Age Owner Branch data TLA Line data Source code
1 : : /*-------------------------------------------------------------------------
2 : : *
3 : : * walreceiver.c
4 : : *
5 : : * The WAL receiver process (walreceiver) is new as of Postgres 9.0. It
6 : : * is the process in the standby server that takes charge of receiving
7 : : * XLOG records from a primary server during streaming replication.
8 : : *
9 : : * When the startup process determines that it's time to start streaming,
10 : : * it instructs postmaster to start walreceiver. Walreceiver first connects
11 : : * to the primary server (it will be served by a walsender process
12 : : * in the primary server), and then keeps receiving XLOG records and
13 : : * writing them to the disk as long as the connection is alive. As XLOG
14 : : * records are received and flushed to disk, it updates the
15 : : * WalRcv->flushedUpto variable in shared memory, to inform the startup
16 : : * process of how far it can proceed with XLOG replay.
17 : : *
18 : : * A WAL receiver cannot directly load GUC parameters used when establishing
19 : : * its connection to the primary. Instead it relies on parameter values
20 : : * that are passed down by the startup process when streaming is requested.
21 : : * This applies, for example, to the replication slot and the connection
22 : : * string to be used for the connection with the primary.
23 : : *
24 : : * If the primary server ends streaming, but doesn't disconnect, walreceiver
25 : : * goes into "waiting" mode, and waits for the startup process to give new
26 : : * instructions. The startup process will treat that the same as
27 : : * disconnection, and will rescan the archive/pg_wal directory. But when the
28 : : * startup process wants to try streaming replication again, it will just
29 : : * nudge the existing walreceiver process that's waiting, instead of launching
30 : : * a new one.
31 : : *
32 : : * Normal termination is by SIGTERM, which instructs the walreceiver to
33 : : * ereport(FATAL). Emergency termination is by SIGQUIT; like any postmaster
34 : : * child process, the walreceiver will simply abort and exit on SIGQUIT. A
35 : : * close of the connection and a FATAL error are treated not as a crash but as
36 : : * normal operation.
37 : : *
38 : : * This file contains the server-facing parts of walreceiver. The libpq-
39 : : * specific parts are in the libpqwalreceiver module. It's loaded
40 : : * dynamically to avoid linking the server with libpq.
41 : : *
42 : : * Portions Copyright (c) 2010-2026, PostgreSQL Global Development Group
43 : : *
44 : : *
45 : : * IDENTIFICATION
46 : : * src/backend/replication/walreceiver.c
47 : : *
48 : : *-------------------------------------------------------------------------
49 : : */
50 : : #include "postgres.h"
51 : :
52 : : #include <unistd.h>
53 : :
54 : : #include "access/htup_details.h"
55 : : #include "access/timeline.h"
56 : : #include "access/transam.h"
57 : : #include "access/xlog_internal.h"
58 : : #include "access/xlogarchive.h"
59 : : #include "access/xlogrecovery.h"
60 : : #include "access/xlogwait.h"
61 : : #include "catalog/pg_authid.h"
62 : : #include "funcapi.h"
63 : : #include "libpq/pqformat.h"
64 : : #include "libpq/pqsignal.h"
65 : : #include "miscadmin.h"
66 : : #include "pgstat.h"
67 : : #include "postmaster/auxprocess.h"
68 : : #include "postmaster/interrupt.h"
69 : : #include "replication/walreceiver.h"
70 : : #include "replication/walsender.h"
71 : : #include "storage/ipc.h"
72 : : #include "storage/proc.h"
73 : : #include "storage/procarray.h"
74 : : #include "storage/procsignal.h"
75 : : #include "tcop/tcopprot.h"
76 : : #include "utils/acl.h"
77 : : #include "utils/builtins.h"
78 : : #include "utils/guc.h"
79 : : #include "utils/pg_lsn.h"
80 : : #include "utils/ps_status.h"
81 : : #include "utils/timestamp.h"
82 : : #include "utils/wait_event.h"
83 : :
84 : :
85 : : /*
86 : : * GUC variables. (Other variables that affect walreceiver are in xlog.c
87 : : * because they're passed down from the startup process, for better
88 : : * synchronization.)
89 : : */
90 : : int wal_receiver_status_interval;
91 : : int wal_receiver_timeout;
92 : : bool hot_standby_feedback;
93 : :
94 : : /* libpqwalreceiver connection */
95 : : static WalReceiverConn *wrconn = NULL;
96 : : WalReceiverFunctionsType *WalReceiverFunctions = NULL;
97 : :
98 : : /*
99 : : * These variables are used similarly to openLogFile/SegNo,
100 : : * but for walreceiver to write the XLOG. recvFileTLI is the TimeLineID
101 : : * corresponding the filename of recvFile.
102 : : */
103 : : static int recvFile = -1;
104 : : static TimeLineID recvFileTLI = 0;
105 : : static XLogSegNo recvSegNo = 0;
106 : :
107 : : /*
108 : : * LogstreamResult indicates the byte positions that we have already
109 : : * written/fsynced.
110 : : */
111 : : static struct
112 : : {
113 : : XLogRecPtr Write; /* last byte + 1 written out in the standby */
114 : : XLogRecPtr Flush; /* last byte + 1 flushed in the standby */
115 : : } LogstreamResult;
116 : :
117 : : /*
118 : : * Reasons to wake up and perform periodic tasks.
119 : : */
120 : : typedef enum WalRcvWakeupReason
121 : : {
122 : : WALRCV_WAKEUP_TERMINATE,
123 : : WALRCV_WAKEUP_PING,
124 : : WALRCV_WAKEUP_REPLY,
125 : : WALRCV_WAKEUP_HSFEEDBACK,
126 : : #define NUM_WALRCV_WAKEUPS (WALRCV_WAKEUP_HSFEEDBACK + 1)
127 : : } WalRcvWakeupReason;
128 : :
129 : : /*
130 : : * Wake up times for periodic tasks.
131 : : */
132 : : static TimestampTz wakeup[NUM_WALRCV_WAKEUPS];
133 : :
134 : : static StringInfoData reply_message;
135 : :
136 : : /* Prototypes for private functions */
137 : : static void WalRcvFetchTimeLineHistoryFiles(TimeLineID first, TimeLineID last);
138 : : static void WalRcvWaitForStartPosition(XLogRecPtr *startpoint, TimeLineID *startpointTLI);
139 : : static void WalRcvDie(int code, Datum arg);
140 : : static void XLogWalRcvProcessMsg(unsigned char type, char *buf, Size len,
141 : : TimeLineID tli);
142 : : static void XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr,
143 : : TimeLineID tli);
144 : : static void XLogWalRcvFlush(bool dying, TimeLineID tli);
145 : : static void XLogWalRcvClose(XLogRecPtr recptr, TimeLineID tli);
146 : : static void XLogWalRcvSendReply(bool force, bool requestReply, bool checkApply);
147 : : static void XLogWalRcvSendHSFeedback(bool immed);
148 : : static void ProcessWalSndrMessage(XLogRecPtr walEnd, TimestampTz sendTime);
149 : : static void WalRcvComputeNextWakeup(WalRcvWakeupReason reason, TimestampTz now);
150 : :
151 : :
152 : : /* Main entry point for walreceiver process */
153 : : void
438 peter@eisentraut.org 154 :CBC 241 : WalReceiverMain(const void *startup_data, size_t startup_data_len)
155 : : {
156 : : char conninfo[MAXCONNINFO];
157 : : char *tmp_conninfo;
158 : : char slotname[NAMEDATALEN];
159 : : bool is_temp_slot;
160 : : XLogRecPtr startpoint;
161 : : TimeLineID startpointTLI;
162 : : TimeLineID primaryTLI;
163 : : bool first_stream;
164 : : WalRcvData *walrcv;
165 : : TimestampTz now;
166 : : char *err;
2957 fujii@postgresql.org 167 : 241 : char *sender_host = NULL;
168 : 241 : int sender_port = 0;
169 : : char *appname;
170 : :
778 heikki.linnakangas@i 171 [ - + ]: 241 : Assert(startup_data_len == 0);
172 : :
173 : 241 : AuxiliaryProcessMainCommon();
174 : :
175 : : /*
176 : : * WalRcv should be set up already (if we are a backend, we inherit this
177 : : * by fork() or EXEC_BACKEND mechanism from the postmaster).
178 : : */
884 179 : 241 : walrcv = WalRcv;
5942 180 [ - + ]: 241 : Assert(walrcv != NULL);
181 : :
182 : : /*
183 : : * Mark walreceiver as running in shared memory.
184 : : *
185 : : * Do this as early as possible, so that if we fail later on, we'll set
186 : : * state to STOPPED. If we die before this, the startup process will keep
187 : : * waiting for us to start up, until it times out.
188 : : */
189 [ - + ]: 241 : SpinLockAcquire(&walrcv->mutex);
190 [ - + ]: 241 : Assert(walrcv->pid == 0);
5912 bruce@momjian.us 191 [ - + + - ]: 241 : switch (walrcv->walRcvState)
192 : : {
5942 heikki.linnakangas@i 193 :UBC 0 : case WALRCV_STOPPING:
194 : : /* If we've already been requested to stop, don't start up. */
195 : 0 : walrcv->walRcvState = WALRCV_STOPPED;
196 : : pg_fallthrough;
197 : :
5942 heikki.linnakangas@i 198 :CBC 1 : case WALRCV_STOPPED:
199 : 1 : SpinLockRelease(&walrcv->mutex);
1880 tmunro@postgresql.or 200 : 1 : ConditionVariableBroadcast(&walrcv->walRcvStoppedCV);
5942 heikki.linnakangas@i 201 : 1 : proc_exit(1);
202 : : break;
203 : :
204 : 240 : case WALRCV_STARTING:
205 : : /* The usual case */
206 : 240 : break;
207 : :
102 michael@paquier.xyz 208 :UNC 0 : case WALRCV_CONNECTING:
4891 heikki.linnakangas@i 209 :EUB : case WALRCV_WAITING:
210 : : case WALRCV_STREAMING:
211 : : case WALRCV_RESTARTING:
212 : : default:
213 : : /* Shouldn't happen */
3136 alvherre@alvh.no-ip. 214 :UBC 0 : SpinLockRelease(&walrcv->mutex);
5942 heikki.linnakangas@i 215 [ # # ]: 0 : elog(PANIC, "walreceiver still running according to shared memory state");
216 : : }
217 : : /* Advertise our PID so that the startup process can kill us */
5942 heikki.linnakangas@i 218 :CBC 240 : walrcv->pid = MyProcPid;
102 michael@paquier.xyz 219 :GNC 240 : walrcv->walRcvState = WALRCV_CONNECTING;
220 : :
221 : : /* Fetch information required to start streaming */
3595 alvherre@alvh.no-ip. 222 :CBC 240 : walrcv->ready_to_display = false;
447 peter@eisentraut.org 223 : 240 : strlcpy(conninfo, walrcv->conninfo, MAXCONNINFO);
224 : 240 : strlcpy(slotname, walrcv->slotname, NAMEDATALEN);
2303 225 : 240 : is_temp_slot = walrcv->is_temp_slot;
5544 heikki.linnakangas@i 226 : 240 : startpoint = walrcv->receiveStart;
4891 227 : 240 : startpointTLI = walrcv->receiveStartTLI;
228 : :
229 : : /*
230 : : * At most one of is_temp_slot and slotname can be set; otherwise,
231 : : * RequestXLogStreaming messed up.
