Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * walreceiverfuncs.c
4 : *
5 : * This file contains functions used by the startup process to communicate
6 : * with the walreceiver process. Functions implementing walreceiver itself
7 : * are in walreceiver.c.
8 : *
9 : * Portions Copyright (c) 2010-2026, PostgreSQL Global Development Group
10 : *
11 : *
12 : * IDENTIFICATION
13 : * src/backend/replication/walreceiverfuncs.c
14 : *
15 : *-------------------------------------------------------------------------
16 : */
17 : #include "postgres.h"
18 :
19 : #include <sys/stat.h>
20 : #include <sys/time.h>
21 : #include <time.h>
22 : #include <unistd.h>
23 : #include <signal.h>
24 :
25 : #include "access/xlog_internal.h"
26 : #include "access/xlogrecovery.h"
27 : #include "pgstat.h"
28 : #include "replication/walreceiver.h"
29 : #include "storage/pmsignal.h"
30 : #include "storage/proc.h"
31 : #include "storage/shmem.h"
32 : #include "utils/timestamp.h"
33 :
34 : WalRcvData *WalRcv = NULL;
35 :
36 : /*
37 : * How long to wait for walreceiver to start up after requesting
38 : * postmaster to launch it. In seconds.
39 : */
40 : #define WALRCV_STARTUP_TIMEOUT 10
41 :
42 : /* Report shared memory space needed by WalRcvShmemInit */
43 : Size
44 4407 : WalRcvShmemSize(void)
45 : {
46 4407 : Size size = 0;
47 :
48 4407 : size = add_size(size, sizeof(WalRcvData));
49 :
50 4407 : return size;
51 : }
52 :
53 : /* Allocate and initialize walreceiver-related shared memory */
54 : void
55 1140 : WalRcvShmemInit(void)
56 : {
57 : bool found;
58 :
59 1140 : WalRcv = (WalRcvData *)
60 1140 : ShmemInitStruct("Wal Receiver Ctl", WalRcvShmemSize(), &found);
61 :
62 1140 : if (!found)
63 : {
64 : /* First time through, so initialize */
65 1140 : MemSet(WalRcv, 0, WalRcvShmemSize());
66 1140 : WalRcv->walRcvState = WALRCV_STOPPED;
67 1140 : ConditionVariableInit(&WalRcv->walRcvStoppedCV);
68 1140 : SpinLockInit(&WalRcv->mutex);
69 1140 : pg_atomic_init_u64(&WalRcv->writtenUpto, 0);
70 1140 : WalRcv->procno = INVALID_PROC_NUMBER;
71 : }
72 1140 : }
73 :
74 : /* Is walreceiver running (or starting up)? */
75 : bool
76 1038 : WalRcvRunning(void)
77 : {
78 1038 : WalRcvData *walrcv = WalRcv;
79 : WalRcvState state;
80 : pg_time_t startTime;
81 :
82 1038 : SpinLockAcquire(&walrcv->mutex);
83 :
84 1038 : state = walrcv->walRcvState;
85 1038 : startTime = walrcv->startTime;
86 :
87 1038 : SpinLockRelease(&walrcv->mutex);
88 :
89 : /*
90 : * If it has taken too long for walreceiver to start up, give up. Setting
91 : * the state to STOPPED ensures that if walreceiver later does start up
92 : * after all, it will see that it's not supposed to be running and die
93 : * without doing anything.
