Line data Source code
1 : /*-------------------------------------------------------------------------
2 : * applyparallelworker.c
3 : * Support routines for applying xact by parallel apply worker
4 : *
5 : * Copyright (c) 2023-2024, PostgreSQL Global Development Group
6 : *
7 : * IDENTIFICATION
8 : * src/backend/replication/logical/applyparallelworker.c
9 : *
10 : * This file contains the code to launch, set up, and teardown a parallel apply
11 : * worker which receives the changes from the leader worker and invokes routines
12 : * to apply those on the subscriber database. Additionally, this file contains
13 : * routines that are intended to support setting up, using, and tearing down a
14 : * ParallelApplyWorkerInfo which is required so the leader worker and parallel
15 : * apply workers can communicate with each other.
16 : *
17 : * The parallel apply workers are assigned (if available) as soon as xact's
18 : * first stream is received for subscriptions that have set their 'streaming'
19 : * option as parallel. The leader apply worker will send changes to this new
20 : * worker via shared memory. We keep this worker assigned till the transaction
21 : * commit is received and also wait for the worker to finish at commit. This
22 : * preserves commit ordering and avoid file I/O in most cases, although we
23 : * still need to spill to a file if there is no worker available. See comments
24 : * atop logical/worker to know more about streamed xacts whose changes are
25 : * spilled to disk. It is important to maintain commit order to avoid failures
26 : * due to: (a) transaction dependencies - say if we insert a row in the first
27 : * transaction and update it in the second transaction on publisher then
28 : * allowing the subscriber to apply both in parallel can lead to failure in the
29 : * update; (b) deadlocks - allowing transactions that update the same set of
30 : * rows/tables in the opposite order to be applied in parallel can lead to
31 : * deadlocks.
32 : *
33 : * A worker pool is used to avoid restarting workers for each streaming
34 : * transaction. We maintain each worker's information (ParallelApplyWorkerInfo)
35 : * in the ParallelApplyWorkerPool. After successfully launching a new worker,
36 : * its information is added to the ParallelApplyWorkerPool. Once the worker
37 : * finishes applying the transaction, it is marked as available for re-use.
38 : * Now, before starting a new worker to apply the streaming transaction, we
39 : * check the list for any available worker. Note that we retain a maximum of
40 : * half the max_parallel_apply_workers_per_subscription workers in the pool and
41 : * after that, we simply exit the worker after applying the transaction.
42 : *
43 : * XXX This worker pool threshold is arbitrary and we can provide a GUC
44 : * variable for this in the future if required.
45 : *
46 : * The leader apply worker will create a separate dynamic shared memory segment
47 : * when each parallel apply worker starts. The reason for this design is that
48 : * we cannot predict how many workers will be needed. It may be possible to
49 : * allocate enough shared memory in one segment based on the maximum number of
50 : * parallel apply workers (max_parallel_apply_workers_per_subscription), but
51 : * this would waste memory if no process is actually started.
52 : *
53 : * The dynamic shared memory segment contains: (a) a shm_mq that is used to
54 : * send changes in the transaction from leader apply worker to parallel apply
55 : * worker; (b) another shm_mq that is used to send errors (and other messages
56 : * reported via elog/ereport) from the parallel apply worker to leader apply
57 : * worker; (c) necessary information to be shared among parallel apply workers
58 : * and the leader apply worker (i.e. members of ParallelApplyWorkerShared).
59 : *
60 : * Locking Considerations
61 : * ----------------------
62 : * We have a risk of deadlock due to concurrently applying the transactions in
63 : * parallel mode that were independent on the publisher side but became
64 : * dependent on the subscriber side due to the different database structures
65 : * (like schema of subscription tables, constraints, etc.) on each side. This
66 : * can happen even without parallel mode when there are concurrent operations
67 : * on the subscriber. In order to detect the deadlocks among leader (LA) and
68 : * parallel apply (PA) workers, we used lmgr locks when the PA waits for the
69 : * next stream (set of changes) and LA waits for PA to finish the transaction.
70 : * An alternative approach could be to not allow parallelism when the schema of
71 : * tables is different between the publisher and subscriber but that would be
72 : * too restrictive and would require the publisher to send much more
73 : * information than it is currently sending.
74 : *
75 : * Consider a case where the subscribed table does not have a unique key on the
76 : * publisher and has a unique key on the subscriber. The deadlock can happen in
77 : * the following ways:
78 : *
79 : * 1) Deadlock between the leader apply worker and a parallel apply worker
80 : *
81 : * Consider that the parallel apply worker (PA) is executing TX-1 and the
82 : * leader apply worker (LA) is executing TX-2 concurrently on the subscriber.
83 : * Now, LA is waiting for PA because of the unique key constraint of the
84 : * subscribed table while PA is waiting for LA to send the next stream of
85 : * changes or transaction finish command message.
86 : *
87 : * In order for lmgr to detect this, we have LA acquire a session lock on the
88 : * remote transaction (by pa_lock_stream()) and have PA wait on the lock before
89 : * trying to receive the next stream of changes. Specifically, LA will acquire
90 : * the lock in AccessExclusive mode before sending the STREAM_STOP and will
91 : * release it if already acquired after sending the STREAM_START, STREAM_ABORT
92 : * (for toplevel transaction), STREAM_PREPARE, and STREAM_COMMIT. The PA will
93 : * acquire the lock in AccessShare mode after processing STREAM_STOP and
94 : * STREAM_ABORT (for subtransaction) and then release the lock immediately
95 : * after acquiring it.
96 : *
97 : * The lock graph for the above example will look as follows:
98 : * LA (waiting to acquire the lock on the unique index) -> PA (waiting to
99 : * acquire the stream lock) -> LA
100 : *
101 : * This way, when PA is waiting for LA for the next stream of changes, we can
102 : * have a wait-edge from PA to LA in lmgr, which will make us detect the
103 : * deadlock between LA and PA.
104 : *
105 : * 2) Deadlock between the leader apply worker and parallel apply workers
106 : *
107 : * This scenario is similar to the first case but TX-1 and TX-2 are executed by
108 : * two parallel apply workers (PA-1 and PA-2 respectively). In this scenario,
109 : * PA-2 is waiting for PA-1 to complete its transaction while PA-1 is waiting
110 : * for subsequent input from LA. Also, LA is waiting for PA-2 to complete its
111 : * transaction in order to preserve the commit order. There is a deadlock among
112 : * the three processes.
113 : *
114 : * In order for lmgr to detect this, we have PA acquire a session lock (this is
115 : * a different lock than referred in the previous case, see
116 : * pa_lock_transaction()) on the transaction being applied and have LA wait on
117 : * the lock before proceeding in the transaction finish commands. Specifically,
118 : * PA will acquire this lock in AccessExclusive mode before executing the first
119 : * message of the transaction and release it at the xact end. LA will acquire
120 : * this lock in AccessShare mode at transaction finish commands and release it
121 : * immediately.
122 : *
123 : * The lock graph for the above example will look as follows:
124 : * LA (waiting to acquire the transaction lock) -> PA-2 (waiting to acquire the
125 : * lock due to unique index constraint) -> PA-1 (waiting to acquire the stream
126 : * lock) -> LA
127 : *
128 : * This way when LA is waiting to finish the transaction end command to preserve
129 : * the commit order, we will be able to detect deadlock, if any.
130 : *
131 : * One might think we can use XactLockTableWait(), but XactLockTableWait()
132 : * considers PREPARED TRANSACTION as still in progress which means the lock
133 : * won't be released even after the parallel apply worker has prepared the
134 : * transaction.
135 : *
136 : * 3) Deadlock when the shm_mq buffer is full
137 : *
138 : * In the previous scenario (ie. PA-1 and PA-2 are executing transactions
139 : * concurrently), if the shm_mq buffer between LA and PA-2 is full, LA has to
140 : * wait to send messages, and this wait doesn't appear in lmgr.
141 : *
142 : * To avoid this wait, we use a non-blocking write and wait with a timeout. If
143 : * the timeout is exceeded, the LA will serialize all the pending messages to
144 : * a file and indicate PA-2 that it needs to read that file for the remaining
145 : * messages. Then LA will start waiting for commit as in the previous case
146 : * which will detect deadlock if any. See pa_send_data() and
147 : * enum TransApplyAction.
148 : *
149 : * Lock types
150 : * ----------
151 : * Both the stream lock and the transaction lock mentioned above are
152 : * session-level locks because both locks could be acquired outside the
153 : * transaction, and the stream lock in the leader needs to persist across
154 : * transaction boundaries i.e. until the end of the streaming transaction.