232 : : */
2230 alvherre@alvh.no-ip. 233 [ - + - - ]: 240 : Assert(!is_temp_slot || (slotname[0] == '\0'));
234 : :
235 : : /* Initialise to a sanish value */
1195 tgl@sss.pgh.pa.us 236 : 240 : now = GetCurrentTimestamp();
3136 alvherre@alvh.no-ip. 237 : 240 : walrcv->lastMsgSendTime =
1274 tmunro@postgresql.or 238 : 240 : walrcv->lastMsgReceiptTime = walrcv->latestWalEndTime = now;
239 : :
240 : : /* Report our proc number so that others can wake us up */
550 heikki.linnakangas@i 241 : 240 : walrcv->procno = MyProcNumber;
242 : :
5942 243 : 240 : SpinLockRelease(&walrcv->mutex);
244 : :
245 : : /* Arrange to clean up at walreceiver exit */
1642 rhaas@postgresql.org 246 : 240 : on_shmem_exit(WalRcvDie, PointerGetDatum(&startpointTLI));
247 : :
248 : : /* Properly accept or ignore signals the postmaster might send us */
2000 fujii@postgresql.org 249 : 240 : pqsignal(SIGHUP, SignalHandlerForConfigReload); /* set flag to read config
250 : : * file */
21 andrew@dunslane.net 251 :GNC 240 : pqsignal(SIGINT, PG_SIG_IGN);
396 heikki.linnakangas@i 252 :CBC 240 : pqsignal(SIGTERM, die); /* request shutdown */
253 : : /* SIGQUIT handler was already set up by InitPostmasterChild */
21 andrew@dunslane.net 254 :GNC 240 : pqsignal(SIGALRM, PG_SIG_IGN);
255 : 240 : pqsignal(SIGPIPE, PG_SIG_IGN);
2353 rhaas@postgresql.org 256 :CBC 240 : pqsignal(SIGUSR1, procsignal_sigusr1_handler);
21 andrew@dunslane.net 257 :GNC 240 : pqsignal(SIGUSR2, PG_SIG_IGN);
258 : :
259 : : /* Reset some signals that are accepted by postmaster but not here */
260 : 240 : pqsignal(SIGCHLD, PG_SIG_DFL);
261 : :
262 : : /* Load the libpq-specific functions */
5942 heikki.linnakangas@i 263 :CBC 240 : load_file("libpqwalreceiver", false);
3443 peter_e@gmx.net 264 [ - + ]: 240 : if (WalReceiverFunctions == NULL)
5942 heikki.linnakangas@i 265 [ # # ]:UBC 0 : elog(ERROR, "libpqwalreceiver didn't initialize correctly");
266 : :
267 : : /* Unblock signals (they were blocked when the postmaster forked us) */
1187 tmunro@postgresql.or 268 :CBC 240 : sigprocmask(SIG_SETMASK, &UnBlockSig, NULL);
269 : :
270 : : /* Establish the connection to the primary for XLOG streaming */
663 tgl@sss.pgh.pa.us 271 [ + + ]: 240 : appname = cluster_name[0] ? cluster_name : "walreceiver";
272 : 240 : wrconn = walrcv_connect(conninfo, true, false, false, appname, &err);
3393 peter_e@gmx.net 273 [ + + ]: 240 : if (!wrconn)
274 [ + - ]: 88 : ereport(ERROR,
275 : : (errcode(ERRCODE_CONNECTION_FAILURE),
276 : : errmsg("streaming replication receiver \"%s\" could not connect to the primary server: %s",
277 : : appname, err)));
278 : :
279 : : /*
280 : : * Save user-visible connection string. This clobbers the original
281 : : * conninfo, for security. Also save host and port of the sender server
282 : : * this walreceiver is connected to.
283 : : */
3443 284 : 152 : tmp_conninfo = walrcv_get_conninfo(wrconn);
2957 fujii@postgresql.org 285 : 152 : walrcv_get_senderinfo(wrconn, &sender_host, &sender_port);
3597 alvherre@alvh.no-ip. 286 [ - + ]: 152 : SpinLockAcquire(&walrcv->mutex);
287 : 152 : memset(walrcv->conninfo, 0, MAXCONNINFO);
288 [ + - ]: 152 : if (tmp_conninfo)
447 peter@eisentraut.org 289 : 152 : strlcpy(walrcv->conninfo, tmp_conninfo, MAXCONNINFO);
290 : :
2957 fujii@postgresql.org 291 : 152 : memset(walrcv->sender_host, 0, NI_MAXHOST);
292 [ + - ]: 152 : if (sender_host)
447 peter@eisentraut.org 293 : 152 : strlcpy(walrcv->sender_host, sender_host, NI_MAXHOST);
294 : :
2957 fujii@postgresql.org 295 : 152 : walrcv->sender_port = sender_port;
3597 alvherre@alvh.no-ip. 296 : 152 : walrcv->ready_to_display = true;
297 : 152 : SpinLockRelease(&walrcv->mutex);
298 : :
3136 299 [ + - ]: 152 : if (tmp_conninfo)
300 : 152 : pfree(tmp_conninfo);
301 : :
2957 fujii@postgresql.org 302 [ + - ]: 152 : if (sender_host)
303 : 152 : pfree(sender_host);
304 : :
305 : : /* Initialize buffers for processing messages */
8 michael@paquier.xyz 306 :GNC 152 : initStringInfo(&reply_message);
307 : :
4891 heikki.linnakangas@i 308 :CBC 152 : first_stream = true;
309 : : for (;;)
5954 310 : 14 : {
311 : : char *primary_sysid;
312 : : char standby_sysid[32];
313 : : WalRcvStreamOptions options;
314 : :
315 : : /*
316 : : * Check that we're connected to a valid server using the
317 : : * IDENTIFY_SYSTEM replication command.
318 : : */
2608 peter@eisentraut.org 319 : 166 : primary_sysid = walrcv_identify_system(wrconn, &primaryTLI);
320 : :
3443 peter_e@gmx.net 321 : 166 : snprintf(standby_sysid, sizeof(standby_sysid), UINT64_FORMAT,
322 : : GetSystemIdentifier());
323 [ - + ]: 166 : if (strcmp(primary_sysid, standby_sysid) != 0)
324 : : {
3443 peter_e@gmx.net 325 [ # # ]:UBC 0 : ereport(ERROR,
326 : : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
327 : : errmsg("database system identifier differs between the primary and standby"),
328 : : errdetail("The primary's identifier is %s, the standby's identifier is %s.",
329 : : primary_sysid, standby_sysid)));
330 : : }
8 michael@paquier.xyz 331 :GNC 166 : pfree(primary_sysid);
332 : :
333 : : /*
334 : : * Confirm that the current timeline of the primary is the same or
335 : : * ahead of ours.
336 : : */
4891 heikki.linnakangas@i 337 [ - + ]:CBC 166 : if (primaryTLI < startpointTLI)
4891 heikki.linnakangas@i 338 [ # # ]:UBC 0 : ereport(ERROR,
339 : : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
340 : : errmsg("highest timeline %u of the primary is behind recovery timeline %u",
341 : : primaryTLI, startpointTLI)));
342 : :
343 : : /*
344 : : * Get any missing history files. We do this always, even when we're
345 : : * not interested in that timeline, so that if we're promoted to
346 : : * become the primary later on, we don't select the same timeline that
347 : : * was already used in the current primary. This isn't bullet-proof -
348 : : * you'll need some external software to manage your cluster if you
349 : : * need to ensure that a unique timeline id is chosen in every case,
350 : : * but let's avoid the confusion of timeline id collisions where we
351 : : * can.
352 : : */
4870 heikki.linnakangas@i 353 :CBC 166 : WalRcvFetchTimeLineHistoryFiles(startpointTLI, primaryTLI);
354 : :
355 : : /*
356 : : * Create temporary replication slot if requested, and update slot
357 : : * name in shared memory. (Note the slot name cannot already be set
358 : : * in this case.)
359 : : */
2230 alvherre@alvh.no-ip. 360 [ - + ]: 166 : if (is_temp_slot)
361 : : {
2230 alvherre@alvh.no-ip. 362 :UBC 0 : snprintf(slotname, sizeof(slotname),
363 : : "pg_walreceiver_%lld",
364 : 0 : (long long int) walrcv_get_backend_pid(wrconn));
365 : :
827 akapila@postgresql.o 366 : 0 : walrcv_create_slot(wrconn, slotname, true, false, false, 0, NULL);
367 : :
2230 alvherre@alvh.no-ip. 368 [ # # ]: 0 : SpinLockAcquire(&walrcv->mutex);
369 : 0 : strlcpy(walrcv->slotname, slotname, NAMEDATALEN);
370 : 0 : SpinLockRelease(&walrcv->mutex);
371 : : }
372 : :
373 : : /*
374 : : * Start streaming.
375 : : *
376 : : * We'll try to start at the requested starting point and timeline,
377 : : * even if it's different from the server's latest timeline. In case
378 : : * we've already reached the end of the old timeline, the server will
379 : : * finish the streaming immediately, and we will go back to await
380 : : * orders from the startup process. If recovery_target_timeline is
381 : : * 'latest', the startup process will scan pg_wal and find the new
382 : : * history file, bump recovery target timeline, and ask us to restart
383 : : * on the new timeline.
384 : : */
3393 peter_e@gmx.net 385 :CBC 166 : options.logical = false;
386 : 166 : options.startpoint = startpoint;
387 [ + + ]: 166 : options.slotname = slotname[0] != '\0' ? slotname : NULL;
388 : 166 : options.proto.physical.startpointTLI = startpointTLI;
389 [ + - ]: 166 : if (walrcv_startstreaming(wrconn, &options))
390 : : {
4891 heikki.linnakangas@i 391 [ + + ]: 165 : if (first_stream)
392 [ + - ]: 151 : ereport(LOG,
393 : : errmsg("started streaming WAL from primary at %X/%08X on timeline %u",
394 : : LSN_FORMAT_ARGS(startpoint), startpointTLI));
395 : : else
396 [ + - ]: 14 : ereport(LOG,
397 : : errmsg("restarted WAL streaming at %X/%08X on timeline %u",
398 : : LSN_FORMAT_ARGS(startpoint), startpointTLI));
399 : 165 : first_stream = false;
400 : :
401 : : /*
402 : : * Switch to STREAMING after a successful connection if current
403 : : * state is CONNECTING. This switch happens after an initial
404 : : * startup, or after a restart as determined by
405 : : * WalRcvWaitForStartPosition().
