSN (restart_lsn) we need to read to replay all
1738 : * transactions that have not yet committed at current_lsn.
1739 : *
1740 : * Just like LogicalIncreaseXminForSlot this only takes effect when the
1741 : * client has confirmed to have received current_lsn.
1742 : */
1743 : void
1744 692 : LogicalIncreaseRestartDecodingForSlot(XLogRecPtr current_lsn, XLogRecPtr restart_lsn)
1745 : {
1746 692 : bool updated_lsn = false;
1747 : ReplicationSlot *slot;
1748 :
1749 692 : slot = MyReplicationSlot;
1750 :
1751 : Assert(slot != NULL);
1752 : Assert(restart_lsn != InvalidXLogRecPtr);
1753 : Assert(current_lsn != InvalidXLogRecPtr);
1754 :
1755 692 : SpinLockAcquire(&slot->mutex);
1756 :
1757 : /* don't overwrite if have a newer restart lsn */
1758 692 : if (restart_lsn <= slot->data.restart_lsn)
1759 : {
1760 20 : SpinLockRelease(&slot->mutex);
1761 : }
1762 :
1763 : /*
1764 : * We might have already flushed far enough to directly accept this lsn,
1765 : * in this case there is no need to check for existing candidate LSNs
1766 : */
1767 672 : else if (current_lsn <= slot->data.confirmed_flush)
1768 : {
1769 534 : slot->candidate_restart_valid = current_lsn;
1770 534 : slot->candidate_restart_lsn = restart_lsn;
1771 534 : SpinLockRelease(&slot->mutex);
1772 :
1773 : /* our candidate can directly be used */
1774 534 : updated_lsn = true;
1775 : }
1776 :
1777 : /*
1778 : * Only increase if the previous values have been applied, otherwise we
1779 : * might never end up updating if the receiver acks too slowly. A missed
1780 : * value here will just cause some extra effort after reconnecting.
1781 : */
1782 138 : else if (slot->candidate_restart_valid == InvalidXLogRecPtr)
1783 : {
1784 76 : slot->candidate_restart_valid = current_lsn;
1785 76 : slot->candidate_restart_lsn = restart_lsn;
1786 76 : SpinLockRelease(&slot->mutex);
1787 :
1788 76 : elog(DEBUG1, "got new restart lsn %X/%X at %X/%X",
1789 : LSN_FORMAT_ARGS(restart_lsn),
1790 : LSN_FORMAT_ARGS(current_lsn));
1791 : }
1792 : else
1793 : {
1794 : XLogRecPtr candidate_restart_lsn;
1795 : XLogRecPtr candidate_restart_valid;
1796 : XLogRecPtr confirmed_flush;
1797 :
1798 62 : candidate_restart_lsn = slot->candidate_restart_lsn;
1799 62 : candidate_restart_valid = slot->candidate_restart_valid;
1800 62 : confirmed_flush = slot->data.confirmed_flush;
1801 62 : SpinLockRelease(&slot->mutex);
1802 :
1803 62 : elog(DEBUG1, "failed to increase restart lsn: proposed %X/%X, after %X/%X, current candidate %X/%X, current after %X/%X, flushed up to %X/%X",
1804 : LSN_FORMAT_ARGS(restart_lsn),
1805 : LSN_FORMAT_ARGS(current_lsn),
1806 : LSN_FORMAT_ARGS(candidate_restart_lsn),
1807 : LSN_FORMAT_ARGS(candidate_restart_valid),
1808 : LSN_FORMAT_ARGS(confirmed_flush));
1809 : }
1810 :
1811 : /* candidates are already valid with the current flush position, apply */
1812 692 : if (updated_lsn)
1813 534 : LogicalConfirmReceivedLocation(slot->data.confirmed_flush);
1814 692 : }
1815 :
1816 : /*
1817 : * Handle a consumer's confirmation having received all changes up to lsn.