155 : *-------------------------------------------------------------------------
156 : */
157 :
158 : #include "postgres.h"
159 :
160 : #include "libpq/pqformat.h"
161 : #include "libpq/pqmq.h"
162 : #include "pgstat.h"
163 : #include "postmaster/interrupt.h"
164 : #include "replication/logicallauncher.h"
165 : #include "replication/logicalworker.h"
166 : #include "replication/origin.h"
167 : #include "replication/worker_internal.h"
168 : #include "storage/ipc.h"
169 : #include "storage/lmgr.h"
170 : #include "tcop/tcopprot.h"
171 : #include "utils/inval.h"
172 : #include "utils/memutils.h"
173 : #include "utils/syscache.h"
174 :
175 : #define PG_LOGICAL_APPLY_SHM_MAGIC 0x787ca067
176 :
177 : /*
178 : * DSM keys for parallel apply worker. Unlike other parallel execution code,
179 : * since we don't need to worry about DSM keys conflicting with plan_node_id we
180 : * can use small integers.
181 : */
182 : #define PARALLEL_APPLY_KEY_SHARED 1
183 : #define PARALLEL_APPLY_KEY_MQ 2
184 : #define PARALLEL_APPLY_KEY_ERROR_QUEUE 3
185 :
186 : /* Queue size of DSM, 16 MB for now. */
187 : #define DSM_QUEUE_SIZE (16 * 1024 * 1024)
188 :
189 : /*
190 : * Error queue size of DSM. It is desirable to make it large enough that a
191 : * typical ErrorResponse can be sent without blocking. That way, a worker that
192 : * errors out can write the whole message into the queue and terminate without
193 : * waiting for the user backend.
194 : */
195 : #define DSM_ERROR_QUEUE_SIZE (16 * 1024)
196 :
197 : /*
198 : * There are three fields in each message received by the parallel apply
199 : * worker: start_lsn, end_lsn and send_time. Because we have updated these
200 : * statistics in the leader apply worker, we can ignore these fields in the
201 : * parallel apply worker (see function LogicalRepApplyLoop).
202 : */
203 : #define SIZE_STATS_MESSAGE (2 * sizeof(XLogRecPtr) + sizeof(TimestampTz))
204 :
205 : /*
206 : * The type of session-level lock on a transaction being applied on a logical
207 : * replication subscriber.
208 : */
209 : #define PARALLEL_APPLY_LOCK_STREAM 0
210 : #define PARALLEL_APPLY_LOCK_XACT 1
211 :
212 : /*
213 : * Hash table entry to map xid to the parallel apply worker state.
214 : */
215 : typedef struct ParallelApplyWorkerEntry
216 : {
217 : TransactionId xid; /* Hash key -- must be first */
218 : ParallelApplyWorkerInfo *winfo;
219 : } ParallelApplyWorkerEntry;
220 :
221 : /*
222 : * A hash table used to cache the state of streaming transactions being applied
223 : * by the parallel apply workers.
224 : */
225 : static HTAB *ParallelApplyTxnHash = NULL;
226 :
227 : /*
228 : * A list (pool) of active parallel apply workers. The information for
229 : * the new worker is added to the list after successfully launching it. The
230 : * list entry is removed if there are already enough workers in the worker
231 : * pool at the end of the transaction. For more information about the worker
232 : * pool, see comments atop this file.
233 : */
234 : static List *ParallelApplyWorkerPool = NIL;
235 :
236 : /*
237 : * Information shared between leader apply worker and parallel apply worker.
238 : */
239 : ParallelApplyWorkerShared *MyParallelShared = NULL;
240 :
241 : /*
242 : * Is there a message sent by a parallel apply worker that the leader apply
243 : * worker needs to receive?
244 : */
245 : volatile sig_atomic_t ParallelApplyMessagePending = false;
246 :
247 : /*
248 : * Cache the parallel apply worker information required for applying the
249 : * current streaming transaction. It is used to save the cost of searching the
250 : * hash table when applying the changes between STREAM_START and STREAM_STOP.
251 : */
252 : static ParallelApplyWorkerInfo *stream_apply_worker = NULL;
253 :
254 : /* A list to maintain subtransactions, if any. */
255 : static List *subxactlist = NIL;
256 :
257 : static void pa_free_worker_info(ParallelApplyWorkerInfo *winfo);
258 : static ParallelTransState pa_get_xact_state(ParallelApplyWorkerShared *wshared);
259 : static PartialFileSetState pa_get_fileset_state(void);
260 :
261 : /*
262 : * Returns true if it is OK to start a parallel apply worker, false otherwise.
263 : */
264 : static bool
265 164 : pa_can_start(void)
266 : {
267 : /* Only leader apply workers can start parallel apply workers. */
268 164 : if (!am_leader_apply_worker())
269 54 : return false;
270 :
271 : /*
272 : * It is good to check for any change in the subscription parameter to
273 : * avoid the case where for a very long time the change doesn't get
274 : * reflected. This can happen when there is a constant flow of streaming
275 : * transactions that are handled by parallel apply workers.
276 : *
277 : * It is better to do it before the below checks so that the latest values
278 : * of subscription can be used for the checks.
279 : */
280 110 : maybe_reread_subscription();
281 :
282 : /*
283 : * Don't start a new parallel apply worker if the subscription is not
284 : * using parallel streaming mode, or if the publisher does not support
285 : * parallel apply.
286 : */
287 110 : if (!MyLogicalRepWorker->parallel_apply)
288 56 : return false;
289 :
290 : /*
291 : * Don't start a new parallel worker if user has set skiplsn as it's
292 : * possible that they want to skip the streaming transaction. For
293 : * streaming transactions, we need to serialize the transaction to a file
294 : * so that we can get the last LSN of the transaction to judge whether to
295 : * skip before starting to apply the change.
296 : *
297 : * One might think that we could allow parallelism if the first lsn of the
298 : * transaction is greater than skiplsn, but we don't send it with the
299 : * STREAM START message, and it doesn't seem worth sending the extra eight
300 : * bytes with the STREAM START to enable parallelism for this case.
301 : */
302 54 : if (!XLogRecPtrIsInvalid(MySubscription->skiplsn))
303 0 : return false;
304 :
305 : /*
306 : * For streaming transactions that are being applied using a parallel
307 : * apply worker, we cannot decide whether to apply the change for a
308 : * relation that is not in the READY state (see
309 : * should_apply_changes_for_rel) as we won't know remote_final_lsn by that
310 : * time. So, we don't start the new parallel apply worker in this case.
311 : */
312 54 : if (!AllTablesyncsReady())
313 0 : return false;
314 :
315 54 : return true;
316 : }
317 :
318 : /*
319 : * Set up a dynamic shared memory segment.
320 : *
321 : * We set up a control region that contains a fixed-size worker info
322 : * (ParallelApplyWorkerShared), a message queue, and an error queue.
323 : *
324 : * Returns true on success, false on failure.
325 : */
326 : static bool
327 20 : pa_setup_dsm(ParallelApplyWorkerInfo *winfo)
328 : {
329 : shm_toc_estimator e;
330 : Size segsize;
331 : dsm_segment *seg;
332 : shm_toc *toc;
333 : ParallelApplyWorkerShared *shared;
334 : shm_mq *mq;
335 20 : Size queue_size = DSM_QUEUE_SIZE;
336 20 : Size error_queue_size = DSM_ERROR_QUEUE_SIZE;
337 :
338 : /*
339 : * Estimate how much shared memory we need.
340 : *
341 : * Because the TOC machinery may choose to insert padding of oddly-sized
342 : * requests, we must estimate each chunk separately.
343 : *
344 : * We need one key to register the location of the header, and two other
345 : * keys to track the locations of the message queue and the error message
346 : * queue.
