1 /* ----------
2 * pgstat.c
3 *
4 * All the statistics collector stuff hacked up in one big, ugly file.
5 *
6 * TODO: - Separate collector, postmaster and backend stuff
7 * into different files.
8 *
9 * - Add some automatic call for pgstat vacuuming.
10 *
11 * - Add a pgstat config column to pg_database, so this
12 * entire thing can be enabled/disabled on a per db basis.
13 *
14 * Copyright (c) 2001-2017, PostgreSQL Global Development Group
15 *
16 * src/backend/postmaster/pgstat.c
17 * ----------
18 */
19 #include "postgres.h"
20
21 #include <unistd.h>
22 #include <fcntl.h>
23 #include <sys/param.h>
24 #include <sys/time.h>
25 #include <sys/socket.h>
26 #include <netdb.h>
27 #include <netinet/in.h>
28 #include <arpa/inet.h>
29 #include <signal.h>
30 #include <time.h>
31 #ifdef HAVE_SYS_SELECT_H
32 #include <sys/select.h>
33 #endif
34
35 #include "pgstat.h"
36
37 #include "access/heapam.h"
38 #include "access/htup_details.h"
39 #include "access/transam.h"
40 #include "access/twophase_rmgr.h"
41 #include "access/xact.h"
42 #include "catalog/pg_database.h"
43 #include "catalog/pg_proc.h"
44 #include "common/ip.h"
45 #include "libpq/libpq.h"
46 #include "libpq/pqsignal.h"
47 #include "mb/pg_wchar.h"
48 #include "miscadmin.h"
49 #include "pg_trace.h"
50 #include "postmaster/autovacuum.h"
51 #include "postmaster/fork_process.h"
52 #include "postmaster/postmaster.h"
53 #include "replication/walsender.h"
54 #include "storage/backendid.h"
55 #include "storage/dsm.h"
56 #include "storage/fd.h"
57 #include "storage/ipc.h"
58 #include "storage/latch.h"
59 #include "storage/lmgr.h"
60 #include "storage/pg_shmem.h"
61 #include "storage/procsignal.h"
62 #include "storage/sinvaladt.h"
63 #include "utils/ascii.h"
64 #include "utils/guc.h"
65 #include "utils/memutils.h"
66 #include "utils/ps_status.h"
67 #include "utils/rel.h"
68 #include "utils/snapmgr.h"
69 #include "utils/timestamp.h"
70 #include "utils/tqual.h"
71
72
73 /* ----------
74 * Timer definitions.
75 * ----------
76 */
77 #define PGSTAT_STAT_INTERVAL 500 /* Minimum time between stats file
78 * updates; in milliseconds. */
79
80 #define PGSTAT_RETRY_DELAY 10 /* How long to wait between checks for a
81 * new file; in milliseconds. */
82
83 #define PGSTAT_MAX_WAIT_TIME 10000 /* Maximum time to wait for a stats
84 * file update; in milliseconds. */
85
86 #define PGSTAT_INQ_INTERVAL 640 /* How often to ping the collector for a
87 * new file; in milliseconds. */
88
89 #define PGSTAT_RESTART_INTERVAL 60 /* How often to attempt to restart a
90 * failed statistics collector; in
91 * seconds. */
92
93 #define PGSTAT_POLL_LOOP_COUNT (PGSTAT_MAX_WAIT_TIME / PGSTAT_RETRY_DELAY)
94 #define PGSTAT_INQ_LOOP_COUNT (PGSTAT_INQ_INTERVAL / PGSTAT_RETRY_DELAY)
95
96 /* Minimum receive buffer size for the collector's socket. */
97 #define PGSTAT_MIN_RCVBUF (100 * 1024)
98
99
100 /* ----------
101 * The initial size hints for the hash tables used in the collector.
102 * ----------
103 */
104 #define PGSTAT_DB_HASH_SIZE 16
105 #define PGSTAT_TAB_HASH_SIZE 512
106 #define PGSTAT_FUNCTION_HASH_SIZE 512
107
108
109 /* ----------
110 * Total number of backends including auxiliary
111 *
112 * We reserve a slot for each possible BackendId, plus one for each
113 * possible auxiliary process type. (This scheme assumes there is not
114 * more than one of any auxiliary process type at a time.) MaxBackends
115 * includes autovacuum workers and background workers as well.
116 * ----------
117 */
118 #define NumBackendStatSlots (MaxBackends + NUM_AUXPROCTYPES)
119
120
121 /* ----------
122 * GUC parameters
123 * ----------
124 */
125 bool pgstat_track_activities = false;
126 bool pgstat_track_counts = false;
127 int pgstat_track_functions = TRACK_FUNC_OFF;
128 int pgstat_track_activity_query_size = 1024;
129
130 /* ----------
131 * Built from GUC parameter
132 * ----------
133 */
134 char *pgstat_stat_directory = NULL;
135 char *pgstat_stat_filename = NULL;
136 char *pgstat_stat_tmpname = NULL;
137
138 /*
139 * BgWriter global statistics counters (unused in other processes).
140 * Stored directly in a stats message structure so it can be sent
141 * without needing to copy things around. We assume this inits to zeroes.
142 */
143 PgStat_MsgBgWriter BgWriterStats;
144
145 /* ----------
146 * Local data
147 * ----------
148 */
149 NON_EXEC_STATIC pgsocket pgStatSock = PGINVALID_SOCKET;
150
151 static struct sockaddr_storage pgStatAddr;
152
153 static time_t last_pgstat_start_time;
154
155 static bool pgStatRunningInCollector = false;
156
157 /*
158 * Structures in which backends store per-table info that's waiting to be
159 * sent to the collector.