94 : */
95 1038 : if (state == WALRCV_STARTING)
96 : {
97 1 : pg_time_t now = (pg_time_t) time(NULL);
98 :
99 1 : if ((now - startTime) > WALRCV_STARTUP_TIMEOUT)
100 : {
101 0 : bool stopped = false;
102 :
103 0 : SpinLockAcquire(&walrcv->mutex);
104 0 : if (walrcv->walRcvState == WALRCV_STARTING)
105 : {
106 0 : state = walrcv->walRcvState = WALRCV_STOPPED;
107 0 : stopped = true;
108 : }
109 0 : SpinLockRelease(&walrcv->mutex);
110 :
111 0 : if (stopped)
112 0 : ConditionVariableBroadcast(&walrcv->walRcvStoppedCV);
113 : }
114 : }
115 :
116 1038 : if (state != WALRCV_STOPPED)
117 44 : return true;
118 : else
119 994 : return false;
120 : }
121 :
122 : /* Return the state of the walreceiver. */
123 : WalRcvState
124 0 : WalRcvGetState(void)
125 : {
126 0 : WalRcvData *walrcv = WalRcv;
127 : WalRcvState state;
128 :
129 0 : SpinLockAcquire(&walrcv->mutex);
130 0 : state = walrcv->walRcvState;
131 0 : SpinLockRelease(&walrcv->mutex);
132 :
133 0 : return state;
134 : }
135 :
136 : /*
137 : * Is walreceiver running and streaming (or at least attempting to connect,
138 : * or starting up)?
139 : */
140 : bool
141 17767 : WalRcvStreaming(void)
142 : {
143 17767 : WalRcvData *walrcv = WalRcv;
144 : WalRcvState state;
145 : pg_time_t startTime;
146 :
147 17767 : SpinLockAcquire(&walrcv->mutex);
148 :
149 17767 : state = walrcv->walRcvState;
150 17767 : startTime = walrcv->startTime;
151 :
152 17767 : SpinLockRelease(&walrcv->mutex);
153 :
154 : /*
155 : * If it has taken too long for walreceiver to start up, give up. Setting
156 : * the state to STOPPED ensures that if walreceiver later does start up
157 : * after all, it will see that it's not supposed to be running and die
158 : * without doing anything.
159 : */
160 17767 : if (state == WALRCV_STARTING)
161 : {
162 297 : pg_time_t now = (pg_time_t) time(NULL);
163 :
164 297 : if ((now - startTime) > WALRCV_STARTUP_TIMEOUT)
165 : {
166 0 : bool stopped = false;
167 :
168 0 : SpinLockAcquire(&walrcv->mutex);
169 0 : if (walrcv->walRcvState == WALRCV_STARTING)
170 : {
171 0 : state = walrcv->walRcvState = WALRCV_STOPPED;
172 0 : stopped = true;
173 : }
174 0 : SpinLockRelease(&walrcv->mutex);
175 :
176 0 : if (stopped)
177 0 : ConditionVariableBroadcast(&walrcv->walRcvStoppedCV);
178 : }
179 : }
180 :
181 17767 : if (state == WALRCV_STREAMING || state == WALRCV_STARTING ||
182 305 : state == WALRCV_CONNECTING || state == WALRCV_RESTARTING)
183 17467 : return true;
184 : else
185 300 : return false;
186 : }
187 :
188 : /*
189 : * Stop walreceiver (if running) and wait for it to die.
190 : * Executed by the Startup process.
191 : */
192 : void
193 990 : ShutdownWalRcv(void)
194 : {
195 990 : WalRcvData *walrcv = WalRcv;
196 990 : pid_t walrcvpid = 0;
197 990 : bool stopped = false;
198 :
199 : /*
200 : * Request walreceiver to stop. Walreceiver will switch to WALRCV_STOPPED
201 : * mode once it's finished, and will also request postmaster to not
202 : * restart itself.
203 : */
204 990 : SpinLockAcquire(&walrcv->mutex);
205 990 : switch (walrcv->walRcvState)
206 : {
207 947 : case WALRCV_STOPPED:
208 947 : break;
209 5 : case WALRCV_STARTING:
210 5 : walrcv->walRcvState = WALRCV_STOPPED;
211 5 : stopped = true;
212 5 : break;
213 :
214 38 : case WALRCV_CONNECTING:
215 : case WALRCV_STREAMING:
216 : case WALRCV_WAITING:
217 : case WALRCV_RESTARTING:
218 38 : walrcv->walRcvState = WALRCV_STOPPING;
219 : /* fall through */
220 38 : case WALRCV_STOPPING:
221 38 : walrcvpid = walrcv->pid;
222 38 : break;
223 : }
224 990 : SpinLockRelease(&walrcv->mutex);
225 :
226 : /* Unnecessary but consistent. */
227 990 : if (stopped)
228 5 : ConditionVariableBroadcast(&walrcv->walRcvStoppedCV);
229 :
230 : /*
231 : * Signal walreceiver process if it was still running.