347 : */
348 20 : shm_toc_initialize_estimator(&e);
349 20 : shm_toc_estimate_chunk(&e, sizeof(ParallelApplyWorkerShared));
350 20 : shm_toc_estimate_chunk(&e, queue_size);
351 20 : shm_toc_estimate_chunk(&e, error_queue_size);
352 :
353 20 : shm_toc_estimate_keys(&e, 3);
354 20 : segsize = shm_toc_estimate(&e);
355 :
356 : /* Create the shared memory segment and establish a table of contents. */
357 20 : seg = dsm_create(shm_toc_estimate(&e), 0);
358 20 : if (!seg)
359 0 : return false;
360 :
361 20 : toc = shm_toc_create(PG_LOGICAL_APPLY_SHM_MAGIC, dsm_segment_address(seg),
362 : segsize);
363 :
364 : /* Set up the header region. */
365 20 : shared = shm_toc_allocate(toc, sizeof(ParallelApplyWorkerShared));
366 20 : SpinLockInit(&shared->mutex);
367 :
368 20 : shared->xact_state = PARALLEL_TRANS_UNKNOWN;
369 20 : pg_atomic_init_u32(&(shared->pending_stream_count), 0);
370 20 : shared->last_commit_end = InvalidXLogRecPtr;
371 20 : shared->fileset_state = FS_EMPTY;
372 :
373 20 : shm_toc_insert(toc, PARALLEL_APPLY_KEY_SHARED, shared);
374 :
375 : /* Set up message queue for the worker. */
376 20 : mq = shm_mq_create(shm_toc_allocate(toc, queue_size), queue_size);
377 20 : shm_toc_insert(toc, PARALLEL_APPLY_KEY_MQ, mq);
378 20 : shm_mq_set_sender(mq, MyProc);
379 :
380 : /* Attach the queue. */
381 20 : winfo->mq_handle = shm_mq_attach(mq, seg, NULL);
382 :
383 : /* Set up error queue for the worker. */
384 20 : mq = shm_mq_create(shm_toc_allocate(toc, error_queue_size),
385 : error_queue_size);
386 20 : shm_toc_insert(toc, PARALLEL_APPLY_KEY_ERROR_QUEUE, mq);
387 20 : shm_mq_set_receiver(mq, MyProc);
388 :
389 : /* Attach the queue. */
390 20 : winfo->error_mq_handle = shm_mq_attach(mq, seg, NULL);
391 :
392 : /* Return results to caller. */
393 20 : winfo->dsm_seg = seg;
394 20 : winfo->shared = shared;
395 :
396 20 : return true;
397 : }
398 :
399 : /*
400 : * Try to get a parallel apply worker from the pool. If none is available then
401 : * start a new one.
402 : */
403 : static ParallelApplyWorkerInfo *
404 54 : pa_launch_parallel_worker(void)
405 : {
406 : MemoryContext oldcontext;
407 : bool launched;
408 : ParallelApplyWorkerInfo *winfo;
409 : ListCell *lc;
410 :
411 : /* Try to get an available parallel apply worker from the worker pool. */
412 58 : foreach(lc, ParallelApplyWorkerPool)
413 : {
414 38 : winfo = (ParallelApplyWorkerInfo *) lfirst(lc);
415 :
416 38 : if (!winfo->in_use)
417 34 : return winfo;
418 : }
419 :
420 : /*
421 : * Start a new parallel apply worker.
422 : *
423 : * The worker info can be used for the lifetime of the worker process, so
424 : * create it in a permanent context.
425 : */
426 20 : oldcontext = MemoryContextSwitchTo(ApplyContext);
427 :
428 20 : winfo = (ParallelApplyWorkerInfo *) palloc0(sizeof(ParallelApplyWorkerInfo));
429 :
430 : /* Setup shared memory. */
431 20 : if (!pa_setup_dsm(winfo))
432 : {
433 0 : MemoryContextSwitchTo(oldcontext);
434 0 : pfree(winfo);
435 0 : return NULL;
436 : }
437 :
438 20 : launched = logicalrep_worker_launch(WORKERTYPE_PARALLEL_APPLY,
439 20 : MyLogicalRepWorker->dbid,
440 20 : MySubscription->oid,
441 20 : MySubscription->name,
442 20 : MyLogicalRepWorker->userid,
443 : InvalidOid,
444 : dsm_segment_handle(winfo->dsm_seg));
445 :
446 20 : if (launched)
447 : {
448 20 : ParallelApplyWorkerPool = lappend(ParallelApplyWorkerPool, winfo);
449 : }
450 : else
451 : {
452 0 : pa_free_worker_info(winfo);
453 0 : winfo = NULL;
454 : }
455 :
456 20 : MemoryContextSwitchTo(oldcontext);
457 :
458 20 : return winfo;
459 : }
460 :
461 : /*
462 : * Allocate a parallel apply worker that will be used for the specified xid.
463 : *
464 : * We first try to get an available worker from the pool, if any and then try
465 : * to launch a new worker. On successful allocation, remember the worker
466 : * information in the hash table so that we can get it later for processing the
467 : * streaming changes.
468 : */
469 : void
470 164 : pa_allocate_worker(TransactionId xid)
471 : {
472 : bool found;
473 164 : ParallelApplyWorkerInfo *winfo = NULL;
474 : ParallelApplyWorkerEntry *entry;
475 :
476 164 : if (!pa_can_start())
477 110 : return;
478 :
479 54 : winfo = pa_launch_parallel_worker();
480 54 : if (!winfo)
481 0 : return;
482 :
483 : /* First time through, initialize parallel apply worker state hashtable. */
484 54 : if (!ParallelApplyTxnHash)
485 : {
486 : HASHCTL ctl;
487 :
488 182 : MemSet(&ctl, 0, sizeof(ctl));
489 14 : ctl.keysize = sizeof(TransactionId);
490 14 : ctl.entrysize = sizeof(ParallelApplyWorkerEntry);
491 14 : ctl.hcxt = ApplyContext;
492 :
493 14 : ParallelApplyTxnHash = hash_create("logical replication parallel apply workers hash",
494 : 16, &ctl,
495 : HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
496 : }
497 :
498 : /* Create an entry for the requested transaction. */
499 54 : entry = hash_search(ParallelApplyTxnHash, &xid, HASH_ENTER, &found);
500 54 : if (found)
501 0 : elog(ERROR, "hash table corrupted");
502 :
503 : /* Update the transaction information in shared memory. */
504 54 : SpinLockAcquire(&winfo->shared->mutex);
505 54 : winfo->shared->xact_state = PARALLEL_TRANS_UNKNOWN;
506 54 : winfo->shared->xid = xid;
507 54 : SpinLockRelease(&winfo->shared->mutex);
508 :
509 54 : winfo->in_use = true;
510 54 : winfo->serialize_changes = false;
511 54 : entry->winfo = winfo;
512 : }
513 :
514 : /*
515 : * Find the assigned worker for the given transaction, if any.
516 : */
517 : ParallelApplyWorkerInfo *
518 514382 : pa_find_worker(TransactionId xid)
519 : {
520 : bool found;
521 : ParallelApplyWorkerEntry *entry;
522 :
523 514382 : if (!TransactionIdIsValid(xid))
524 160042 : return NULL;
525 :
526 354340 : if (!ParallelApplyTxnHash)
527 206478 : return NULL;
528 :
529 : /* Return the cached parallel apply worker if valid. */
530 147862 : if (stream_apply_worker)
531 147288 : return stream_apply_worker;
532 :
533 : /* Find an entry for the requested transaction. */
534 574 : entry = hash_search(ParallelApplyTxnHash, &xid, HASH_FIND, &found);
535 574 : if (found)
536 : {
537 : /* The worker must not have exited. */
538 : Assert(entry->winfo->in_use);
539 574 : return entry->winfo;
540 : }
541 :
542 0 : return NULL;
543 : }
544 :
545 : /*
546 : * Makes the worker available for reuse.
547 : *
548 : * This removes the parallel apply worker entry from the hash table so that it
549 : * can't be used. If there are enough workers in the pool, it stops the worker
550 : * and frees the corresponding info. Otherwise it just marks the worker as
551 : * available for reuse.
552 : *
553 : * For more information about the worker pool, see comments atop this file.
554 : */
555 : static void
556 48 : pa_free_worker(ParallelApplyWorkerInfo *winfo)
557 : {
558 : Assert(!am_parallel_apply_worker());
559 : Assert(winfo->in_use);
560 : Assert(pa_get_xact_state(winfo->shared) == PARALLEL_TRANS_FINISHED);
561 :
562 48 : if (!hash_search(ParallelApplyTxnHash, &winfo->shared->xid, HASH_REMOVE, NULL))
563 0 : elog(ERROR, "hash table corrupted");
564 :
565 : /*
566 : * Stop the worker if there are enough workers in the pool.