1818 : */
1819 : void
1820 93232 : LogicalConfirmReceivedLocation(XLogRecPtr lsn)
1821 : {
1822 : Assert(lsn != InvalidXLogRecPtr);
1823 :
1824 : /* Do an unlocked check for candidate_lsn first. */
1825 93232 : if (MyReplicationSlot->candidate_xmin_lsn != InvalidXLogRecPtr ||
1826 93092 : MyReplicationSlot->candidate_restart_valid != InvalidXLogRecPtr)
1827 708 : {
1828 708 : bool updated_xmin = false;
1829 708 : bool updated_restart = false;
1830 : XLogRecPtr restart_lsn pg_attribute_unused();
1831 :
1832 708 : SpinLockAcquire(&MyReplicationSlot->mutex);
1833 :
1834 : /* remember the old restart lsn */
1835 708 : restart_lsn = MyReplicationSlot->data.restart_lsn;
1836 :
1837 : /*
1838 : * Prevent moving the confirmed_flush backwards, as this could lead to
1839 : * data duplication issues caused by replicating already replicated
1840 : * changes.
1841 : *
1842 : * This can happen when a client acknowledges an LSN it doesn't have
1843 : * to do anything for, and thus didn't store persistently. After a
1844 : * restart, the client can send the prior LSN that it stored
1845 : * persistently as an acknowledgement, but we need to ignore such an
1846 : * LSN. See similar case handling in CreateDecodingContext.
1847 : */
1848 708 : if (lsn > MyReplicationSlot->data.confirmed_flush)
1849 68 : MyReplicationSlot->data.confirmed_flush = lsn;
1850 :
1851 : /* if we're past the location required for bumping xmin, do so */
1852 708 : if (MyReplicationSlot->candidate_xmin_lsn != InvalidXLogRecPtr &&
1853 140 : MyReplicationSlot->candidate_xmin_lsn <= lsn)
1854 : {
1855 : /*
1856 : * We have to write the changed xmin to disk *before* we change
1857 : * the in-memory value, otherwise after a crash we wouldn't know
1858 : * that some catalog tuples might have been removed already.
1859 : *
1860 : * Ensure that by first writing to ->xmin and only update
1861 : * ->effective_xmin once the new state is synced to disk. After a
1862 : * crash ->effective_xmin is set to ->xmin.
1863 : */
1864 136 : if (TransactionIdIsValid(MyReplicationSlot->candidate_catalog_xmin) &&
1865 136 : MyReplicationSlot->data.catalog_xmin != MyReplicationSlot->candidate_catalog_xmin)
1866 : {
1867 136 : MyReplicationSlot->data.catalog_xmin = MyReplicationSlot->candidate_catalog_xmin;
1868 136 : MyReplicationSlot->candidate_catalog_xmin = InvalidTransactionId;
1869 136 : MyReplicationSlot->candidate_xmin_lsn = InvalidXLogRecPtr;
1870 136 : updated_xmin = true;
1871 : }
1872 : }
1873 :
1874 708 : if (MyReplicationSlot->candidate_restart_valid != InvalidXLogRecPtr &&
1875 606 : MyReplicationSlot->candidate_restart_valid <= lsn)
1876 : {
1877 : Assert(MyReplicationSlot->candidate_restart_lsn != InvalidXLogRecPtr);
1878 :
1879 602 : MyReplicationSlot->data.restart_lsn = MyReplicationSlot->candidate_restart_lsn;
1880 602 : MyReplicationSlot->candidate_restart_lsn = InvalidXLogRecPtr;
1881 602 : MyReplicationSlot->candidate_restart_valid = InvalidXLogRecPtr;
1882 602 : updated_restart = true;
1883 : }
1884 :
1885 708 : SpinLockRelease(&MyReplicationSlot->mutex);
1886 :
1887 : /* first write new xmin to disk, so we know what's up after a crash */
1888 708 : if (updated_xmin || updated_restart)
1889 : {
1890 : #ifdef USE_INJECTION_POINTS
1891 : XLogSegNo seg1,
1892 : seg2;
1893 :
1894 704 : XLByteToSeg(restart_lsn, seg1, wal_segment_size);
1895 704 : XLByteToSeg(MyReplicationSlot->data.restart_lsn, seg2, wal_segment_size);
1896 :
1897 : /* trigger injection point, but only if segment changes */
1898 704 : if (seg1 != seg2)
1899 12 : INJECTION_POINT("logical-replication-slot-advance-segment", NULL);
1900 : #endif
1901 :
1902 704 : ReplicationSlotMarkDirty();
1903 704 : ReplicationSlotSave();
1904 704 : elog(DEBUG1, "updated xmin: %u restart: %u", updated_xmin, updated_restart);
1905 : }
1906 :
1907 : /*
1908 : * Now the new xmin is safely on disk, we can let the global value
1909 : * advance. We do not take ProcArrayLock or similar since we only
1910 : * advance xmin here and there's not much harm done by a concurrent
1911 : * computation missing that.