406 : : */
102 michael@paquier.xyz 407 :GNC 165 : SpinLockAcquire(&walrcv->mutex);
408 [ + - ]: 165 : if (walrcv->walRcvState == WALRCV_CONNECTING)
409 : 165 : walrcv->walRcvState = WALRCV_STREAMING;
410 : 165 : SpinLockRelease(&walrcv->mutex);
411 : :
412 : : /* Initialize LogstreamResult for processing messages */
4884 heikki.linnakangas@i 413 :CBC 165 : LogstreamResult.Write = LogstreamResult.Flush = GetXLogReplayRecPtr(NULL);
414 : :
415 : : /* Initialize nap wakeup times. */
1274 tmunro@postgresql.or 416 : 165 : now = GetCurrentTimestamp();
417 [ + + ]: 825 : for (int i = 0; i < NUM_WALRCV_WAKEUPS; ++i)
418 : 660 : WalRcvComputeNextWakeup(i, now);
419 : :
420 : : /* Send initial reply/feedback messages. */
40 fujii@postgresql.org 421 :GNC 165 : XLogWalRcvSendReply(true, false, false);
1265 tmunro@postgresql.or 422 :CBC 165 : XLogWalRcvSendHSFeedback(true);
423 : :
424 : : /* Loop until end-of-streaming or error */
425 : : for (;;)
4954 heikki.linnakangas@i 426 : 57147 : {
427 : : char *buf;
428 : : int len;
3689 rhaas@postgresql.org 429 : 57312 : bool endofwal = false;
3673 tgl@sss.pgh.pa.us 430 : 57312 : pgsocket wait_fd = PGINVALID_SOCKET;
431 : : int rc;
432 : : TimestampTz nextWakeup;
433 : : long nap;
434 : :
435 : : /*
436 : : * Exit walreceiver if we're not in recovery. This should not
437 : : * happen, but cross-check the status here.
438 : : */
4891 heikki.linnakangas@i 439 [ - + ]: 57312 : if (!RecoveryInProgress())
4891 heikki.linnakangas@i 440 [ # # ]:UBC 0 : ereport(FATAL,
441 : : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
442 : : errmsg("cannot continue WAL streaming, recovery has already ended")));
443 : :
444 : : /* Process any requests or signals received recently */
396 heikki.linnakangas@i 445 [ + + ]:CBC 57312 : CHECK_FOR_INTERRUPTS();
446 : :
2000 fujii@postgresql.org 447 [ + + ]: 57312 : if (ConfigReloadPending)
448 : : {
449 : 27 : ConfigReloadPending = false;
4891 heikki.linnakangas@i 450 : 27 : ProcessConfigFile(PGC_SIGHUP);
451 : : /* recompute wakeup times */
1274 tmunro@postgresql.or 452 : 27 : now = GetCurrentTimestamp();
453 [ + + ]: 135 : for (int i = 0; i < NUM_WALRCV_WAKEUPS; ++i)
454 : 108 : WalRcvComputeNextWakeup(i, now);
4838 simon@2ndQuadrant.co 455 : 27 : XLogWalRcvSendHSFeedback(true);
456 : : }
457 : :
458 : : /* See if we can read data immediately */
3443 peter_e@gmx.net 459 : 57312 : len = walrcv_receive(wrconn, &buf, &wait_fd);
4891 heikki.linnakangas@i 460 [ + + ]: 57283 : if (len != 0)
461 : : {
462 : : /*
463 : : * Process the received data, and any subsequent data we
464 : : * can read without blocking.
465 : : */
466 : : for (;;)
467 : : {
468 [ + + ]: 152013 : if (len > 0)
469 : : {
470 : : /*
471 : : * Something was received from primary, so adjust
472 : : * the ping and terminate wakeup times.
473 : : */
1195 tgl@sss.pgh.pa.us 474 : 108876 : now = GetCurrentTimestamp();
1274 tmunro@postgresql.or 475 : 108876 : WalRcvComputeNextWakeup(WALRCV_WAKEUP_TERMINATE,
476 : : now);
477 : 108876 : WalRcvComputeNextWakeup(WALRCV_WAKEUP_PING, now);
1642 rhaas@postgresql.org 478 : 108876 : XLogWalRcvProcessMsg(buf[0], &buf[1], len - 1,
479 : : startpointTLI);
480 : : }
4891 heikki.linnakangas@i 481 [ + + ]: 43137 : else if (len == 0)
482 : 43090 : break;
483 [ + - ]: 47 : else if (len < 0)
484 : : {
485 [ + - ]: 47 : ereport(LOG,
486 : : (errmsg("replication terminated by primary server"),
487 : : errdetail("End of WAL reached on timeline %u at %X/%08X.",
488 : : startpointTLI,
489 : : LSN_FORMAT_ARGS(LogstreamResult.Write))));
490 : 47 : endofwal = true;
491 : 47 : break;
492 : : }
3443 peter_e@gmx.net 493 : 108876 : len = walrcv_receive(wrconn, &buf, &wait_fd);
494 : : }
495 : :
496 : : /* Let the primary know that we received some data. */
40 fujii@postgresql.org 497 :GNC 43137 : XLogWalRcvSendReply(false, false, false);
498 : :
499 : : /*
500 : : * If we've written some records, flush them to disk and
501 : : * let the startup process and primary server know about
502 : : * them.
503 : : */
1642 rhaas@postgresql.org 504 :CBC 43136 : XLogWalRcvFlush(false, startpointTLI);
505 : : }
506 : :
507 : : /* Check if we need to exit the streaming loop. */
3689 508 [ + + ]: 57281 : if (endofwal)
509 : 46 : break;
510 : :
511 : : /* Find the soonest wakeup time, to limit our nap. */
1195 tgl@sss.pgh.pa.us 512 : 57235 : nextWakeup = TIMESTAMP_INFINITY;
1274 tmunro@postgresql.or 513 [ + + ]: 286175 : for (int i = 0; i < NUM_WALRCV_WAKEUPS; ++i)
514 : 228940 : nextWakeup = Min(wakeup[i], nextWakeup);
515 : :
516 : : /* Calculate the nap time, clamping as necessary. */
1195 tgl@sss.pgh.pa.us 517 : 57235 : now = GetCurrentTimestamp();
518 : 57235 : nap = TimestampDifferenceMilliseconds(now, nextWakeup);
519 : :
520 : : /*
521 : : * Ideally we would reuse a WaitEventSet object repeatedly
522 : : * here to avoid the overheads of WaitLatchOrSocket on epoll
523 : : * systems, but we can't be sure that libpq (or any other
524 : : * walreceiver implementation) has the same socket (even if
525 : : * the fd is the same number, it may have been closed and
526 : : * reopened since the last time). In future, if there is a
527 : : * function for removing sockets from WaitEventSet, then we
528 : : * could add and remove just the socket each time, potentially
529 : : * avoiding some system calls.
530 : : */
3689 rhaas@postgresql.org 531 [ - + ]: 57235 : Assert(wait_fd != PGINVALID_SOCKET);
2000 fujii@postgresql.org 532 : 57235 : rc = WaitLatchOrSocket(MyLatch,
533 : : WL_EXIT_ON_PM_DEATH | WL_SOCKET_READABLE |
534 : : WL_TIMEOUT | WL_LATCH_SET,
535 : : wait_fd,
536 : : nap,
537 : : WAIT_EVENT_WAL_RECEIVER_MAIN);
3689 rhaas@postgresql.org 538 [ + + ]: 57235 : if (rc & WL_LATCH_SET)
539 : : {
2000 fujii@postgresql.org 540 : 12722 : ResetLatch(MyLatch);
396 heikki.linnakangas@i 541 [ + + ]: 12722 : CHECK_FOR_INTERRUPTS();
542 : :
40 fujii@postgresql.org 543 [ + + ]:GNC 12634 : if (walrcv->apply_reply_requested)
544 : : {
545 : : /*
546 : : * The recovery process has asked us to send apply
547 : : * feedback now. Make sure the flag is really set to
548 : : * false in shared memory before sending the reply, so
549 : : * we don't miss a new request for a reply.
550 : : */
551 : 12579 : walrcv->apply_reply_requested = false;
3689 rhaas@postgresql.org 552 :CBC 12579 : pg_memory_barrier();
40 fujii@postgresql.org 553 :GNC 12579 : XLogWalRcvSendReply(false, false, true);
554 : : }
555 : : }
3689 rhaas@postgresql.org 556 [ + + ]:CBC 57147 : if (rc & WL_TIMEOUT)
557 : : {
558 : : /*
559 : : * We didn't receive anything new. If we haven't heard
560 : : * anything from the server for more than
561 : : * wal_receiver_timeout / 2, ping the server. Also, if
562 : : * it's been longer than wal_receiver_status_interval
563 : : * since the last update we sent, send a status update to
564 : : * the primary anyway, to report any progress in applying
565 : : * WAL.
566 : : */
4724 bruce@momjian.us 567 : 4 : bool requestReply = false;
568 : :
569 : : /*
570 : : * Report pending statistics to the cumulative stats
571 : : * system. This location is useful for the report as it
572 : : * is not within a tight loop in the WAL receiver, to
573 : : * avoid bloating pgstats with requests, while also making
574 : : * sure that the reports happen each time a status update
575 : : * is sent.
576 : : */
426 michael@paquier.xyz 577 : 4 : pgstat_report_wal(false);
578 : :
579 : : /*
580 : : * Check if time since last receive from primary has
581 : : * reached the configured limit.
582 : : */
1195 tgl@sss.pgh.pa.us 583 : 4 : now = GetCurrentTimestamp();
1274 tmunro@postgresql.or 584 [ - + ]: 4 : if (now >= wakeup[WALRCV_WAKEUP_TERMINATE])
1274 tmunro@postgresql.or 585 [ # # ]:UBC 0 : ereport(ERROR,
586 : : (errcode(ERRCODE_CONNECTION_FAILURE),
587 : : errmsg("terminating walreceiver due to timeout")));
588 : :
589 : : /*
590 : : * If we didn't receive anything new for half of receiver
591 : : * replication timeout, then ping the server.
592 : : */
1274 tmunro@postgresql.or 593 [ - + ]:CBC 4 : if (now >= wakeup[WALRCV_WAKEUP_PING])
594 : : {
1274 tmunro@postgresql.or 595 :UBC 0 : requestReply = true;
1195 tgl@sss.pgh.pa.us 596 : 0 : wakeup[WALRCV_WAKEUP_PING] = TIMESTAMP_INFINITY;
597 : : }
598 : :
40 fujii@postgresql.org 599 :GNC 4 : XLogWalRcvSendReply(requestReply, requestReply, false);
4838 simon@2ndQuadrant.co 600 :CBC 4 : XLogWalRcvSendHSFeedback(false);
601 : : }
602 : : }
603 : :
604 : : /*
605 : : * The backend finished streaming. Exit streaming COPY-mode from
606 : : * our side, too.
607 : : */
3443 peter_e@gmx.net 608 : 46 : walrcv_endstreaming(wrconn, &primaryTLI);
609 : :
610 : : /*
611 : : * If the server had switched to a new timeline that we didn't
612 : : * know about when we began streaming, fetch its timeline history
613 : : * file now.