567 : *
568 : * XXX Additionally, we also stop the worker if the leader apply worker
569 : * serialize part of the transaction data due to a send timeout. This is
570 : * because the message could be partially written to the queue and there
571 : * is no way to clean the queue other than resending the message until it
572 : * succeeds. Instead of trying to send the data which anyway would have
573 : * been serialized and then letting the parallel apply worker deal with
574 : * the spurious message, we stop the worker.
575 : */
576 48 : if (winfo->serialize_changes ||
577 40 : list_length(ParallelApplyWorkerPool) >
578 40 : (max_parallel_apply_workers_per_subscription / 2))
579 : {
580 10 : logicalrep_pa_worker_stop(winfo);
581 10 : pa_free_worker_info(winfo);
582 :
583 10 : return;
584 : }
585 :
586 38 : winfo->in_use = false;
587 38 : winfo->serialize_changes = false;
588 : }
589 :
590 : /*
591 : * Free the parallel apply worker information and unlink the files with
592 : * serialized changes if any.
593 : */
594 : static void
595 10 : pa_free_worker_info(ParallelApplyWorkerInfo *winfo)
596 : {
597 : Assert(winfo);
598 :
599 10 : if (winfo->mq_handle)
600 10 : shm_mq_detach(winfo->mq_handle);
601 :
602 10 : if (winfo->error_mq_handle)
603 0 : shm_mq_detach(winfo->error_mq_handle);
604 :
605 : /* Unlink the files with serialized changes. */
606 10 : if (winfo->serialize_changes)
607 8 : stream_cleanup_files(MyLogicalRepWorker->subid, winfo->shared->xid);
608 :
609 10 : if (winfo->dsm_seg)
610 10 : dsm_detach(winfo->dsm_seg);
611 :
612 : /* Remove from the worker pool. */
613 10 : ParallelApplyWorkerPool = list_delete_ptr(ParallelApplyWorkerPool, winfo);
614 :
615 10 : pfree(winfo);
616 10 : }
617 :
618 : /*
619 : * Detach the error queue for all parallel apply workers.
620 : */
621 : void
622 398 : pa_detach_all_error_mq(void)
623 : {
624 : ListCell *lc;
625 :
626 408 : foreach(lc, ParallelApplyWorkerPool)
627 : {
628 10 : ParallelApplyWorkerInfo *winfo = (ParallelApplyWorkerInfo *) lfirst(lc);
629 :
630 10 : if (winfo->error_mq_handle)
631 : {
632 10 : shm_mq_detach(winfo->error_mq_handle);
633 10 : winfo->error_mq_handle = NULL;
634 : }
635 : }
636 398 : }
637 :
638 : /*
639 : * Check if there are any pending spooled messages.
640 : */
641 : static bool
642 32 : pa_has_spooled_message_pending()
643 : {
644 : PartialFileSetState fileset_state;
645 :
646 32 : fileset_state = pa_get_fileset_state();
647 :
648 32 : return (fileset_state != FS_EMPTY);
649 : }
650 :
651 : /*
652 : * Replay the spooled messages once the leader apply worker has finished
653 : * serializing changes to the file.
654 : *
655 : * Returns false if there aren't any pending spooled messages, true otherwise.
656 : */
657 : static bool
658 114 : pa_process_spooled_messages_if_required(void)
659 : {
660 : PartialFileSetState fileset_state;
661 :
662 114 : fileset_state = pa_get_fileset_state();
663 :
664 114 : if (fileset_state == FS_EMPTY)
665 98 : return false;
666 :
667 : /*
668 : * If the leader apply worker is busy serializing the partial changes then
669 : * acquire the stream lock now and wait for the leader worker to finish
670 : * serializing the changes. Otherwise, the parallel apply worker won't get
671 : * a chance to receive a STREAM_STOP (and acquire the stream lock) until
672 : * the leader had serialized all changes which can lead to undetected
673 : * deadlock.
674 : *
675 : * Note that the fileset state can be FS_SERIALIZE_DONE once the leader
676 : * worker has finished serializing the changes.
677 : */
678 16 : if (fileset_state == FS_SERIALIZE_IN_PROGRESS)
679 : {
680 0 : pa_lock_stream(MyParallelShared->xid, AccessShareLock);
681 0 : pa_unlock_stream(MyParallelShared->xid, AccessShareLock);
682 :
683 0 : fileset_state = pa_get_fileset_state();
684 : }
685 :
686 : /*
687 : * We cannot read the file immediately after the leader has serialized all
688 : * changes to the file because there may still be messages in the memory
689 : * queue. We will apply all spooled messages the next time we call this
690 : * function and that will ensure there are no messages left in the memory
691 : * queue.
692 : */
693 16 : if (fileset_state == FS_SERIALIZE_DONE)
694 : {
695 8 : pa_set_fileset_state(MyParallelShared, FS_READY);
696 : }
697 8 : else if (fileset_state == FS_READY)
698 : {
699 8 : apply_spooled_messages(&MyParallelShared->fileset,
700 8 : MyParallelShared->xid,
701 : InvalidXLogRecPtr);
702 8 : pa_set_fileset_state(MyParallelShared, FS_EMPTY);
703 : }
704 :
705 16 : return true;
706 : }
707 :
708 : /*
709 : * Interrupt handler for main loop of parallel apply worker.
710 : */
711 : static void
712 128932 : ProcessParallelApplyInterrupts(void)
713 : {
714 128932 : CHECK_FOR_INTERRUPTS();
715 :
716 128926 : if (ShutdownRequestPending)
717 : {
718 10 : ereport(LOG,
719 : (errmsg("logical replication parallel apply worker for subscription \"%s\" has finished",
720 : MySubscription->name)));
721 :
722 10 : proc_exit(0);
723 : }
724 :
725 128916 : if (ConfigReloadPending)
726 : {
727 8 : ConfigReloadPending = false;
728 8 : ProcessConfigFile(PGC_SIGHUP);
729 : }
730 128916 : }
731 :
732 : /* Parallel apply worker main loop. */
733 : static void
734 20 : LogicalParallelApplyLoop(shm_mq_handle *mqh)
735 : {
736 : shm_mq_result shmq_res;
737 : ErrorContextCallback errcallback;
738 20 : MemoryContext oldcxt = CurrentMemoryContext;
739 :
740 : /*
741 : * Init the ApplyMessageContext which we clean up after each replication
742 : * protocol message.
743 : */
744 20 : ApplyMessageContext = AllocSetContextCreate(ApplyContext,
745 : "ApplyMessageContext",
746 : ALLOCSET_DEFAULT_SIZES);
747 :
748 : /*
749 : * Push apply error context callback. Fields will be filled while applying
750 : * a change.
751 : */
752 20 : errcallback.callback = apply_error_callback;
753 20 : errcallback.previous = error_context_stack;
754 20 : error_context_stack = &errcallback;
755 :
756 : for (;;)
757 128912 : {
758 : void *data;
759 : Size len;
760 :
761 128932 : ProcessParallelApplyInterrupts();
762 :
763 : /* Ensure we are reading the data into our memory context. */
764 128916 : MemoryContextSwitchTo(ApplyMessageContext);
765 :
766 128916 : shmq_res = shm_mq_receive(mqh, &len, &data, true);
767 :
768 128916 : if (shmq_res == SHM_MQ_SUCCESS)
769 : {
770 : StringInfoData s;
771 : int c;
772 :
773 128802 : if (len == 0)
774 0 : elog(ERROR, "invalid message length");
775 :
776 128802 : initReadOnlyStringInfo(&s, data, len);
777 :
778 : /*
779 : * The first byte of messages sent from leader apply worker to
780 : * parallel apply workers can only be 'w'.
781 : */
782 128802 : c = pq_getmsgbyte(&s);
783 128802 : if (c != 'w')
784 0 : elog(ERROR, "unexpected message \"%c\"", c);
785 :
786 : /*
787 : * Ignore statistics fields that have been updated by the leader
788 : * apply worker.
789 : *
790 : * XXX We can avoid sending the statistics fields from the leader
791 : * apply worker but for that, it needs to rebuild the entire
792 : * message by removing these fields which could be more work than
793 : * simply ignoring these fields in the parallel apply worker.