1912 : */
1913 708 : if (updated_xmin)
1914 : {
1915 136 : SpinLockAcquire(&MyReplicationSlot->mutex);
1916 136 : MyReplicationSlot->effective_catalog_xmin = MyReplicationSlot->data.catalog_xmin;
1917 136 : SpinLockRelease(&MyReplicationSlot->mutex);
1918 :
1919 136 : ReplicationSlotsComputeRequiredXmin(false);
1920 136 : ReplicationSlotsComputeRequiredLSN();
1921 : }
1922 : }
1923 : else
1924 : {
1925 92524 : SpinLockAcquire(&MyReplicationSlot->mutex);
1926 :
1927 : /*
1928 : * Prevent moving the confirmed_flush backwards. See comments above
1929 : * for the details.
1930 : */
1931 92524 : if (lsn > MyReplicationSlot->data.confirmed_flush)
1932 91056 : MyReplicationSlot->data.confirmed_flush = lsn;
1933 :
1934 92524 : SpinLockRelease(&MyReplicationSlot->mutex);
1935 : }
1936 93232 : }
1937 :
1938 : /*
1939 : * Clear logical streaming state during (sub)transaction abort.
1940 : */
1941 : void
1942 58412 : ResetLogicalStreamingState(void)
1943 : {
1944 58412 : CheckXidAlive = InvalidTransactionId;
1945 58412 : bsysscan = false;
1946 58412 : }
1947 :
1948 : /*
1949 : * Report stats for a slot.
1950 : */
1951 : void
1952 10362 : UpdateDecodingStats(LogicalDecodingContext *ctx)
1953 : {
1954 10362 : ReorderBuffer *rb = ctx->reorder;
1955 : PgStat_StatReplSlotEntry repSlotStat;
1956 :
1957 : /* Nothing to do if we don't have any replication stats to be sent. */
1958 10362 : if (rb->spillBytes <= 0 && rb->streamBytes <= 0 && rb->totalBytes <= 0)
1959 484 : return;
1960 :
1961 9878 : elog(DEBUG2, "UpdateDecodingStats: updating stats %p %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64,
1962 : rb,
1963 : rb->spillTxns,
1964 : rb->spillCount,
1965 : rb->spillBytes,
1966 : rb->streamTxns,
1967 : rb->streamCount,
1968 : rb->streamBytes,
1969 : rb->totalTxns,
1970 : rb->totalBytes);
1971 :
1972 9878 : repSlotStat.spill_txns = rb->spillTxns;
1973 9878 : repSlotStat.spill_count = rb->spillCount;
1974 9878 : repSlotStat.spill_bytes = rb->spillBytes;
1975 9878 : repSlotStat.stream_txns = rb->streamTxns;
1976 9878 : repSlotStat.stream_count = rb->streamCount;
1977 9878 : repSlotStat.stream_bytes = rb->streamBytes;
1978 9878 : repSlotStat.total_txns = rb->totalTxns;
1979 9878 : repSlotStat.total_bytes = rb->totalBytes;
1980 :
1981 9878 : pgstat_report_replslot(ctx->slot, &repSlotStat);
1982 :
1983 9878 : rb->spillTxns = 0;
1984 9878 : rb->spillCount = 0;
1985 9878 : rb->spillBytes = 0;
1986 9878 : rb->streamTxns = 0;
1987 9878 : rb->streamCount = 0;
1988 9878 : rb->streamBytes = 0;
1989 9878 : rb->totalTxns = 0;
1990 9878 : rb->totalBytes = 0;
1991 : }
1992 :
1993 : /*
1994 : * Read up to the end of WAL starting from the decoding slot's restart_lsn.