614 : : */
4855 heikki.linnakangas@i 615 : 14 : WalRcvFetchTimeLineHistoryFiles(startpointTLI, primaryTLI);
616 : : }
617 : : else
4891 heikki.linnakangas@i 618 [ # # ]:UBC 0 : ereport(LOG,
619 : : (errmsg("primary server contains no more WAL on requested timeline %u",
620 : : startpointTLI)));
621 : :
622 : : /*
623 : : * End of WAL reached on the requested timeline. Close the last
624 : : * segment, and await for new orders from the startup process.
625 : : */
4891 heikki.linnakangas@i 626 [ + + ]:CBC 14 : if (recvFile >= 0)
627 : : {
628 : : char xlogfname[MAXFNAMELEN];
629 : :
1642 rhaas@postgresql.org 630 : 13 : XLogWalRcvFlush(false, startpointTLI);
2345 michael@paquier.xyz 631 : 13 : XLogFileName(xlogfname, recvFileTLI, recvSegNo, wal_segment_size);
4891 heikki.linnakangas@i 632 [ - + ]: 13 : if (close(recvFile) != 0)
4891 heikki.linnakangas@i 633 [ # # ]:UBC 0 : ereport(PANIC,
634 : : (errcode_for_file_access(),
635 : : errmsg("could not close WAL segment %s: %m",
636 : : xlogfname)));
637 : :
638 : : /*
639 : : * Create .done file forcibly to prevent the streamed segment from
640 : : * being archived later.
641 : : */
4008 heikki.linnakangas@i 642 [ + - ]:CBC 13 : if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
643 : 13 : XLogArchiveForceDone(xlogfname);
644 : : else
1704 alvherre@alvh.no-ip. 645 :UBC 0 : XLogArchiveNotify(xlogfname);
646 : : }
4891 heikki.linnakangas@i 647 :CBC 14 : recvFile = -1;
648 : :
649 [ + + ]: 14 : elog(DEBUG1, "walreceiver ended streaming and awaits new instructions");
650 : 14 : WalRcvWaitForStartPosition(&startpoint, &startpointTLI);
651 : : }
652 : : /* not reached */
653 : : }
654 : :
655 : : /*
656 : : * Wait for startup process to set receiveStart and receiveStartTLI.
657 : : */
658 : : static void
659 : 14 : WalRcvWaitForStartPosition(XLogRecPtr *startpoint, TimeLineID *startpointTLI)
660 : : {
3864 rhaas@postgresql.org 661 : 14 : WalRcvData *walrcv = WalRcv;
662 : : int state;
663 : :
4891 heikki.linnakangas@i 664 [ - + ]: 14 : SpinLockAcquire(&walrcv->mutex);
665 : 14 : state = walrcv->walRcvState;
102 michael@paquier.xyz 666 [ - + - - ]:GNC 14 : if (state != WALRCV_STREAMING && state != WALRCV_CONNECTING)
667 : : {
4891 heikki.linnakangas@i 668 :UBC 0 : SpinLockRelease(&walrcv->mutex);
669 [ # # ]: 0 : if (state == WALRCV_STOPPING)
670 : 0 : proc_exit(0);
671 : : else
672 [ # # ]: 0 : elog(FATAL, "unexpected walreceiver state");
673 : : }
4891 heikki.linnakangas@i 674 :CBC 14 : walrcv->walRcvState = WALRCV_WAITING;
675 : 14 : walrcv->receiveStart = InvalidXLogRecPtr;
676 : 14 : walrcv->receiveStartTLI = 0;
677 : 14 : SpinLockRelease(&walrcv->mutex);
678 : :
2246 peter@eisentraut.org 679 : 14 : set_ps_display("idle");
680 : :
681 : : /*
682 : : * nudge startup process to notice that we've stopped streaming and are
683 : : * now waiting for instructions.
684 : : */
4891 heikki.linnakangas@i 685 : 14 : WakeupRecovery();
686 : : for (;;)
687 : : {
2000 fujii@postgresql.org 688 : 28 : ResetLatch(MyLatch);
689 : :
396 heikki.linnakangas@i 690 [ - + ]: 28 : CHECK_FOR_INTERRUPTS();
691 : :
4891 692 [ - + ]: 28 : SpinLockAcquire(&walrcv->mutex);
693 [ + + - + : 28 : Assert(walrcv->walRcvState == WALRCV_RESTARTING ||
- - ]
694 : : walrcv->walRcvState == WALRCV_WAITING ||
695 : : walrcv->walRcvState == WALRCV_STOPPING);
696 [ + + ]: 28 : if (walrcv->walRcvState == WALRCV_RESTARTING)
697 : : {
698 : : /*
699 : : * No need to handle changes in primary_conninfo or
700 : : * primary_slot_name here. Startup process will signal us to
701 : : * terminate in case those change.
702 : : */
703 : 14 : *startpoint = walrcv->receiveStart;
704 : 14 : *startpointTLI = walrcv->receiveStartTLI;
102 michael@paquier.xyz 705 :GNC 14 : walrcv->walRcvState = WALRCV_CONNECTING;
4891 heikki.linnakangas@i 706 :CBC 14 : SpinLockRelease(&walrcv->mutex);
707 : 14 : break;
708 : : }
709 [ - + ]: 14 : if (walrcv->walRcvState == WALRCV_STOPPING)
710 : : {
711 : : /*
712 : : * We should've received SIGTERM if the startup process wants us
713 : : * to die, but might as well check it here too.
714 : : */
4891 heikki.linnakangas@i 715 :UBC 0 : SpinLockRelease(&walrcv->mutex);
19 fujii@postgresql.org 716 :UNC 0 : proc_exit(1);
717 : : }
4891 heikki.linnakangas@i 718 :CBC 14 : SpinLockRelease(&walrcv->mutex);
719 : :
2000 fujii@postgresql.org 720 : 14 : (void) WaitLatch(MyLatch, WL_LATCH_SET | WL_EXIT_ON_PM_DEATH, 0,
721 : : WAIT_EVENT_WAL_RECEIVER_WAIT_START);
722 : : }
723 : :
4891 heikki.linnakangas@i 724 [ + - ]: 14 : if (update_process_title)
725 : : {
726 : : char activitymsg[50];
727 : :
302 alvherre@kurilemu.de 728 :GNC 14 : snprintf(activitymsg, sizeof(activitymsg), "restarting at %X/%08X",
1897 peter@eisentraut.org 729 :CBC 14 : LSN_FORMAT_ARGS(*startpoint));
2246 730 : 14 : set_ps_display(activitymsg);
731 : : }
4891 heikki.linnakangas@i 732 : 14 : }
733 : :
734 : : /*
735 : : * Fetch any missing timeline history files between 'first' and 'last'
736 : : * (inclusive) from the server.
737 : : */
738 : : static void
739 : 180 : WalRcvFetchTimeLineHistoryFiles(TimeLineID first, TimeLineID last)
740 : : {
741 : : TimeLineID tli;
742 : :
743 [ + + ]: 385 : for (tli = first; tli <= last; tli++)
744 : : {
745 : : /* there's no history file for timeline 1 */
4870 746 [ + + + + ]: 205 : if (tli != 1 && !existsTimeLineHistory(tli))
747 : : {
748 : : char *fname;
749 : : char *content;
750 : : int len;
751 : : char expectedfname[MAXFNAMELEN];
752 : :
4891 753 [ + - ]: 13 : ereport(LOG,
754 : : (errmsg("fetching timeline history file for timeline %u from primary server",
755 : : tli)));
756 : :
3443 peter_e@gmx.net 757 : 13 : walrcv_readtimelinehistoryfile(wrconn, tli, &fname, &content, &len);
758 : :
759 : : /*
760 : : * Check that the filename on the primary matches what we
761 : : * calculated ourselves. This is just a sanity check, it should
762 : : * always match.
763 : : */
4891 heikki.linnakangas@i 764 : 13 : TLHistoryFileName(expectedfname, tli);
765 [ - + ]: 13 : if (strcmp(fname, expectedfname) != 0)
4891 heikki.linnakangas@i 766 [ # # ]:UBC 0 : ereport(ERROR,
767 : : (errcode(ERRCODE_PROTOCOL_VIOLATION),
768 : : errmsg_internal("primary reported unexpected file name for timeline history file of timeline %u",
769 : : tli)));
770 : :
771 : : /*
772 : : * Write the file to pg_wal.
773 : : */
4891 heikki.linnakangas@i 774 :CBC 13 : writeTimeLineHistoryFile(tli, content, len);
775 : :
776 : : /*
777 : : * Mark the streamed history file as ready for archiving if
778 : : * archive_mode is always.
779 : : */
2044 fujii@postgresql.org 780 [ + - ]: 13 : if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
781 : 13 : XLogArchiveForceDone(fname);
782 : : else
1704 alvherre@alvh.no-ip. 783 :UBC 0 : XLogArchiveNotify(fname);
784 : :
4891 heikki.linnakangas@i 785 :CBC 13 : pfree(fname);
786 : 13 : pfree(content);
787 : : }
788 : : }
5954 789 : 180 : }
790 : :
791 : : /*
792 : : * Mark us as STOPPED in shared memory at exit.
793 : : */
794 : : static void
5942 795 : 240 : WalRcvDie(int code, Datum arg)
796 : : {
3864 rhaas@postgresql.org 797 : 240 : WalRcvData *walrcv = WalRcv;
1642 798 : 240 : TimeLineID *startpointTLI_p = (TimeLineID *) DatumGetPointer(arg);
799 : :
800 [ - + ]: 240 : Assert(*startpointTLI_p != 0);
801 : :
802 : : /* Ensure that all WAL records received are flushed to disk */
803 : 240 : XLogWalRcvFlush(true, *startpointTLI_p);
804 : :
805 : : /* Mark ourselves inactive in shared memory */
5954 heikki.linnakangas@i 806 [ - + ]: 240 : SpinLockAcquire(&walrcv->mutex);
4891 807 [ + + + + : 240 : Assert(walrcv->walRcvState == WALRCV_STREAMING ||
+ - + - +
- - + ]
808 : : walrcv->walRcvState == WALRCV_CONNECTING ||
809 : : walrcv->walRcvState == WALRCV_RESTARTING ||
810 : : walrcv->walRcvState == WALRCV_STARTING ||
811 : : walrcv->walRcvState == WALRCV_WAITING ||
812 : : walrcv->walRcvState == WALRCV_STOPPING);
813 [ - + ]: 240 : Assert(walrcv->pid == MyProcPid);
5942 814 : 240 : walrcv->walRcvState = WALRCV_STOPPED;
5954 815 : 240 : walrcv->pid = 0;
550 816 : 240 : walrcv->procno = INVALID_PROC_NUMBER;
3595 alvherre@alvh.no-ip. 817 : 240 : walrcv->ready_to_display = false;
5954 heikki.linnakangas@i 818 : 240 : SpinLockRelease(&walrcv->mutex);
819 : :
1880 tmunro@postgresql.or 820 : 240 : ConditionVariableBroadcast(&walrcv->walRcvStoppedCV);
821 : :
822 : : /* Terminate the connection gracefully. */
3443 peter_e@gmx.net 823 [ + + ]: 240 : if (wrconn != NULL)
824 : 152 : walrcv_disconnect(wrconn);
825 : :
826 : : /* Wake up the startup process to notice promptly that we're gone */
4891 heikki.linnakangas@i 827 : 240 : WakeupRecovery();
5954 828 : 240 : }
829 : :
830 : : /*
831 : : * Accept the message from XLOG stream, and process it.