794 : */
795 128802 : s.cursor += SIZE_STATS_MESSAGE;
796 :
797 128802 : apply_dispatch(&s);
798 : }
799 114 : else if (shmq_res == SHM_MQ_WOULD_BLOCK)
800 : {
801 : /* Replay the changes from the file, if any. */
802 114 : if (!pa_process_spooled_messages_if_required())
803 : {
804 : int rc;
805 :
806 : /* Wait for more work. */
807 98 : rc = WaitLatch(MyLatch,
808 : WL_LATCH_SET | WL_TIMEOUT | WL_EXIT_ON_PM_DEATH,
809 : 1000L,
810 : WAIT_EVENT_LOGICAL_PARALLEL_APPLY_MAIN);
811 :
812 98 : if (rc & WL_LATCH_SET)
813 92 : ResetLatch(MyLatch);
814 : }
815 : }
816 : else
817 : {
818 : Assert(shmq_res == SHM_MQ_DETACHED);
819 :
820 0 : ereport(ERROR,
821 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
822 : errmsg("lost connection to the logical replication apply worker")));
823 : }
824 :
825 128912 : MemoryContextReset(ApplyMessageContext);
826 128912 : MemoryContextSwitchTo(oldcxt);
827 : }
828 :
829 : /* Pop the error context stack. */
830 : error_context_stack = errcallback.previous;
831 :
832 : MemoryContextSwitchTo(oldcxt);
833 : }
834 :
835 : /*
836 : * Make sure the leader apply worker tries to read from our error queue one more
837 : * time. This guards against the case where we exit uncleanly without sending
838 : * an ErrorResponse, for example because some code calls proc_exit directly.
839 : *
840 : * Also explicitly detach from dsm segment to invoke on_dsm_detach callbacks,
841 : * if any. See ParallelWorkerShutdown for details.
842 : */
843 : static void
844 20 : pa_shutdown(int code, Datum arg)
845 : {
846 20 : SendProcSignal(MyLogicalRepWorker->leader_pid,
847 : PROCSIG_PARALLEL_APPLY_MESSAGE,
848 : INVALID_PROC_NUMBER);
849 :
850 20 : dsm_detach((dsm_segment *) DatumGetPointer(arg));
851 20 : }
852 :
853 : /*
854 : * Parallel apply worker entry point.
855 : */
856 : void
857 20 : ParallelApplyWorkerMain(Datum main_arg)
858 : {
859 : ParallelApplyWorkerShared *shared;
860 : dsm_handle handle;
861 : dsm_segment *seg;
862 : shm_toc *toc;
863 : shm_mq *mq;
864 : shm_mq_handle *mqh;
865 : shm_mq_handle *error_mqh;
866 : RepOriginId originid;
867 20 : int worker_slot = DatumGetInt32(main_arg);
868 : char originname[NAMEDATALEN];
869 :
870 20 : InitializingApplyWorker = true;
871 :
872 : /* Setup signal handling. */
873 20 : pqsignal(SIGHUP, SignalHandlerForConfigReload);
874 20 : pqsignal(SIGINT, SignalHandlerForShutdownRequest);
875 20 : pqsignal(SIGTERM, die);
876 20 : BackgroundWorkerUnblockSignals();
877 :
878 : /*
879 : * Attach to the dynamic shared memory segment for the parallel apply, and
880 : * find its table of contents.
881 : *
882 : * Like parallel query, we don't need resource owner by this time. See
883 : * ParallelWorkerMain.
884 : */
885 20 : memcpy(&handle, MyBgworkerEntry->bgw_extra, sizeof(dsm_handle));
886 20 : seg = dsm_attach(handle);
887 20 : if (!seg)
888 0 : ereport(ERROR,
889 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
890 : errmsg("could not map dynamic shared memory segment")));
891 :
892 20 : toc = shm_toc_attach(PG_LOGICAL_APPLY_SHM_MAGIC, dsm_segment_address(seg));
893 20 : if (!toc)
894 0 : ereport(ERROR,
895 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
896 : errmsg("invalid magic number in dynamic shared memory segment")));
897 :
898 : /* Look up the shared information. */
899 20 : shared = shm_toc_lookup(toc, PARALLEL_APPLY_KEY_SHARED, false);
900 20 : MyParallelShared = shared;
901 :
902 : /*
903 : * Attach to the message queue.
904 : */
905 20 : mq = shm_toc_lookup(toc, PARALLEL_APPLY_KEY_MQ, false);
906 20 : shm_mq_set_receiver(mq, MyProc);
907 20 : mqh = shm_mq_attach(mq, seg, NULL);
908 :
909 : /*
910 : * Primary initialization is complete. Now, we can attach to our slot.
911 : * This is to ensure that the leader apply worker does not write data to
912 : * the uninitialized memory queue.
913 : */
914 20 : logicalrep_worker_attach(worker_slot);
915 :
916 : /*
917 : * Register the shutdown callback after we are attached to the worker
918 : * slot. This is to ensure that MyLogicalRepWorker remains valid when this
919 : * callback is invoked.
920 : */
921 20 : before_shmem_exit(pa_shutdown, PointerGetDatum(seg));
922 :
923 20 : SpinLockAcquire(&MyParallelShared->mutex);
924 20 : MyParallelShared->logicalrep_worker_generation = MyLogicalRepWorker->generation;
925 20 : MyParallelShared->logicalrep_worker_slot_no = worker_slot;
926 20 : SpinLockRelease(&MyParallelShared->mutex);
927 :
928 : /*
929 : * Attach to the error queue.
930 : */
931 20 : mq = shm_toc_lookup(toc, PARALLEL_APPLY_KEY_ERROR_QUEUE, false);
932 20 : shm_mq_set_sender(mq, MyProc);
933 20 : error_mqh = shm_mq_attach(mq, seg, NULL);
934 :
935 20 : pq_redirect_to_shm_mq(seg, error_mqh);
936 20 : pq_set_parallel_leader(MyLogicalRepWorker->leader_pid,
937 : INVALID_PROC_NUMBER);
938 :
939 20 : MyLogicalRepWorker->last_send_time = MyLogicalRepWorker->last_recv_time =
940 20 : MyLogicalRepWorker->reply_time = 0;
941 :
942 20 : InitializeLogRepWorker();
943 :
944 20 : InitializingApplyWorker = false;
945 :
946 : /* Setup replication origin tracking. */
947 20 : StartTransactionCommand();
948 20 : ReplicationOriginNameForLogicalRep(MySubscription->oid, InvalidOid,
949 : originname, sizeof(originname));
950 20 : originid = replorigin_by_name(originname, false);
951 :
952 : /*
953 : * The parallel apply worker doesn't need to monopolize this replication
954 : * origin which was already acquired by its leader process.
955 : */
956 20 : replorigin_session_setup(originid, MyLogicalRepWorker->leader_pid);
957 20 : replorigin_session_origin = originid;
958 20 : CommitTransactionCommand();
959 :
960 : /*
961 : * Setup callback for syscache so that we know when something changes in
962 : * the subscription relation state.
963 : */
964 20 : CacheRegisterSyscacheCallback(SUBSCRIPTIONRELMAP,
965 : invalidate_syncing_table_states,
966 : (Datum) 0);
967 :
968 20 : set_apply_error_context_origin(originname);
969 :
970 20 : LogicalParallelApplyLoop(mqh);
971 :
972 : /*
973 : * The parallel apply worker must not get here because the parallel apply
974 : * worker will only stop when it receives a SIGTERM or SIGINT from the
975 : * leader, or when there is an error. None of these cases will allow the
976 : * code to reach here.
977 : */
978 : Assert(false);
979 0 : }
980 :
981 : /*
982 : * Handle receipt of an interrupt indicating a parallel apply worker message.
983 : *
984 : * Note: this is called within a signal handler! All we can do is set a flag
985 : * that will cause the next CHECK_FOR_INTERRUPTS() to invoke
986 : * HandleParallelApplyMessages().
987 : */
988 : void
989 28 : HandleParallelApplyMessageInterrupt(void)
990 : {
991 28 : InterruptPending = true;
992 28 : ParallelApplyMessagePending = true;
993 28 : SetLatch(MyLatch);
994 28 : }
995 :
996 : /*
997 : * Handle a single protocol message received from a single parallel apply
998 : * worker.
999 : */
1000 : static void
1001 2 : HandleParallelApplyMessage(StringInfo msg)
1002 : {
1003 : char msgtype;
1004 :
1005 2 : msgtype = pq_getmsgbyte(msg);
1006 :
1007 2 : switch (msgtype)
1008 : {
1009 2 : case 'E': /* ErrorResponse */
1010 : {
1011 : ErrorData edata;
1012 :
1013 : /* Parse ErrorResponse. */
1014 2 : pq_parse_errornotice(msg, &edata);
1015 :
1016 : /*
1017 : * If desired, add a context line to show that this is a
1018 : * message propagated from a parallel apply worker. Otherwise,
1019 : * it can sometimes be confusing to understand what actually
1020 : * happened.