1995 : * Return true if any meaningful/decodable WAL records are encountered,
1996 : * otherwise false.
1997 : */
1998 : bool
1999 10 : LogicalReplicationSlotHasPendingWal(XLogRecPtr end_of_wal)
2000 : {
2001 10 : bool has_pending_wal = false;
2002 :
2003 : Assert(MyReplicationSlot);
2004 :
2005 10 : PG_TRY();
2006 : {
2007 : LogicalDecodingContext *ctx;
2008 :
2009 : /*
2010 : * Create our decoding context in fast_forward mode, passing start_lsn
2011 : * as InvalidXLogRecPtr, so that we start processing from the slot's
2012 : * confirmed_flush.
2013 : */
2014 20 : ctx = CreateDecodingContext(InvalidXLogRecPtr,
2015 : NIL,
2016 : true, /* fast_forward */
2017 10 : XL_ROUTINE(.page_read = read_local_xlog_page,
2018 : .segment_open = wal_segment_open,
2019 : .segment_close = wal_segment_close),
2020 : NULL, NULL, NULL);
2021 :
2022 : /*
2023 : * Start reading at the slot's restart_lsn, which we know points to a
2024 : * valid record.
2025 : */
2026 10 : XLogBeginRead(ctx->reader, MyReplicationSlot->data.restart_lsn);
2027 :
2028 : /* Invalidate non-timetravel entries */
2029 10 : InvalidateSystemCaches();
2030 :
2031 : /* Loop until the end of WAL or some changes are processed */
2032 300 : while (!has_pending_wal && ctx->reader->EndRecPtr < end_of_wal)
2033 : {
2034 : XLogRecord *record;
2035 290 : char *errm = NULL;
2036 :
2037 290 : record = XLogReadRecord(ctx->reader, &errm);
2038 :
2039 290 : if (errm)
2040 0 : elog(ERROR, "could not find record for logical decoding: %s", errm);
2041 :
2042 290 : if (record != NULL)
2043 290 : LogicalDecodingProcessRecord(ctx, ctx->reader);
2044 :
2045 290 : has_pending_wal = ctx->processing_required;
2046 :
2047 290 : CHECK_FOR_INTERRUPTS();
2048 : }
2049 :
2050 : /* Clean up */
2051 10 : FreeDecodingContext(ctx);
2052 10 : InvalidateSystemCaches();
2053 : }
2054 0 : PG_CATCH();
2055 : {
2056 : /* clear all timetravel entries */
2057 0 : InvalidateSystemCaches();
2058 :
2059 0 : PG_RE_THROW();
2060 : }
2061 10 : PG_END_TRY();
2062 :
2063 10 : return has_pending_wal;
2064 : }
2065 :
2066 : /*
2067 : * Helper function for advancing our logical replication slot forward.
2068 : *
2069 : * The slot's restart_lsn is used as start point for reading records, while
2070 : * confirmed_flush is used as base point for the decoding context.
2071 : *
2072 : * We cannot just do LogicalConfirmReceivedLocation to update confirmed_flush,
2073 : * because we need to digest WAL to advance restart_lsn allowing to recycle
2074 : * WAL and removal of old catalog tuples. As decoding is done in fast_forward
2075 : * mode, no changes are generated anyway.
2076 : *
2077 : * *found_consistent_snapshot will be true if the initial decoding snapshot has
2078 : * been built; Otherwise, it will be false.
2079 : */
2080 : XLogRecPtr
2081 32 : LogicalSlotAdvanceAndCheckSnapState(XLogRecPtr moveto,
2082 : bool *found_consistent_snapshot)
2083 : {
2084 : LogicalDecodingContext *ctx;
2085 32 : ResourceOwner old_resowner = CurrentResourceOwner;
2086 : XLogRecPtr retlsn;
2087 :
2088 : Assert(moveto != InvalidXLogRecPtr);
2089 :
2090 32 : if (found_consistent_snapshot)
2091 10 : *found_consistent_snapshot = false;
2092 :
2093 32 : PG_TRY();
2094 : {
2095 : /*
2096 : * Create our decoding context in fast_forward mode, passing start_lsn
2097 : * as InvalidXLogRecPtr, so that we start processing from my slot's
2098 : * confirmed_flush.