832 : : */
833 : : static void
1642 rhaas@postgresql.org 834 : 108876 : XLogWalRcvProcessMsg(unsigned char type, char *buf, Size len, TimeLineID tli)
835 : : {
836 : : int hdrlen;
837 : : XLogRecPtr dataStart;
838 : : XLogRecPtr walEnd;
839 : : TimestampTz sendTime;
840 : : bool replyRequested;
841 : :
5935 heikki.linnakangas@i 842 [ + + - ]: 108876 : switch (type)
843 : : {
272 nathan@postgresql.or 844 :GNC 108610 : case PqReplMsg_WALData:
845 : : {
846 : : StringInfoData incoming_message;
847 : :
4927 heikki.linnakangas@i 848 :CBC 108610 : hdrlen = sizeof(int64) + sizeof(int64) + sizeof(int64);
849 [ - + ]: 108610 : if (len < hdrlen)
5912 bruce@momjian.us 850 [ # # ]:UBC 0 : ereport(ERROR,
851 : : (errcode(ERRCODE_PROTOCOL_VIOLATION),
852 : : errmsg_internal("invalid WAL message received from primary")));
853 : :
854 : : /* initialize a StringInfo with the given buffer */
910 drowley@postgresql.o 855 :CBC 108610 : initReadOnlyStringInfo(&incoming_message, buf, hdrlen);
856 : :
857 : : /* read the fields */
4927 heikki.linnakangas@i 858 : 108610 : dataStart = pq_getmsgint64(&incoming_message);
859 : 108610 : walEnd = pq_getmsgint64(&incoming_message);
3358 tgl@sss.pgh.pa.us 860 : 108610 : sendTime = pq_getmsgint64(&incoming_message);
4927 heikki.linnakangas@i 861 : 108610 : ProcessWalSndrMessage(walEnd, sendTime);
862 : :
863 : 108610 : buf += hdrlen;
864 : 108610 : len -= hdrlen;
1642 rhaas@postgresql.org 865 : 108610 : XLogWalRcvWrite(buf, len, dataStart, tli);
5912 bruce@momjian.us 866 : 108610 : break;
867 : : }
272 nathan@postgresql.or 868 :GNC 266 : case PqReplMsg_Keepalive:
869 : : {
870 : : StringInfoData incoming_message;
871 : :
4927 heikki.linnakangas@i 872 :CBC 266 : hdrlen = sizeof(int64) + sizeof(int64) + sizeof(char);
873 [ - + ]: 266 : if (len != hdrlen)
5239 simon@2ndQuadrant.co 874 [ # # ]:UBC 0 : ereport(ERROR,
875 : : (errcode(ERRCODE_PROTOCOL_VIOLATION),
876 : : errmsg_internal("invalid keepalive message received from primary")));
877 : :
878 : : /* initialize a StringInfo with the given buffer */
910 drowley@postgresql.o 879 :CBC 266 : initReadOnlyStringInfo(&incoming_message, buf, hdrlen);
880 : :
881 : : /* read the fields */
4927 heikki.linnakangas@i 882 : 266 : walEnd = pq_getmsgint64(&incoming_message);
3358 tgl@sss.pgh.pa.us 883 : 266 : sendTime = pq_getmsgint64(&incoming_message);
4927 heikki.linnakangas@i 884 : 266 : replyRequested = pq_getmsgbyte(&incoming_message);
885 : :
886 : 266 : ProcessWalSndrMessage(walEnd, sendTime);
887 : :
888 : : /* If the primary requested a reply, send one immediately */
889 [ + - ]: 266 : if (replyRequested)
40 fujii@postgresql.org 890 :GNC 266 : XLogWalRcvSendReply(true, false, false);
5239 simon@2ndQuadrant.co 891 :CBC 266 : break;
892 : : }
5935 heikki.linnakangas@i 893 :UBC 0 : default:
894 [ # # ]: 0 : ereport(ERROR,
895 : : (errcode(ERRCODE_PROTOCOL_VIOLATION),
896 : : errmsg_internal("invalid replication message type %d",
897 : : type)));
898 : : }
5935 heikki.linnakangas@i 899 :CBC 108876 : }
900 : :
901 : : /*
902 : : * Write XLOG data to disk.
903 : : */
904 : : static void
1642 rhaas@postgresql.org 905 : 108610 : XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr, TimeLineID tli)
906 : : {
907 : : int startoff;
908 : : int byteswritten;
909 : : instr_time start;
910 : :
911 [ - + ]: 108610 : Assert(tli != 0);
912 : :
5954 heikki.linnakangas@i 913 [ + + ]: 217658 : while (nbytes > 0)
914 : : {
915 : : int segbytes;
916 : :
917 : : /* Close the current segment if it's completed */
1699 fujii@postgresql.org 918 [ + + + + ]: 109048 : if (recvFile >= 0 && !XLByteInSeg(recptr, recvSegNo, wal_segment_size))
1642 rhaas@postgresql.org 919 : 438 : XLogWalRcvClose(recptr, tli);
920 : :
1699 fujii@postgresql.org 921 [ + + ]: 109048 : if (recvFile < 0)
922 : : {
923 : : /* Create/use new log file */
3150 andres@anarazel.de 924 : 893 : XLByteToSeg(recptr, recvSegNo, wal_segment_size);
1642 rhaas@postgresql.org 925 : 893 : recvFile = XLogFileInit(recvSegNo, tli);
926 : 893 : recvFileTLI = tli;
927 : : }
928 : :
929 : : /* Calculate the start offset of the received logs */
3150 andres@anarazel.de 930 : 109048 : startoff = XLogSegmentOffset(recptr, wal_segment_size);
931 : :
932 [ + + ]: 109048 : if (startoff + nbytes > wal_segment_size)
933 : 438 : segbytes = wal_segment_size - startoff;
934 : : else
5954 heikki.linnakangas@i 935 : 108610 : segbytes = nbytes;
936 : :
937 : : /* OK to write the logs */
938 : 109048 : errno = 0;
939 : :
940 : : /*
941 : : * Measure I/O timing to write WAL data, for pg_stat_io.
942 : : */
425 michael@paquier.xyz 943 : 109048 : start = pgstat_prepare_io_time(track_wal_io_timing);
944 : :
945 : 109048 : pgstat_report_wait_start(WAIT_EVENT_WAL_WRITE);
173 michael@paquier.xyz 946 :GNC 109048 : byteswritten = pg_pwrite(recvFile, buf, segbytes, (pgoff_t) startoff);
425 michael@paquier.xyz 947 :CBC 109048 : pgstat_report_wait_end();
948 : :
949 : 109048 : pgstat_count_io_op_time(IOOBJECT_WAL, IOCONTEXT_NORMAL,
950 : : IOOP_WRITE, start, 1, byteswritten);
951 : :
5954 heikki.linnakangas@i 952 [ - + ]: 109048 : if (byteswritten <= 0)
953 : : {
954 : : char xlogfname[MAXFNAMELEN];
955 : : int save_errno;
956 : :
957 : : /* if write didn't set errno, assume no disk space */
5954 heikki.linnakangas@i 958 [ # # ]:UBC 0 : if (errno == 0)
959 : 0 : errno = ENOSPC;
960 : :
2345 michael@paquier.xyz 961 : 0 : save_errno = errno;
962 : 0 : XLogFileName(xlogfname, recvFileTLI, recvSegNo, wal_segment_size);
963 : 0 : errno = save_errno;
5954 heikki.linnakangas@i 964 [ # # ]: 0 : ereport(PANIC,
965 : : (errcode_for_file_access(),
966 : : errmsg("could not write to WAL segment %s "
967 : : "at offset %d, length %d: %m",
968 : : xlogfname, startoff, segbytes)));
969 : : }
970 : :
971 : : /* Update state for write */
4876 alvherre@alvh.no-ip. 972 :CBC 109048 : recptr += byteswritten;
973 : :
5954 heikki.linnakangas@i 974 : 109048 : nbytes -= byteswritten;
975 : 109048 : buf += byteswritten;
976 : :
5912 bruce@momjian.us 977 : 109048 : LogstreamResult.Write = recptr;
978 : : }
979 : :
980 : : /* Update shared-memory status */
2 akorotkov@postgresql 981 :GNC 108610 : pg_atomic_write_membarrier_u64(&WalRcv->writtenUpto, LogstreamResult.Write);
982 : :
983 : : /*
984 : : * Wake up processes waiting for standby write LSN to reach current write
985 : : * position.
986 : : */
987 : 108610 : WaitLSNWakeup(WAIT_LSN_TYPE_STANDBY_WRITE, LogstreamResult.Write);
988 : :
989 : : /*
990 : : * Close the current segment if it's fully written up in the last cycle of
991 : : * the loop, to create its archive notification file soon. Otherwise WAL
992 : : * archiving of the segment will be delayed until any data in the next
993 : : * segment is received and written.
994 : : */
1699 fujii@postgresql.org 995 [ + - + + ]:CBC 108610 : if (recvFile >= 0 && !XLByteInSeg(recptr, recvSegNo, wal_segment_size))
1642 rhaas@postgresql.org 996 : 314 : XLogWalRcvClose(recptr, tli);
5954 heikki.linnakangas@i 997 : 108610 : }
998 : :
999 : : /*
1000 : : * Flush the log to disk.
1001 : : *
1002 : : * If we're in the midst of dying, it's unwise to do anything that might throw
1003 : : * an error, so we skip sending a reply in that case.
1004 : : */
1005 : : static void
1642 rhaas@postgresql.org 1006 : 44141 : XLogWalRcvFlush(bool dying, TimeLineID tli)
1007 : : {
1008 [ - + ]: 44141 : Assert(tli != 0);
1009 : :
4876 alvherre@alvh.no-ip. 1010 [ + + ]: 44141 : if (LogstreamResult.Flush < LogstreamResult.Write)
1011 : : {
3864 rhaas@postgresql.org 1012 : 43436 : WalRcvData *walrcv = WalRcv;
1013 : :
1642 1014 : 43436 : issue_xlog_fsync(recvFile, recvSegNo, tli);
1015 : :
5954 heikki.linnakangas@i 1016 : 43436 : LogstreamResult.Flush = LogstreamResult.Write;
1017 : :
1018 : : /* Update shared-memory status */
1019 [ - + ]: 43436 : SpinLockAcquire(&walrcv->mutex);
2218 tmunro@postgresql.or 1020 [ + - ]: 43436 : if (walrcv->flushedUpto < LogstreamResult.Flush)
1021 : : {
1022 : 43436 : walrcv->latestChunkStart = walrcv->flushedUpto;
1023 : 43436 : walrcv->flushedUpto = LogstreamResult.Flush;
1642 rhaas@postgresql.org 1024 : 43436 : walrcv->receivedTLI = tli;
1025 : : }
5954 heikki.linnakangas@i 1026 : 43436 : SpinLockRelease(&walrcv->mutex);
1027 : :
1028 : : /*
1029 : : * Wake up processes waiting for standby flush LSN to reach current
1030 : : * flush position.