1021 : */
1022 2 : if (edata.context)
1023 2 : edata.context = psprintf("%s\n%s", edata.context,
1024 : _("logical replication parallel apply worker"));
1025 : else
1026 0 : edata.context = pstrdup(_("logical replication parallel apply worker"));
1027 :
1028 : /*
1029 : * Context beyond that should use the error context callbacks
1030 : * that were in effect in LogicalRepApplyLoop().
1031 : */
1032 2 : error_context_stack = apply_error_context_stack;
1033 :
1034 : /*
1035 : * The actual error must have been reported by the parallel
1036 : * apply worker.
1037 : */
1038 2 : ereport(ERROR,
1039 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1040 : errmsg("logical replication parallel apply worker exited due to error"),
1041 : errcontext("%s", edata.context)));
1042 : }
1043 :
1044 : /*
1045 : * Don't need to do anything about NoticeResponse and
1046 : * NotifyResponse as the logical replication worker doesn't need
1047 : * to send messages to the client.
1048 : */
1049 0 : case 'N':
1050 : case 'A':
1051 0 : break;
1052 :
1053 0 : default:
1054 0 : elog(ERROR, "unrecognized message type received from logical replication parallel apply worker: %c (message length %d bytes)",
1055 : msgtype, msg->len);
1056 : }
1057 0 : }
1058 :
1059 : /*
1060 : * Handle any queued protocol messages received from parallel apply workers.
1061 : */
1062 : void
1063 12 : HandleParallelApplyMessages(void)
1064 : {
1065 : ListCell *lc;
1066 : MemoryContext oldcontext;
1067 :
1068 : static MemoryContext hpam_context = NULL;
1069 :
1070 : /*
1071 : * This is invoked from ProcessInterrupts(), and since some of the
1072 : * functions it calls contain CHECK_FOR_INTERRUPTS(), there is a potential
1073 : * for recursive calls if more signals are received while this runs. It's
1074 : * unclear that recursive entry would be safe, and it doesn't seem useful
1075 : * even if it is safe, so let's block interrupts until done.
1076 : */
1077 12 : HOLD_INTERRUPTS();
1078 :
1079 : /*
1080 : * Moreover, CurrentMemoryContext might be pointing almost anywhere. We
1081 : * don't want to risk leaking data into long-lived contexts, so let's do
1082 : * our work here in a private context that we can reset on each use.
1083 : */
1084 12 : if (!hpam_context) /* first time through? */
1085 10 : hpam_context = AllocSetContextCreate(TopMemoryContext,
1086 : "HandleParallelApplyMessages",
1087 : ALLOCSET_DEFAULT_SIZES);
1088 : else
1089 2 : MemoryContextReset(hpam_context);
1090 :
1091 12 : oldcontext = MemoryContextSwitchTo(hpam_context);
1092 :
1093 12 : ParallelApplyMessagePending = false;
1094 :
1095 24 : foreach(lc, ParallelApplyWorkerPool)
1096 : {
1097 : shm_mq_result res;
1098 : Size nbytes;
1099 : void *data;
1100 14 : ParallelApplyWorkerInfo *winfo = (ParallelApplyWorkerInfo *) lfirst(lc);
1101 :
1102 : /*
1103 : * The leader will detach from the error queue and set it to NULL
1104 : * before preparing to stop all parallel apply workers, so we don't
1105 : * need to handle error messages anymore. See
1106 : * logicalrep_worker_detach.
1107 : */
1108 14 : if (!winfo->error_mq_handle)
1109 12 : continue;
1110 :
1111 4 : res = shm_mq_receive(winfo->error_mq_handle, &nbytes, &data, true);
1112 :
1113 4 : if (res == SHM_MQ_WOULD_BLOCK)
1114 2 : continue;
1115 2 : else if (res == SHM_MQ_SUCCESS)
1116 : {
1117 : StringInfoData msg;
1118 :
1119 2 : initStringInfo(&msg);
1120 2 : appendBinaryStringInfo(&msg, data, nbytes);
1121 2 : HandleParallelApplyMessage(&msg);
1122 0 : pfree(msg.data);
1123 : }
1124 : else
1125 0 : ereport(ERROR,
1126 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1127 : errmsg("lost connection to the logical replication parallel apply worker")));
1128 : }
1129 :
1130 10 : MemoryContextSwitchTo(oldcontext);
1131 :
1132 : /* Might as well clear the context on our way out */
1133 10 : MemoryContextReset(hpam_context);
1134 :
1135 10 : RESUME_INTERRUPTS();
1136 10 : }
1137 :
1138 : /*
1139 : * Send the data to the specified parallel apply worker via shared-memory
1140 : * queue.
1141 : *
1142 : * Returns false if the attempt to send data via shared memory times out, true
1143 : * otherwise.
1144 : */
1145 : bool
1146 137788 : pa_send_data(ParallelApplyWorkerInfo *winfo, Size nbytes, const void *data)
1147 : {
1148 : int rc;
1149 : shm_mq_result result;
1150 137788 : TimestampTz startTime = 0;
1151 :
1152 : Assert(!IsTransactionState());
1153 : Assert(!winfo->serialize_changes);
1154 :
1155 : /*
1156 : * We don't try to send data to parallel worker for 'immediate' mode. This
1157 : * is primarily used for testing purposes.
1158 : */
1159 137788 : if (unlikely(debug_logical_replication_streaming == DEBUG_LOGICAL_REP_STREAMING_IMMEDIATE))
1160 8 : return false;
1161 :
1162 : /*
1163 : * This timeout is a bit arbitrary but testing revealed that it is sufficient
1164 : * to send the message unless the parallel apply worker is waiting on some
1165 : * lock or there is a serious resource crunch. See the comments atop this file
1166 : * to know why we are using a non-blocking way to send the message.
1167 : */
1168 : #define SHM_SEND_RETRY_INTERVAL_MS 1000
1169 : #define SHM_SEND_TIMEOUT_MS (10000 - SHM_SEND_RETRY_INTERVAL_MS)
1170 :
1171 : for (;;)
1172 : {
1173 137780 : result = shm_mq_send(winfo->mq_handle, nbytes, data, true, true);
1174 :
1175 137780 : if (result == SHM_MQ_SUCCESS)
1176 137780 : return true;
1177 0 : else if (result == SHM_MQ_DETACHED)
1178 0 : ereport(ERROR,
1179 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1180 : errmsg("could not send data to shared-memory queue")));
1181 :
1182 : Assert(result == SHM_MQ_WOULD_BLOCK);
1183 :
1184 : /* Wait before retrying. */
1185 0 : rc = WaitLatch(MyLatch,
1186 : WL_LATCH_SET | WL_TIMEOUT | WL_EXIT_ON_PM_DEATH,
1187 : SHM_SEND_RETRY_INTERVAL_MS,
1188 : WAIT_EVENT_LOGICAL_APPLY_SEND_DATA);
1189 :
1190 0 : if (rc & WL_LATCH_SET)
1191 : {
1192 0 : ResetLatch(MyLatch);
1193 0 : CHECK_FOR_INTERRUPTS();
1194 : }
1195 :
1196 0 : if (startTime == 0)
1197 0 : startTime = GetCurrentTimestamp();
1198 0 : else if (TimestampDifferenceExceeds(startTime, GetCurrentTimestamp(),
1199 : SHM_SEND_TIMEOUT_MS))
1200 0 : return false;
1201 : }
1202 : }
1203 :
1204 : /*
1205 : * Switch to PARTIAL_SERIALIZE mode for the current transaction -- this means
1206 : * that the current data and any subsequent data for this transaction will be
1207 : * serialized to a file. This is done to prevent possible deadlocks with
1208 : * another parallel apply worker (refer to the comments atop this file).
1209 : */
1210 : void
1211 8 : pa_switch_to_partial_serialize(ParallelApplyWorkerInfo *winfo,
1212 : bool stream_locked)
1213 : {
1214 8 : ereport(LOG,
1215 : (errmsg("logical replication apply worker will serialize the remaining changes of remote transaction %u to a file",
1216 : winfo->shared->xid)));
1217 :
1218 : /*
1219 : * The parallel apply worker could be stuck for some reason (say waiting
1220 : * on some lock by other backend), so stop trying to send data directly to
1221 : * it and start serializing data to the file instead.