232 : */
233 990 : if (walrcvpid != 0)
234 38 : kill(walrcvpid, SIGTERM);
235 :
236 : /*
237 : * Wait for walreceiver to acknowledge its death by setting state to
238 : * WALRCV_STOPPED.
239 : */
240 990 : ConditionVariablePrepareToSleep(&walrcv->walRcvStoppedCV);
241 1027 : while (WalRcvRunning())
242 37 : ConditionVariableSleep(&walrcv->walRcvStoppedCV,
243 : WAIT_EVENT_WAL_RECEIVER_EXIT);
244 990 : ConditionVariableCancelSleep();
245 990 : }
246 :
247 : /*
248 : * Request postmaster to start walreceiver.
249 : *
250 : * "recptr" indicates the position where streaming should begin. "conninfo"
251 : * is a libpq connection string to use. "slotname" is, optionally, the name
252 : * of a replication slot to acquire. "create_temp_slot" indicates to create
253 : * a temporary slot when no "slotname" is given.
254 : *
255 : * WAL receivers do not directly load GUC parameters used for the connection
256 : * to the primary, and rely on the values passed down by the caller of this
257 : * routine instead. Hence, the addition of any new parameters should happen
258 : * through this code path.
259 : */
260 : void
261 199 : RequestXLogStreaming(TimeLineID tli, XLogRecPtr recptr, const char *conninfo,
262 : const char *slotname, bool create_temp_slot)
263 : {
264 199 : WalRcvData *walrcv = WalRcv;
265 199 : bool launch = false;
266 199 : pg_time_t now = (pg_time_t) time(NULL);
267 : ProcNumber walrcv_proc;
268 :
269 : /*
270 : * We always start at the beginning of the segment. That prevents a broken
271 : * segment (i.e., with no records in the first half of a segment) from
272 : * being created by XLOG streaming, which might cause trouble later on if
273 : * the segment is e.g archived.
274 : */
275 199 : if (XLogSegmentOffset(recptr, wal_segment_size) != 0)
276 199 : recptr -= XLogSegmentOffset(recptr, wal_segment_size);
277 :
278 199 : SpinLockAcquire(&walrcv->mutex);
279 :
280 : /* It better be stopped if we try to restart it */
281 : Assert(walrcv->walRcvState == WALRCV_STOPPED ||
282 : walrcv->walRcvState == WALRCV_WAITING);
283 :
284 199 : if (conninfo != NULL)
285 199 : strlcpy(walrcv->conninfo, conninfo, MAXCONNINFO);
286 : else
287 0 : walrcv->conninfo[0] = '\0';
288 :
289 : /*
290 : * Use configured replication slot if present, and ignore the value of
291 : * create_temp_slot as the slot name should be persistent. Otherwise, use
292 : * create_temp_slot to determine whether this WAL receiver should create a
293 : * temporary slot by itself and use it, or not.
294 : */
295 199 : if (slotname != NULL && slotname[0] != '\0')
296 : {
297 53 : strlcpy(walrcv->slotname, slotname, NAMEDATALEN);
298 53 : walrcv->is_temp_slot = false;
299 : }
300 : else
301 : {
302 146 : walrcv->slotname[0] = '\0';
303 146 : walrcv->is_temp_slot = create_temp_slot;
304 : }
305 :
306 199 : if (walrcv->walRcvState == WALRCV_STOPPED)
307 : {
308 192 : launch = true;
309 192 : walrcv->walRcvState = WALRCV_STARTING;
310 : }
311 : else
312 7 : walrcv->walRcvState = WALRCV_RESTARTING;
313 199 : walrcv->startTime = now;
314 :
315 : /*
316 : * If this is the first startup of walreceiver (on this timeline),
317 : * initialize flushedUpto and latestChunkStart to the starting point.