2099 : */
2100 64 : ctx = CreateDecodingContext(InvalidXLogRecPtr,
2101 : NIL,
2102 : true, /* fast_forward */
2103 32 : XL_ROUTINE(.page_read = read_local_xlog_page,
2104 : .segment_open = wal_segment_open,
2105 : .segment_close = wal_segment_close),
2106 : NULL, NULL, NULL);
2107 :
2108 : /*
2109 : * Wait for specified streaming replication standby servers (if any)
2110 : * to confirm receipt of WAL up to moveto lsn.
2111 : */
2112 32 : WaitForStandbyConfirmation(moveto);
2113 :
2114 : /*
2115 : * Start reading at the slot's restart_lsn, which we know to point to
2116 : * a valid record.
2117 : */
2118 32 : XLogBeginRead(ctx->reader, MyReplicationSlot->data.restart_lsn);
2119 :
2120 : /* invalidate non-timetravel entries */
2121 32 : InvalidateSystemCaches();
2122 :
2123 : /* Decode records until we reach the requested target */
2124 4184 : while (ctx->reader->EndRecPtr < moveto)
2125 : {
2126 4152 : char *errm = NULL;
2127 : XLogRecord *record;
2128 :
2129 : /*
2130 : * Read records. No changes are generated in fast_forward mode,
2131 : * but snapbuilder/slot statuses are updated properly.
2132 : */
2133 4152 : record = XLogReadRecord(ctx->reader, &errm);
2134 4152 : if (errm)
2135 0 : elog(ERROR, "could not find record while advancing replication slot: %s",
2136 : errm);
2137 :
2138 : /*
2139 : * Process the record. Storage-level changes are ignored in
2140 : * fast_forward mode, but other modules (such as snapbuilder)
2141 : * might still have critical updates to do.
2142 : */
2143 4152 : if (record)
2144 4152 : LogicalDecodingProcessRecord(ctx, ctx->reader);
2145 :
2146 4152 : CHECK_FOR_INTERRUPTS();
2147 : }
2148 :
2149 32 : if (found_consistent_snapshot && DecodingContextReady(ctx))
2150 10 : *found_consistent_snapshot = true;
2151 :
2152 : /*
2153 : * Logical decoding could have clobbered CurrentResourceOwner during
2154 : * transaction management, so restore the executor's value. (This is
2155 : * a kluge, but it's not worth cleaning up right now.)
2156 : */
2157 32 : CurrentResourceOwner = old_resowner;
2158 :
2159 32 : if (ctx->reader->EndRecPtr != InvalidXLogRecPtr)
2160 : {
2161 32 : LogicalConfirmReceivedLocation(moveto);
2162 :
2163 : /*
2164 : * If only the confirmed_flush LSN has changed the slot won't get
2165 : * marked as dirty by the above. Callers on the walsender
2166 : * interface are expected to keep track of their own progress and
2167 : * don't need it written out. But SQL-interface users cannot
2168 : * specify their own start positions and it's harder for them to
2169 : * keep track of their progress, so we should make more of an
2170 : * effort to save it for them.
2171 : *
2172 : * Dirty the slot so it is written out at the next checkpoint. The
2173 : * LSN position advanced to may still be lost on a crash but this
2174 : * makes the data consistent after a clean shutdown.
2175 : */
2176 32 : ReplicationSlotMarkDirty();
2177 : }
2178 :
2179 32 : retlsn = MyReplicationSlot->data.confirmed_flush;
2180 :
2181 : /* free context, call shutdown callback */
2182 32 : FreeDecodingContext(ctx);
2183 :
2184 32 : InvalidateSystemCaches();
2185 : }
2186 0 : PG_CATCH();
2187 : {
2188 : /* clear all timetravel entries */
2189 0 : InvalidateSystemCaches();
2190 :
2191 0 : PG_RE_THROW();
2192 : }
2193 32 : PG_END_TRY();
2194 :
2195 32 : return retlsn;
2196 : }
|