1222 : */
1223 8 : winfo->serialize_changes = true;
1224 :
1225 : /* Initialize the stream fileset. */
1226 8 : stream_start_internal(winfo->shared->xid, true);
1227 :
1228 : /*
1229 : * Acquires the stream lock if not already to make sure that the parallel
1230 : * apply worker will wait for the leader to release the stream lock until
1231 : * the end of the transaction.
1232 : */
1233 8 : if (!stream_locked)
1234 8 : pa_lock_stream(winfo->shared->xid, AccessExclusiveLock);
1235 :
1236 8 : pa_set_fileset_state(winfo->shared, FS_SERIALIZE_IN_PROGRESS);
1237 8 : }
1238 :
1239 : /*
1240 : * Wait until the parallel apply worker's transaction state has reached or
1241 : * exceeded the given xact_state.
1242 : */
1243 : static void
1244 588 : pa_wait_for_xact_state(ParallelApplyWorkerInfo *winfo,
1245 : ParallelTransState xact_state)
1246 : {
1247 : for (;;)
1248 : {
1249 : /*
1250 : * Stop if the transaction state has reached or exceeded the given
1251 : * xact_state.
1252 : */
1253 588 : if (pa_get_xact_state(winfo->shared) >= xact_state)
1254 50 : break;
1255 :
1256 : /* Wait to be signalled. */
1257 538 : (void) WaitLatch(MyLatch,
1258 : WL_LATCH_SET | WL_TIMEOUT | WL_EXIT_ON_PM_DEATH,
1259 : 10L,
1260 : WAIT_EVENT_LOGICAL_PARALLEL_APPLY_STATE_CHANGE);
1261 :
1262 : /* Reset the latch so we don't spin. */
1263 538 : ResetLatch(MyLatch);
1264 :
1265 : /* An interrupt may have occurred while we were waiting. */
1266 538 : CHECK_FOR_INTERRUPTS();
1267 : }
1268 50 : }
1269 :
1270 : /*
1271 : * Wait until the parallel apply worker's transaction finishes.
1272 : */
1273 : static void
1274 50 : pa_wait_for_xact_finish(ParallelApplyWorkerInfo *winfo)
1275 : {
1276 : /*
1277 : * Wait until the parallel apply worker set the state to
1278 : * PARALLEL_TRANS_STARTED which means it has acquired the transaction
1279 : * lock. This is to prevent leader apply worker from acquiring the
1280 : * transaction lock earlier than the parallel apply worker.
1281 : */
1282 50 : pa_wait_for_xact_state(winfo, PARALLEL_TRANS_STARTED);
1283 :
1284 : /*
1285 : * Wait for the transaction lock to be released. This is required to
1286 : * detect deadlock among leader and parallel apply workers. Refer to the
1287 : * comments atop this file.
1288 : */
1289 50 : pa_lock_transaction(winfo->shared->xid, AccessShareLock);
1290 48 : pa_unlock_transaction(winfo->shared->xid, AccessShareLock);
1291 :
1292 : /*
1293 : * Check if the state becomes PARALLEL_TRANS_FINISHED in case the parallel
1294 : * apply worker failed while applying changes causing the lock to be
1295 : * released.
1296 : */
1297 48 : if (pa_get_xact_state(winfo->shared) != PARALLEL_TRANS_FINISHED)
1298 0 : ereport(ERROR,
1299 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1300 : errmsg("lost connection to the logical replication parallel apply worker")));
1301 48 : }
1302 :
1303 : /*
1304 : * Set the transaction state for a given parallel apply worker.
1305 : */
1306 : void
1307 102 : pa_set_xact_state(ParallelApplyWorkerShared *wshared,
1308 : ParallelTransState xact_state)
1309 : {
1310 102 : SpinLockAcquire(&wshared->mutex);
1311 102 : wshared->xact_state = xact_state;
1312 102 : SpinLockRelease(&wshared->mutex);
1313 102 : }
1314 :
1315 : /*
1316 : * Get the transaction state for a given parallel apply worker.
1317 : */
1318 : static ParallelTransState
1319 636 : pa_get_xact_state(ParallelApplyWorkerShared *wshared)
1320 : {
1321 : ParallelTransState xact_state;
1322 :
1323 636 : SpinLockAcquire(&wshared->mutex);
1324 636 : xact_state = wshared->xact_state;
1325 636 : SpinLockRelease(&wshared->mutex);
1326 :
1327 636 : return xact_state;
1328 : }
1329 :
1330 : /*
1331 : * Cache the parallel apply worker information.
1332 : */
1333 : void
1334 1008 : pa_set_stream_apply_worker(ParallelApplyWorkerInfo *winfo)
1335 : {
1336 1008 : stream_apply_worker = winfo;
1337 1008 : }
1338 :
1339 : /*
1340 : * Form a unique savepoint name for the streaming transaction.
1341 : *
1342 : * Note that different subscriptions for publications on different nodes can
1343 : * receive same remote xid, so we need to use subscription id along with it.
1344 : *
1345 : * Returns the name in the supplied buffer.
1346 : */
1347 : static void
1348 54 : pa_savepoint_name(Oid suboid, TransactionId xid, char *spname, Size szsp)
1349 : {
1350 54 : snprintf(spname, szsp, "pg_sp_%u_%u", suboid, xid);
1351 54 : }
1352 :
1353 : /*
1354 : * Define a savepoint for a subxact in parallel apply worker if needed.
1355 : *
1356 : * The parallel apply worker can figure out if a new subtransaction was
1357 : * started by checking if the new change arrived with a different xid. In that
1358 : * case define a named savepoint, so that we are able to rollback to it
1359 : * if required.
1360 : */
1361 : void
1362 137846 : pa_start_subtrans(TransactionId current_xid, TransactionId top_xid)
1363 : {
1364 137846 : if (current_xid != top_xid &&
1365 104 : !list_member_xid(subxactlist, current_xid))
1366 : {
1367 : MemoryContext oldctx;
1368 : char spname[NAMEDATALEN];
1369 :
1370 34 : pa_savepoint_name(MySubscription->oid, current_xid,
1371 : spname, sizeof(spname));
1372 :
1373 34 : elog(DEBUG1, "defining savepoint %s in logical replication parallel apply worker", spname);
1374 :
1375 : /* We must be in transaction block to define the SAVEPOINT. */
1376 34 : if (!IsTransactionBlock())
1377 : {
1378 10 : if (!IsTransactionState())
1379 0 : StartTransactionCommand();
1380 :
1381 10 : BeginTransactionBlock();
1382 10 : CommitTransactionCommand();
1383 : }
1384 :
1385 34 : DefineSavepoint(spname);
1386 :
1387 : /*
1388 : * CommitTransactionCommand is needed to start a subtransaction after
1389 : * issuing a SAVEPOINT inside a transaction block (see
1390 : * StartSubTransaction()).
1391 : */
1392 34 : CommitTransactionCommand();
1393 :
1394 34 : oldctx = MemoryContextSwitchTo(TopTransactionContext);
1395 34 : subxactlist = lappend_xid(subxactlist, current_xid);
1396 34 : MemoryContextSwitchTo(oldctx);
1397 : }
1398 137846 : }
1399 :
1400 : /* Reset the list that maintains subtransactions. */
1401 : void
1402 48 : pa_reset_subtrans(void)
1403 : {
1404 : /*
1405 : * We don't need to free this explicitly as the allocated memory will be
1406 : * freed at the transaction end.
1407 : */
1408 48 : subxactlist = NIL;
1409 48 : }
1410 :
1411 : /*
1412 : * Handle STREAM ABORT message when the transaction was applied in a parallel
1413 : * apply worker.
1414 : */
1415 : void
1416 24 : pa_stream_abort(LogicalRepStreamAbortData *abort_data)
1417 : {
1418 24 : TransactionId xid = abort_data->xid;
1419 24 : TransactionId subxid = abort_data->subxid;
1420 :
1421 : /*
1422 : * Update origin state so we can restart streaming from correct position
1423 : * in case of crash.
1424 : */
1425 24 : replorigin_session_origin_lsn = abort_data->abort_lsn;
1426 24 : replorigin_session_origin_timestamp = abort_data->abort_time;
1427 :
1428 : /*
1429 : * If the two XIDs are the same, it's in fact abort of toplevel xact, so
1430 : * just free the subxactlist.