1031 : : */
2 akorotkov@postgresql 1032 :GNC 43436 : WaitLSNWakeup(WAIT_LSN_TYPE_STANDBY_FLUSH, LogstreamResult.Flush);
1033 : :
1034 : : /* Signal the startup process and walsender that new WAL has arrived */
5711 heikki.linnakangas@i 1035 :CBC 43436 : WakeupRecovery();
5404 simon@2ndQuadrant.co 1036 [ + - + - ]: 43436 : if (AllowCascadeReplication())
1123 andres@anarazel.de 1037 : 43436 : WalSndWakeup(true, false);
1038 : :
1039 : : /* Report XLOG streaming progress in PS display */
5811 tgl@sss.pgh.pa.us 1040 [ + - ]: 43436 : if (update_process_title)
1041 : : {
1042 : : char activitymsg[50];
1043 : :
302 alvherre@kurilemu.de 1044 :GNC 43436 : snprintf(activitymsg, sizeof(activitymsg), "streaming %X/%08X",
1897 peter@eisentraut.org 1045 :CBC 43436 : LSN_FORMAT_ARGS(LogstreamResult.Write));
2246 1046 : 43436 : set_ps_display(activitymsg);
1047 : : }
1048 : :
1049 : : /* Also let the primary know that we made some progress */
5557 rhaas@postgresql.org 1050 [ + + ]: 43436 : if (!dying)
1051 : : {
40 fujii@postgresql.org 1052 :GNC 43434 : XLogWalRcvSendReply(false, false, false);
4492 heikki.linnakangas@i 1053 :CBC 43434 : XLogWalRcvSendHSFeedback(false);
1054 : : }
1055 : : }
5954 1056 : 44141 : }
1057 : :
1058 : : /*
1059 : : * Close the current segment.
1060 : : *
1061 : : * Flush the segment to disk before closing it. Otherwise we have to
1062 : : * reopen and fsync it later.
1063 : : *
1064 : : * Create an archive notification file since the segment is known completed.
1065 : : */
1066 : : static void
1642 rhaas@postgresql.org 1067 : 752 : XLogWalRcvClose(XLogRecPtr recptr, TimeLineID tli)
1068 : : {
1069 : : char xlogfname[MAXFNAMELEN];
1070 : :
1699 fujii@postgresql.org 1071 [ + - - + ]: 752 : Assert(recvFile >= 0 && !XLByteInSeg(recptr, recvSegNo, wal_segment_size));
1642 rhaas@postgresql.org 1072 [ - + ]: 752 : Assert(tli != 0);
1073 : :
1074 : : /*
1075 : : * fsync() and close current file before we switch to next one. We would
1076 : : * otherwise have to reopen this file to fsync it later
1077 : : */
1078 : 752 : XLogWalRcvFlush(false, tli);
1079 : :
1699 fujii@postgresql.org 1080 : 752 : XLogFileName(xlogfname, recvFileTLI, recvSegNo, wal_segment_size);
1081 : :
1082 : : /*
1083 : : * XLOG segment files will be re-read by recovery in startup process soon,
1084 : : * so we don't advise the OS to release cache pages associated with the
1085 : : * file like XLogFileClose() does.
1086 : : */
1087 [ - + ]: 752 : if (close(recvFile) != 0)
1699 fujii@postgresql.org 1088 [ # # ]:UBC 0 : ereport(PANIC,
1089 : : (errcode_for_file_access(),
1090 : : errmsg("could not close WAL segment %s: %m",
1091 : : xlogfname)));
1092 : :
1093 : : /*
1094 : : * Create .done file forcibly to prevent the streamed segment from being
1095 : : * archived later.
1096 : : */
1699 fujii@postgresql.org 1097 [ + - ]:CBC 752 : if (XLogArchiveMode != ARCHIVE_MODE_ALWAYS)
1098 : 752 : XLogArchiveForceDone(xlogfname);
1099 : : else
1699 fujii@postgresql.org 1100 :UBC 0 : XLogArchiveNotify(xlogfname);
1101 : :
1699 fujii@postgresql.org 1102 :CBC 752 : recvFile = -1;
1103 : 752 : }
1104 : :
1105 : : /*
1106 : : * Send reply message to primary, indicating our current WAL locations and
1107 : : * time.
1108 : : *
1109 : : * The message is sent if 'force' is set, if enough time has passed since the
1110 : : * last update to reach wal_receiver_status_interval, or if WAL locations have
1111 : : * advanced since the previous status update. If wal_receiver_status_interval
1112 : : * is disabled and 'force' is false, this function does nothing. Set 'force' to
1113 : : * send the message unconditionally.
1114 : : *
1115 : : * Whether WAL locations are considered "advanced" depends on 'checkApply'.
1116 : : * If 'checkApply' is false, only the write and flush locations are checked.
1117 : : * This should be used when the call is triggered by write/flush activity
1118 : : * (e.g., after walreceiver writes or flushes WAL), and avoids the
1119 : : * apply-location check, which requires a spinlock. If 'checkApply' is true,
1120 : : * the apply location is also considered. This should be used when the apply
1121 : : * location is expected to advance (e.g., when the startup process requests
1122 : : * an apply notification).
1123 : : *
1124 : : * If 'requestReply' is true, requests the server to reply immediately upon
1125 : : * receiving this message. This is used for heartbeats, when approaching
1126 : : * wal_receiver_timeout.
1127 : : */
1128 : : static void
40 fujii@postgresql.org 1129 :GNC 99585 : XLogWalRcvSendReply(bool force, bool requestReply, bool checkApply)
1130 : : {
1131 : : static XLogRecPtr writePtr = InvalidXLogRecPtr;
1132 : : static XLogRecPtr flushPtr = InvalidXLogRecPtr;
1133 : : static XLogRecPtr applyPtr = InvalidXLogRecPtr;
1134 : 99585 : XLogRecPtr latestApplyPtr = InvalidXLogRecPtr;
1135 : : TimestampTz now;
1136 : :
1137 : : /*
1138 : : * If the user doesn't want status to be reported to the primary, be sure
1139 : : * to exit before doing anything at all.
1140 : : */
4954 heikki.linnakangas@i 1141 [ + + - + ]:CBC 99585 : if (!force && wal_receiver_status_interval <= 0)
5563 heikki.linnakangas@i 1142 :UBC 0 : return;
1143 : :
1144 : : /* Get current timestamp. */
5563 heikki.linnakangas@i 1145 :CBC 99585 : now = GetCurrentTimestamp();
1146 : :
1147 : : /*
1148 : : * We can compare the write and flush positions to the last message we
1149 : : * sent without taking any lock, but the apply position requires a spin
1150 : : * lock, so we don't check that unless it is expected to advance since the
1151 : : * previous update, i.e., when 'checkApply' is true.
1152 : : */
40 fujii@postgresql.org 1153 [ + + + - ]:GNC 99585 : if (!force && now < wakeup[WALRCV_WAKEUP_REPLY])
1154 : : {
1155 [ + + ]: 99154 : if (checkApply)
1156 : 12579 : latestApplyPtr = GetXLogReplayRecPtr(NULL);
1157 : :
1158 [ + + ]: 99154 : if (writePtr == LogstreamResult.Write
1159 [ + + ]: 55561 : && flushPtr == LogstreamResult.Flush
1160 [ + + + + ]: 12879 : && (!checkApply || applyPtr == latestApplyPtr))
1161 : 1976 : return;
1162 : : }
1163 : :
1164 : : /* Make sure we wake up when it's time to send another reply. */
1274 tmunro@postgresql.or 1165 :CBC 97609 : WalRcvComputeNextWakeup(WALRCV_WAKEUP_REPLY, now);
1166 : :
1167 : : /* Construct a new message */
4927 heikki.linnakangas@i 1168 : 97609 : writePtr = LogstreamResult.Write;
1169 : 97609 : flushPtr = LogstreamResult.Flush;
19 fujii@postgresql.org 1170 :GNC 97609 : applyPtr = XLogRecPtrIsValid(latestApplyPtr) ?
1171 [ + + ]: 97609 : latestApplyPtr : GetXLogReplayRecPtr(NULL);
1172 : :
4927 heikki.linnakangas@i 1173 :CBC 97609 : resetStringInfo(&reply_message);
272 nathan@postgresql.or 1174 :GNC 97609 : pq_sendbyte(&reply_message, PqReplMsg_StandbyStatusUpdate);
4927 heikki.linnakangas@i 1175 :CBC 97609 : pq_sendint64(&reply_message, writePtr);
1176 : 97609 : pq_sendint64(&reply_message, flushPtr);
1177 : 97609 : pq_sendint64(&reply_message, applyPtr);
3358 tgl@sss.pgh.pa.us 1178 : 97609 : pq_sendint64(&reply_message, GetCurrentTimestamp());
4927 heikki.linnakangas@i 1179 : 97609 : pq_sendbyte(&reply_message, requestReply ? 1 : 0);
1180 : :
1181 : : /* Send it */
302 alvherre@kurilemu.de 1182 [ + + - + ]:GNC 97609 : elog(DEBUG2, "sending write %X/%08X flush %X/%08X apply %X/%08X%s",
1183 : : LSN_FORMAT_ARGS(writePtr),
1184 : : LSN_FORMAT_ARGS(flushPtr),
1185 : : LSN_FORMAT_ARGS(applyPtr),
1186 : : requestReply ? " (reply requested)" : "");
1187 : :
3443 peter_e@gmx.net 1188 :CBC 97609 : walrcv_send(wrconn, reply_message.data, reply_message.len);
1189 : : }
1190 : :
1191 : : /*
1192 : : * Send hot standby feedback message to primary, plus the current time,
1193 : : * in case they don't have a watch.
1194 : : *
1195 : : * If the user disables feedback, send one final message to tell sender
1196 : : * to forget about the xmin on this standby. We also send this message
1197 : : * on first connect because a previous connection might have set xmin
1198 : : * on a replication slot. (If we're not using a slot it's harmless to
1199 : : * send a feedback message explicitly setting InvalidTransactionId).
1200 : : */
1201 : : static void
4838 simon@2ndQuadrant.co 1202 : 43630 : XLogWalRcvSendHSFeedback(bool immed)
1203 : : {
1204 : : TimestampTz now;
1205 : : FullTransactionId nextFullXid;
1206 : : TransactionId nextXid;
1207 : : uint32 xmin_epoch,
1208 : : catalog_xmin_epoch;
1209 : : TransactionId xmin,
1210 : : catalog_xmin;
1211 : :
1212 : : /* initially true so we always send at least one feedback message */
1213 : : static bool primary_has_standby_xmin = true;
1214 : :
1215 : : /*
1216 : : * If the user doesn't want status to be reported to the primary, be sure
1217 : : * to exit before doing anything at all.