318 : */
319 199 : if (!XLogRecPtrIsValid(walrcv->receiveStart) || walrcv->receivedTLI != tli)
320 : {
321 107 : walrcv->flushedUpto = recptr;
322 107 : walrcv->receivedTLI = tli;
323 107 : walrcv->latestChunkStart = recptr;
324 : }
325 199 : walrcv->receiveStart = recptr;
326 199 : walrcv->receiveStartTLI = tli;
327 :
328 199 : walrcv_proc = walrcv->procno;
329 :
330 199 : SpinLockRelease(&walrcv->mutex);
331 :
332 199 : if (launch)
333 192 : SendPostmasterSignal(PMSIGNAL_START_WALRECEIVER);
334 7 : else if (walrcv_proc != INVALID_PROC_NUMBER)
335 7 : SetLatch(&GetPGProcByNumber(walrcv_proc)->procLatch);
336 199 : }
337 :
338 : /*
339 : * Returns the last+1 byte position that walreceiver has flushed.
340 : *
341 : * Optionally, returns the previous chunk start, that is the first byte
342 : * written in the most recent walreceiver flush cycle. Callers not
343 : * interested in that value may pass NULL for latestChunkStart. Same for
344 : * receiveTLI.
345 : */
346 : XLogRecPtr
347 16907 : GetWalRcvFlushRecPtr(XLogRecPtr *latestChunkStart, TimeLineID *receiveTLI)
348 : {
349 16907 : WalRcvData *walrcv = WalRcv;
350 : XLogRecPtr recptr;
351 :
352 16907 : SpinLockAcquire(&walrcv->mutex);
353 16907 : recptr = walrcv->flushedUpto;
354 16907 : if (latestChunkStart)
355 15706 : *latestChunkStart = walrcv->latestChunkStart;
356 16907 : if (receiveTLI)
357 16676 : *receiveTLI = walrcv->receivedTLI;
358 16907 : SpinLockRelease(&walrcv->mutex);
359 :
360 16907 : return recptr;
361 : }
362 :
363 : /*
364 : * Returns the last+1 byte position that walreceiver has written.
365 : * This returns a recently written value without taking a lock.
366 : */
367 : XLogRecPtr
368 29 : GetWalRcvWriteRecPtr(void)
369 : {
370 29 : WalRcvData *walrcv = WalRcv;
371 :
372 29 : return pg_atomic_read_u64(&walrcv->writtenUpto);
373 : }
374 :
375 : /*
376 : * Returns the replication apply delay in ms or -1
377 : * if the apply delay info is not available
378 : */
379 : int
380 400 : GetReplicationApplyDelay(void)
381 : {
382 400 : WalRcvData *walrcv = WalRcv;
383 : XLogRecPtr receivePtr;
384 : XLogRecPtr replayPtr;
385 : TimestampTz chunkReplayStartTime;
386 :
387 400 : SpinLockAcquire(&walrcv->mutex);
388 400 : receivePtr = walrcv->flushedUpto;
389 400 : SpinLockRelease(&walrcv->mutex);
390 :
391 400 : replayPtr = GetXLogReplayRecPtr(NULL);
392 :
393 400 : if (receivePtr == replayPtr)
394 141 : return 0;
395 :
396 259 : chunkReplayStartTime = GetCurrentChunkReplayStartTime();
397 :
398 259 : if (chunkReplayStartTime == 0)
399 30 : return -1;
400 :
401 229 : return TimestampDifferenceMilliseconds(chunkReplayStartTime,
402 : GetCurrentTimestamp());
403 : }
404 :
405 : /*
406 : * Returns the network latency in ms, note that this includes any
407 : * difference in clock settings between the servers, as well as timezone.
408 : */
409 : int
410 400 : GetReplicationTransferLatency(void)
411 : {
412 400 : WalRcvData *walrcv = WalRcv;
413 : TimestampTz lastMsgSendTime;
414 : TimestampTz lastMsgReceiptTime;
415 :
416 400 : SpinLockAcquire(&walrcv->mutex);
417 400 : lastMsgSendTime = walrcv->lastMsgSendTime;
418 400 : lastMsgReceiptTime = walrcv->lastMsgReceiptTime;
419 400 : SpinLockRelease(&walrcv->mutex);
420 :
421 400 : return TimestampDifferenceMilliseconds(lastMsgSendTime,
422 : lastMsgReceiptTime);
423 : }
|