1431 : */
1432 24 : if (subxid == xid)
1433 : {
1434 4 : pa_set_xact_state(MyParallelShared, PARALLEL_TRANS_FINISHED);
1435 :
1436 : /*
1437 : * Release the lock as we might be processing an empty streaming
1438 : * transaction in which case the lock won't be released during
1439 : * transaction rollback.
1440 : *
1441 : * Note that it's ok to release the transaction lock before aborting
1442 : * the transaction because even if the parallel apply worker dies due
1443 : * to crash or some other reason, such a transaction would still be
1444 : * considered aborted.
1445 : */
1446 4 : pa_unlock_transaction(xid, AccessExclusiveLock);
1447 :
1448 4 : AbortCurrentTransaction();
1449 :
1450 4 : if (IsTransactionBlock())
1451 : {
1452 2 : EndTransactionBlock(false);
1453 2 : CommitTransactionCommand();
1454 : }
1455 :
1456 4 : pa_reset_subtrans();
1457 :
1458 4 : pgstat_report_activity(STATE_IDLE, NULL);
1459 : }
1460 : else
1461 : {
1462 : /* OK, so it's a subxact. Rollback to the savepoint. */
1463 : int i;
1464 : char spname[NAMEDATALEN];
1465 :
1466 20 : pa_savepoint_name(MySubscription->oid, subxid, spname, sizeof(spname));
1467 :
1468 20 : elog(DEBUG1, "rolling back to savepoint %s in logical replication parallel apply worker", spname);
1469 :
1470 : /*
1471 : * Search the subxactlist, determine the offset tracked for the
1472 : * subxact, and truncate the list.
1473 : *
1474 : * Note that for an empty sub-transaction we won't find the subxid
1475 : * here.
1476 : */
1477 24 : for (i = list_length(subxactlist) - 1; i >= 0; i--)
1478 : {
1479 22 : TransactionId xid_tmp = lfirst_xid(list_nth_cell(subxactlist, i));
1480 :
1481 22 : if (xid_tmp == subxid)
1482 : {
1483 18 : RollbackToSavepoint(spname);
1484 18 : CommitTransactionCommand();
1485 18 : subxactlist = list_truncate(subxactlist, i);
1486 18 : break;
1487 : }
1488 : }
1489 : }
1490 24 : }
1491 :
1492 : /*
1493 : * Set the fileset state for a particular parallel apply worker. The fileset
1494 : * will be set once the leader worker serialized all changes to the file
1495 : * so that it can be used by parallel apply worker.
1496 : */
1497 : void
1498 32 : pa_set_fileset_state(ParallelApplyWorkerShared *wshared,
1499 : PartialFileSetState fileset_state)
1500 : {
1501 32 : SpinLockAcquire(&wshared->mutex);
1502 32 : wshared->fileset_state = fileset_state;
1503 :
1504 32 : if (fileset_state == FS_SERIALIZE_DONE)
1505 : {
1506 : Assert(am_leader_apply_worker());
1507 : Assert(MyLogicalRepWorker->stream_fileset);
1508 8 : wshared->fileset = *MyLogicalRepWorker->stream_fileset;
1509 : }
1510 :
1511 32 : SpinLockRelease(&wshared->mutex);
1512 32 : }
1513 :
1514 : /*
1515 : * Get the fileset state for the current parallel apply worker.
1516 : */
1517 : static PartialFileSetState
1518 146 : pa_get_fileset_state(void)
1519 : {
1520 : PartialFileSetState fileset_state;
1521 :
1522 : Assert(am_parallel_apply_worker());
1523 :
1524 146 : SpinLockAcquire(&MyParallelShared->mutex);
1525 146 : fileset_state = MyParallelShared->fileset_state;
1526 146 : SpinLockRelease(&MyParallelShared->mutex);
1527 :
1528 146 : return fileset_state;
1529 : }
1530 :
1531 : /*
1532 : * Helper functions to acquire and release a lock for each stream block.
1533 : *
1534 : * Set locktag_field4 to PARALLEL_APPLY_LOCK_STREAM to indicate that it's a
1535 : * stream lock.
1536 : *
1537 : * Refer to the comments atop this file to see how the stream lock is used.
1538 : */
1539 : void
1540 548 : pa_lock_stream(TransactionId xid, LOCKMODE lockmode)
1541 : {
1542 548 : LockApplyTransactionForSession(MyLogicalRepWorker->subid, xid,
1543 : PARALLEL_APPLY_LOCK_STREAM, lockmode);
1544 544 : }
1545 :
1546 : void
1547 540 : pa_unlock_stream(TransactionId xid, LOCKMODE lockmode)
1548 : {
1549 540 : UnlockApplyTransactionForSession(MyLogicalRepWorker->subid, xid,
1550 : PARALLEL_APPLY_LOCK_STREAM, lockmode);
1551 540 : }
1552 :
1553 : /*
1554 : * Helper functions to acquire and release a lock for each local transaction
1555 : * apply.
1556 : *
1557 : * Set locktag_field4 to PARALLEL_APPLY_LOCK_XACT to indicate that it's a
1558 : * transaction lock.
1559 : *
1560 : * Note that all the callers must pass a remote transaction ID instead of a
1561 : * local transaction ID as xid. This is because the local transaction ID will
1562 : * only be assigned while applying the first change in the parallel apply but
1563 : * it's possible that the first change in the parallel apply worker is blocked
1564 : * by a concurrently executing transaction in another parallel apply worker. We
1565 : * can only communicate the local transaction id to the leader after applying
1566 : * the first change so it won't be able to wait after sending the xact finish
1567 : * command using this lock.
1568 : *
1569 : * Refer to the comments atop this file to see how the transaction lock is
1570 : * used.
1571 : */
1572 : void
1573 104 : pa_lock_transaction(TransactionId xid, LOCKMODE lockmode)
1574 : {
1575 104 : LockApplyTransactionForSession(MyLogicalRepWorker->subid, xid,
1576 : PARALLEL_APPLY_LOCK_XACT, lockmode);
1577 102 : }
1578 :
1579 : void
1580 96 : pa_unlock_transaction(TransactionId xid, LOCKMODE lockmode)
1581 : {
1582 96 : UnlockApplyTransactionForSession(MyLogicalRepWorker->subid, xid,
1583 : PARALLEL_APPLY_LOCK_XACT, lockmode);
1584 96 : }
1585 :
1586 : /*
1587 : * Decrement the number of pending streaming blocks and wait on the stream lock
1588 : * if there is no pending block available.
1589 : */
1590 : void
1591 504 : pa_decr_and_wait_stream_block(void)
1592 : {
1593 : Assert(am_parallel_apply_worker());
1594 :
1595 : /*
1596 : * It is only possible to not have any pending stream chunks when we are
1597 : * applying spooled messages.
1598 : */
1599 504 : if (pg_atomic_read_u32(&MyParallelShared->pending_stream_count) == 0)
1600 : {
1601 32 : if (pa_has_spooled_message_pending())
1602 32 : return;
1603 :
1604 0 : elog(ERROR, "invalid pending streaming chunk 0");
1605 : }
1606 :
1607 472 : if (pg_atomic_sub_fetch_u32(&MyParallelShared->pending_stream_count, 1) == 0)
1608 : {
1609 48 : pa_lock_stream(MyParallelShared->xid, AccessShareLock);
1610 44 : pa_unlock_stream(MyParallelShared->xid, AccessShareLock);
1611 : }
1612 : }
1613 :
1614 : /*
1615 : * Finish processing the streaming transaction in the leader apply worker.
1616 : */
1617 : void
1618 50 : pa_xact_finish(ParallelApplyWorkerInfo *winfo, XLogRecPtr remote_lsn)
1619 : {
1620 : Assert(am_leader_apply_worker());
1621 :
1622 : /*
1623 : * Unlock the shared object lock so that parallel apply worker can
1624 : * continue to receive and apply changes.
1625 : */
1626 50 : pa_unlock_stream(winfo->shared->xid, AccessExclusiveLock);
1627 :
1628 : /*
1629 : * Wait for that worker to finish. This is necessary to maintain commit
1630 : * order which avoids failures due to transaction dependencies and
1631 : * deadlocks.
1632 : */
1633 50 : pa_wait_for_xact_finish(winfo);
1634 :
1635 48 : if (!XLogRecPtrIsInvalid(remote_lsn))
1636 44 : store_flush_position(remote_lsn, winfo->shared->last_commit_end);
1637 :
1638 48 : pa_free_worker(winfo);
1639 48 : }
|