1218 : : */
1219 [ + - + + ]: 43630 : if ((wal_receiver_status_interval <= 0 || !hot_standby_feedback) &&
2151 andres@anarazel.de 1220 [ + + ]: 42959 : !primary_has_standby_xmin)
5555 simon@2ndQuadrant.co 1221 : 43456 : return;
1222 : :
1223 : : /* Get current timestamp. */
1224 : 809 : now = GetCurrentTimestamp();
1225 : :
1226 : : /* Send feedback at most once per wal_receiver_status_interval. */
1274 tmunro@postgresql.or 1227 [ + + + + ]: 809 : if (!immed && now < wakeup[WALRCV_WAKEUP_HSFEEDBACK])
1228 : 634 : return;
1229 : :
1230 : : /* Make sure we wake up when it's time to send feedback again. */
1231 : 175 : WalRcvComputeNextWakeup(WALRCV_WAKEUP_HSFEEDBACK, now);
1232 : :
1233 : : /*
1234 : : * If Hot Standby is not yet accepting connections there is nothing to
1235 : : * send. Check this after the interval has expired to reduce number of
1236 : : * calls.
1237 : : *
1238 : : * Bailing out here also ensures that we don't send feedback until we've
1239 : : * read our own replication slot state, so we don't tell the primary to
1240 : : * discard needed xmin or catalog_xmin from any slots that may exist on
1241 : : * this replica.
1242 : : */
5555 simon@2ndQuadrant.co 1243 [ + + ]: 175 : if (!HotStandbyActive())
1244 : 1 : return;
1245 : :
1246 : : /*
1247 : : * Make the expensive call to get the oldest xmin once we are certain
1248 : : * everything else has been checked.
1249 : : */
4838 1250 [ + + ]: 174 : if (hot_standby_feedback)
1251 : : {
2092 andres@anarazel.de 1252 : 51 : GetReplicationHorizons(&xmin, &catalog_xmin);
1253 : : }
1254 : : else
1255 : : {
4838 simon@2ndQuadrant.co 1256 : 123 : xmin = InvalidTransactionId;
3328 1257 : 123 : catalog_xmin = InvalidTransactionId;
1258 : : }
1259 : :
1260 : : /*
1261 : : * Get epoch and adjust if nextXid and oldestXmin are different sides of
1262 : : * the epoch boundary.
1263 : : */
2595 tmunro@postgresql.or 1264 : 174 : nextFullXid = ReadNextFullTransactionId();
1265 : 174 : nextXid = XidFromFullTransactionId(nextFullXid);
1266 : 174 : xmin_epoch = EpochFromFullTransactionId(nextFullXid);
3328 simon@2ndQuadrant.co 1267 : 174 : catalog_xmin_epoch = xmin_epoch;
5555 1268 [ - + ]: 174 : if (nextXid < xmin)
3275 bruce@momjian.us 1269 :UBC 0 : xmin_epoch--;
3328 simon@2ndQuadrant.co 1270 [ - + ]:CBC 174 : if (nextXid < catalog_xmin)
3275 bruce@momjian.us 1271 :UBC 0 : catalog_xmin_epoch--;
1272 : :
3328 simon@2ndQuadrant.co 1273 [ + + ]:CBC 174 : elog(DEBUG2, "sending hot standby feedback xmin %u epoch %u catalog_xmin %u catalog_xmin_epoch %u",
1274 : : xmin, xmin_epoch, catalog_xmin, catalog_xmin_epoch);
1275 : :
1276 : : /* Construct the message and send it. */
4927 heikki.linnakangas@i 1277 : 174 : resetStringInfo(&reply_message);
272 nathan@postgresql.or 1278 :GNC 174 : pq_sendbyte(&reply_message, PqReplMsg_HotStandbyFeedback);
3358 tgl@sss.pgh.pa.us 1279 :CBC 174 : pq_sendint64(&reply_message, GetCurrentTimestamp());
3128 andres@anarazel.de 1280 : 174 : pq_sendint32(&reply_message, xmin);
1281 : 174 : pq_sendint32(&reply_message, xmin_epoch);
1282 : 174 : pq_sendint32(&reply_message, catalog_xmin);
1283 : 174 : pq_sendint32(&reply_message, catalog_xmin_epoch);
3443 peter_e@gmx.net 1284 : 174 : walrcv_send(wrconn, reply_message.data, reply_message.len);
3328 simon@2ndQuadrant.co 1285 [ + + - + ]: 174 : if (TransactionIdIsValid(xmin) || TransactionIdIsValid(catalog_xmin))
2151 andres@anarazel.de 1286 : 51 : primary_has_standby_xmin = true;
1287 : : else
1288 : 123 : primary_has_standby_xmin = false;
1289 : : }
1290 : :
1291 : : /*
1292 : : * Update shared memory status upon receiving a message from primary.
1293 : : *
1294 : : * 'walEnd' and 'sendTime' are the end-of-WAL and timestamp of the latest
1295 : : * message, reported by primary.
1296 : : */
1297 : : static void
5239 simon@2ndQuadrant.co 1298 : 108876 : ProcessWalSndrMessage(XLogRecPtr walEnd, TimestampTz sendTime)
1299 : : {
3864 rhaas@postgresql.org 1300 : 108876 : WalRcvData *walrcv = WalRcv;
5239 simon@2ndQuadrant.co 1301 : 108876 : TimestampTz lastMsgReceiptTime = GetCurrentTimestamp();
1302 : :
1303 : : /* Update shared-memory status */
1304 [ + + ]: 108876 : SpinLockAcquire(&walrcv->mutex);
4876 alvherre@alvh.no-ip. 1305 [ + + ]: 108876 : if (walrcv->latestWalEnd < walEnd)
5017 simon@2ndQuadrant.co 1306 : 24862 : walrcv->latestWalEndTime = sendTime;
1307 : 108876 : walrcv->latestWalEnd = walEnd;
5239 1308 : 108876 : walrcv->lastMsgSendTime = sendTime;
1309 : 108876 : walrcv->lastMsgReceiptTime = lastMsgReceiptTime;
1310 : 108876 : SpinLockRelease(&walrcv->mutex);
1311 : :
1989 tgl@sss.pgh.pa.us 1312 [ + + ]: 108876 : if (message_level_is_interesting(DEBUG2))
1313 : : {
1314 : : char *sendtime;
1315 : : char *receipttime;
1316 : : int applyDelay;
1317 : :
1318 : : /* Copy because timestamptz_to_str returns a static buffer */
4414 1319 : 417 : sendtime = pstrdup(timestamptz_to_str(sendTime));
1320 : 417 : receipttime = pstrdup(timestamptz_to_str(lastMsgReceiptTime));
4070 ishii@postgresql.org 1321 : 417 : applyDelay = GetReplicationApplyDelay();
1322 : :
1323 : : /* apply delay is not available */
1324 [ + + ]: 417 : if (applyDelay == -1)
1325 [ + - ]: 3 : elog(DEBUG2, "sendtime %s receipttime %s replication apply delay (N/A) transfer latency %d ms",
1326 : : sendtime,
1327 : : receipttime,
1328 : : GetReplicationTransferLatency());
1329 : : else
1330 [ + - ]: 414 : elog(DEBUG2, "sendtime %s receipttime %s replication apply delay %d ms transfer latency %d ms",
1331 : : sendtime,
1332 : : receipttime,
1333 : : applyDelay,
1334 : : GetReplicationTransferLatency());
1335 : :
4414 tgl@sss.pgh.pa.us 1336 : 417 : pfree(sendtime);
1337 : 417 : pfree(receipttime);
1338 : : }
5239 simon@2ndQuadrant.co 1339 : 108876 : }
1340 : :
1341 : : /*
1342 : : * Compute the next wakeup time for a given wakeup reason. Can be called to
1343 : : * initialize a wakeup time, to adjust it for the next wakeup, or to
1344 : : * reinitialize it when GUCs have changed. We ask the caller to pass in the
1345 : : * value of "now" because this frequently avoids multiple calls of
1346 : : * GetCurrentTimestamp(). It had better be a reasonably up-to-date value
1347 : : * though.
1348 : : */
1349 : : static void
1274 tmunro@postgresql.or 1350 : 316304 : WalRcvComputeNextWakeup(WalRcvWakeupReason reason, TimestampTz now)
1351 : : {
1352 [ + + + + : 316304 : switch (reason)
- ]
1353 : : {
1354 : 109068 : case WALRCV_WAKEUP_TERMINATE:
1355 [ - + ]: 109068 : if (wal_receiver_timeout <= 0)
1195 tgl@sss.pgh.pa.us 1356 :UBC 0 : wakeup[reason] = TIMESTAMP_INFINITY;
1357 : : else
1195 tgl@sss.pgh.pa.us 1358 :CBC 109068 : wakeup[reason] = TimestampTzPlusMilliseconds(now, wal_receiver_timeout);
1274 tmunro@postgresql.or 1359 : 109068 : break;
1360 : 109068 : case WALRCV_WAKEUP_PING:
1361 [ - + ]: 109068 : if (wal_receiver_timeout <= 0)
1195 tgl@sss.pgh.pa.us 1362 :UBC 0 : wakeup[reason] = TIMESTAMP_INFINITY;
1363 : : else
1195 tgl@sss.pgh.pa.us 1364 :CBC 109068 : wakeup[reason] = TimestampTzPlusMilliseconds(now, wal_receiver_timeout / 2);
1274 tmunro@postgresql.or 1365 : 109068 : break;
1366 : 367 : case WALRCV_WAKEUP_HSFEEDBACK:
1367 [ + + - + ]: 367 : if (!hot_standby_feedback || wal_receiver_status_interval <= 0)
1195 tgl@sss.pgh.pa.us 1368 : 269 : wakeup[reason] = TIMESTAMP_INFINITY;
1369 : : else
1370 : 98 : wakeup[reason] = TimestampTzPlusSeconds(now, wal_receiver_status_interval);
1274 tmunro@postgresql.or 1371 : 367 : break;
1372 : 97801 : case WALRCV_WAKEUP_REPLY:
1373 [ - + ]: 97801 : if (wal_receiver_status_interval <= 0)
1195 tgl@sss.pgh.pa.us 1374 :UBC 0 : wakeup[reason] = TIMESTAMP_INFINITY;
1375 : : else
1195 tgl@sss.pgh.pa.us 1376 :CBC 97801 : wakeup[reason] = TimestampTzPlusSeconds(now, wal_receiver_status_interval);
1274 tmunro@postgresql.or 1377 : 97801 : break;
1378 : : /* there's intentionally no default: here */
1379 : : }
1380 : 316304 : }
1381 : :
1382 : : /*
1383 : : * Wake up the walreceiver main loop.
1384 : : *
1385 : : * This is called by the startup process whenever interesting xlog records
1386 : : * are applied, so that walreceiver can check if it needs to send an apply
1387 : : * notification back to the primary which may be waiting in a COMMIT with
1388 : : * synchronous_commit = remote_apply.
1389 : : */
1390 : : void
40 fujii@postgresql.org 1391 :GNC 12412 : WalRcvRequestApplyReply(void)
1392 : : {
1393 : : ProcNumber procno;
1394 : :
1395 : 12412 : WalRcv->apply_reply_requested = true;
1396 : : /* fetching the proc number is probably atomic, but don't rely on it */
3136 tgl@sss.pgh.pa.us 1397 [ - + ]:CBC 12412 : SpinLockAcquire(&WalRcv->mutex);
550 heikki.linnakangas@i 1398 : 12412 : procno = WalRcv->procno;
3136 tgl@sss.pgh.pa.us 1399 : 12412 : SpinLockRelease(&WalRcv->mutex);
550 heikki.linnakangas@i 1400 [ + + ]: 12412 : if (procno != INVALID_PROC_NUMBER)
1401 : 12233 : SetLatch(&GetPGProcByNumber(procno)->procLatch);
3689 rhaas@postgresql.org 1402 : 12412 : }
1403 : :
1404 : : /*
1405 : : * Return a string constant representing the state. This is used
1406 : : * in system functions and views, and should *not* be translated.
1407 : : */
1408 : : static const char *
3771 alvherre@alvh.no-ip. 1409 : 19 : WalRcvGetStateString(WalRcvState state)
1410 : : {
1411 [ - - - + : 19 : switch (state)
- - - - ]
1412 : : {
3771 alvherre@alvh.no-ip. 1413 :UBC 0 : case WALRCV_STOPPED:
1414 : 0 : return "stopped";
1415 : 0 : case WALRCV_STARTING:
1416 : 0 : return "starting";
102 michael@paquier.xyz 1417 :UNC 0 : case WALRCV_CONNECTING:
1418 : 0 : return "connecting";
3771 alvherre@alvh.no-ip. 1419 :CBC 19 : case WALRCV_STREAMING:
1420 : 19 : return "streaming";
3771 alvherre@alvh.no-ip. 1421 :UBC 0 : case WALRCV_WAITING:
1422 : 0 : return "waiting";
1423 : 0 : case WALRCV_RESTARTING:
1424 : 0 : return "restarting";
1425 : 0 : case WALRCV_STOPPING:
1426 : 0 : return "stopping";
1427 : : }
1428 : 0 : return "UNKNOWN";
1429 : : }
1430 : :
1431 : : /*
1432 : : * Returns activity of WAL receiver, including pid, state and xlog locations
1433 : : * received from the WAL sender of another server.
1434 : : */
1435 : : Datum
3771 alvherre@alvh.no-ip. 1436 :CBC 31 : pg_stat_get_wal_receiver(PG_FUNCTION_ARGS)
1437 : : {
1438 : : TupleDesc tupdesc;
1439 : : Datum *values;
1440 : : bool *nulls;
1441 : : int pid;
1442 : : bool ready_to_display;
1443 : : WalRcvState state;
1444 : : XLogRecPtr receive_start_lsn;
1445 : : TimeLineID receive_start_tli;
1446 : : XLogRecPtr written_lsn;
1447 : : XLogRecPtr flushed_lsn;
1448 : : TimeLineID received_tli;
1449 : : TimestampTz last_send_time;
1450 : : TimestampTz last_receipt_time;
1451 : : XLogRecPtr latest_end_lsn;
1452 : : TimestampTz latest_end_time;
1453 : : char sender_host[NI_MAXHOST];
2957 fujii@postgresql.org 1454 : 31 : int sender_port = 0;
1455 : : char slotname[NAMEDATALEN];
1456 : : char conninfo[MAXCONNINFO];
1457 : :
1458 : : /* Take a lock to ensure value consistency */
3231 alvherre@alvh.no-ip. 1459 [ - + ]: 31 : SpinLockAcquire(&WalRcv->mutex);
1460 : 31 : pid = (int) WalRcv->pid;
1461 : 31 : ready_to_display = WalRcv->ready_to_display;
1462 : 31 : state = WalRcv->walRcvState;
1463 : 31 : receive_start_lsn = WalRcv->receiveStart;
1464 : 31 : receive_start_tli = WalRcv->receiveStartTLI;
2179 michael@paquier.xyz 1465 : 31 : flushed_lsn = WalRcv->flushedUpto;
3231 alvherre@alvh.no-ip. 1466 : 31 : received_tli = WalRcv->receivedTLI;
1467 : 31 : last_send_time = WalRcv->lastMsgSendTime;
1468 : 31 : last_receipt_time = WalRcv->lastMsgReceiptTime;
1469 : 31 : latest_end_lsn = WalRcv->latestWalEnd;
1470 : 31 : latest_end_time = WalRcv->latestWalEndTime;
447 peter@eisentraut.org 1471 : 31 : strlcpy(slotname, WalRcv->slotname, sizeof(slotname));
1472 : 31 : strlcpy(sender_host, WalRcv->sender_host, sizeof(sender_host));
2957 fujii@postgresql.org 1473 : 31 : sender_port = WalRcv->sender_port;
447 peter@eisentraut.org 1474 : 31 : strlcpy(conninfo, WalRcv->conninfo, sizeof(conninfo));
3231 alvherre@alvh.no-ip. 1475 : 31 : SpinLockRelease(&WalRcv->mutex);
1476 : :
1477 : : /*
1478 : : * No WAL receiver (or not ready yet), just return a tuple with NULL
1479 : : * values
1480 : : */
1481 [ + + + + ]: 31 : if (pid == 0 || !ready_to_display)
3595 1482 : 12 : PG_RETURN_NULL();
1483 : :
1484 : : /*
1485 : : * Read "writtenUpto" without holding a spinlock. Note that it may not be
1486 : : * consistent with the other shared variables of the WAL receiver
1487 : : * protected by a spinlock, but this should not be used for data integrity
1488 : : * checks.
1489 : : */
1902 fujii@postgresql.org 1490 : 19 : written_lsn = pg_atomic_read_u64(&WalRcv->writtenUpto);
1491 : :
1492 : : /* determine result type */
3597 alvherre@alvh.no-ip. 1493 [ - + ]: 19 : if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
3597 alvherre@alvh.no-ip. 1494 [ # # ]:UBC 0 : elog(ERROR, "return type must be a row type");
1495 : :
146 michael@paquier.xyz 1496 :GNC 19 : values = palloc0_array(Datum, tupdesc->natts);
1497 : 19 : nulls = palloc0_array(bool, tupdesc->natts);
1498 : :
1499 : : /* Fetch values */
3231 alvherre@alvh.no-ip. 1500 :CBC 19 : values[0] = Int32GetDatum(pid);
1501 : :
1499 mail@joeconway.com 1502 [ - + ]: 19 : if (!has_privs_of_role(GetUserId(), ROLE_PG_READ_ALL_STATS))
1503 : : {
1504 : : /*
1505 : : * Only superusers and roles with privileges of pg_read_all_stats can
1506 : : * see details. Other users only get the pid value to know whether it
1507 : : * is a WAL receiver, but no details.
1508 : : */
1404 peter@eisentraut.org 1509 :UBC 0 : memset(&nulls[1], true, sizeof(bool) * (tupdesc->natts - 1));
1510 : : }
1511 : : else
1512 : : {
3771 alvherre@alvh.no-ip. 1513 :CBC 19 : values[1] = CStringGetTextDatum(WalRcvGetStateString(state));
1514 : :
180 alvherre@kurilemu.de 1515 [ - + ]:GNC 19 : if (!XLogRecPtrIsValid(receive_start_lsn))
3771 alvherre@alvh.no-ip. 1516 :UBC 0 : nulls[2] = true;
1517 : : else
3771 alvherre@alvh.no-ip. 1518 :CBC 19 : values[2] = LSNGetDatum(receive_start_lsn);
1519 : 19 : values[3] = Int32GetDatum(receive_start_tli);
180 alvherre@kurilemu.de 1520 [ - + ]:GNC 19 : if (!XLogRecPtrIsValid(written_lsn))
3771 alvherre@alvh.no-ip. 1521 :UBC 0 : nulls[4] = true;
1522 : : else
2179 michael@paquier.xyz 1523 :CBC 19 : values[4] = LSNGetDatum(written_lsn);
180 alvherre@kurilemu.de 1524 [ - + ]:GNC 19 : if (!XLogRecPtrIsValid(flushed_lsn))
2179 michael@paquier.xyz 1525 :UBC 0 : nulls[5] = true;
1526 : : else
2179 michael@paquier.xyz 1527 :CBC 19 : values[5] = LSNGetDatum(flushed_lsn);
1528 : 19 : values[6] = Int32GetDatum(received_tli);
3771 alvherre@alvh.no-ip. 1529 [ - + ]: 19 : if (last_send_time == 0)
2179 michael@paquier.xyz 1530 :UBC 0 : nulls[7] = true;
1531 : : else
2179 michael@paquier.xyz 1532 :CBC 19 : values[7] = TimestampTzGetDatum(last_send_time);
3771 alvherre@alvh.no-ip. 1533 [ - + ]: 19 : if (last_receipt_time == 0)
2179 michael@paquier.xyz 1534 :UBC 0 : nulls[8] = true;
1535 : : else
2179 michael@paquier.xyz 1536 :CBC 19 : values[8] = TimestampTzGetDatum(last_receipt_time);
180 alvherre@kurilemu.de 1537 [ - + ]:GNC 19 : if (!XLogRecPtrIsValid(latest_end_lsn))
2179 michael@paquier.xyz 1538 :UBC 0 : nulls[9] = true;
1539 : : else
2179 michael@paquier.xyz 1540 :CBC 19 : values[9] = LSNGetDatum(latest_end_lsn);
3771 alvherre@alvh.no-ip. 1541 [ - + ]: 19 : if (latest_end_time == 0)
2179 michael@paquier.xyz 1542 :UBC 0 : nulls[10] = true;
1543 : : else
2179 michael@paquier.xyz 1544 :CBC 19 : values[10] = TimestampTzGetDatum(latest_end_time);
3771 alvherre@alvh.no-ip. 1545 [ + + ]: 19 : if (*slotname == '\0')
2179 michael@paquier.xyz 1546 : 17 : nulls[11] = true;
1547 : : else
2179 michael@paquier.xyz 1548 :GBC 2 : values[11] = CStringGetTextDatum(slotname);
2957 fujii@postgresql.org 1549 [ - + ]:CBC 19 : if (*sender_host == '\0')
2179 michael@paquier.xyz 1550 :UBC 0 : nulls[12] = true;
1551 : : else
2179 michael@paquier.xyz 1552 :CBC 19 : values[12] = CStringGetTextDatum(sender_host);
2957 fujii@postgresql.org 1553 [ - + ]: 19 : if (sender_port == 0)
2179 michael@paquier.xyz 1554 :UBC 0 : nulls[13] = true;
1555 : : else
2179 michael@paquier.xyz 1556 :CBC 19 : values[13] = Int32GetDatum(sender_port);
2957 fujii@postgresql.org 1557 [ - + ]: 19 : if (*conninfo == '\0')
2179 michael@paquier.xyz 1558 :UBC 0 : nulls[14] = true;
1559 : : else
2179 michael@paquier.xyz 1560 :CBC 19 : values[14] = CStringGetTextDatum(conninfo);
1561 : : }
1562 : :
1563 : : /* Returns the record as Datum */
3231 alvherre@alvh.no-ip. 1564 : 19 : PG_RETURN_DATUM(HeapTupleGetDatum(heap_form_tuple(tupdesc, values, nulls)));
1565 : : }
|