160 *
161 * NOTE: once allocated, TabStatusArray structures are never moved or deleted
162 * for the life of the backend. Also, we zero out the t_id fields of the
163 * contained PgStat_TableStatus structs whenever they are not actively in use.
164 * This allows relcache pgstat_info pointers to be treated as long-lived data,
165 * avoiding repeated searches in pgstat_initstats() when a relation is
166 * repeatedly opened during a transaction.
167 */
168 #define TABSTAT_QUANTUM 100 /* we alloc this many at a time */
169
170 typedef struct TabStatusArray
171 {
172 struct TabStatusArray *tsa_next; /* link to next array, if any */
173 int tsa_used; /* # entries currently used */
174 PgStat_TableStatus tsa_entries[TABSTAT_QUANTUM]; /* per-table data */
175 } TabStatusArray;
176
177 static TabStatusArray *pgStatTabList = NULL;
178
179 /*
180 * pgStatTabHash entry: map from relation OID to PgStat_TableStatus pointer
181 */
182 typedef struct TabStatHashEntry
183 {
184 Oid t_id;
185 PgStat_TableStatus *tsa_entry;
186 } TabStatHashEntry;
187
188 /*
189 * Hash table for O(1) t_id -> tsa_entry lookup
190 */
191 static HTAB *pgStatTabHash = NULL;
192
193 /*
194 * Backends store per-function info that's waiting to be sent to the collector
195 * in this hash table (indexed by function OID).
196 */
197 static HTAB *pgStatFunctions = NULL;
198
199 /*
200 * Indicates if backend has some function stats that it hasn't yet
201 * sent to the collector.
202 */
203 static bool have_function_stats = false;
204
205 /*
206 * Tuple insertion/deletion counts for an open transaction can't be propagated
207 * into PgStat_TableStatus counters until we know if it is going to commit
208 * or abort. Hence, we keep these counts in per-subxact structs that live
209 * in TopTransactionContext. This data structure is designed on the assumption
210 * that subxacts won't usually modify very many tables.
211 */
212 typedef struct PgStat_SubXactStatus
213 {
214 int nest_level; /* subtransaction nest level */
215 struct PgStat_SubXactStatus *prev; /* higher-level subxact if any */
216 PgStat_TableXactStatus *first; /* head of list for this subxact */
217 } PgStat_SubXactStatus;
218
219 static PgStat_SubXactStatus *pgStatXactStack = NULL;
220
221 static int pgStatXactCommit = 0;
222 static int pgStatXactRollback = 0;
223 PgStat_Counter pgStatBlockReadTime = 0;
224 PgStat_Counter pgStatBlockWriteTime = 0;
225
226 /* Record that's written to 2PC state file when pgstat state is persisted */
227 typedef struct TwoPhasePgStatRecord
228 {
229 PgStat_Counter tuples_inserted; /* tuples inserted in xact */
230 PgStat_Counter tuples_updated; /* tuples updated in xact */
231 PgStat_Counter tuples_deleted; /* tuples deleted in xact */
232 PgStat_Counter inserted_pre_trunc; /* tuples inserted prior to truncate */
233 PgStat_Counter updated_pre_trunc; /* tuples updated prior to truncate */
234 PgStat_Counter deleted_pre_trunc; /* tuples deleted prior to truncate */
235 Oid t_id; /* table's OID */
236 bool t_shared; /* is it a shared catalog? */
237 bool t_truncated; /* was the relation truncated? */
238 } TwoPhasePgStatRecord;
239
240 /*
241 * Info about current "snapshot" of stats file
242 */
243 static MemoryContext pgStatLocalContext = NULL;
244 static HTAB *pgStatDBHash = NULL;
245
246 /* Status for backends including auxiliary */
247 static LocalPgBackendStatus *localBackendStatusTable = NULL;
248
249 /* Total number of backends including auxiliary */
250 static int localNumBackends = 0;
251
252 /*
253 * Cluster wide statistics, kept in the stats collector.
254 * Contains statistics that are not collected per database
255 * or per table.
256 */
257 static PgStat_ArchiverStats archiverStats;
258 static PgStat_GlobalStats globalStats;
259
260 /*
261 * List of OIDs of databases we need to write out. If an entry is InvalidOid,
262 * it means to write only the shared-catalog stats ("DB 0"); otherwise, we
263 * will write both that DB's data and the shared stats.
264 */
265 static List *pending_write_requests = NIL;
266
267 /* Signal handler flags */
268 static volatile bool need_exit = false;
269 static volatile bool got_SIGHUP = false;
270
271 /*
272 * Total time charged to functions so far in the current backend.
273 * We use this to help separate "self" and "other" time charges.
274 * (We assume this initializes to zero.)
275 */
276 static instr_time total_func_time;
277
278
279 /* ----------
280 * Local function forward declarations
281 * ----------
282 */
283 #ifdef EXEC_BACKEND
284 static pid_t pgstat_forkexec(void);
285 #endif
286
287 NON_EXEC_STATIC void PgstatCollectorMain(int argc, char *argv[]) pg_attribute_noreturn();
288 static void pgstat_exit(SIGNAL_ARGS);
289 static void pgstat_beshutdown_hook(int code, Datum arg);
290 static void pgstat_sighup_handler(SIGNAL_ARGS);
291
292 static PgStat_StatDBEntry *pgstat_get_db_entry(Oid databaseid, bool create);
293 static PgStat_StatTabEntry *pgstat_get_tab_entry(PgStat_StatDBEntry *dbentry,
294 Oid tableoid, bool create);
295 static void pgstat_write_statsfiles(bool permanent, bool allDbs);
296 static void pgstat_write_db_statsfile(PgStat_StatDBEntry *dbentry, bool permanent);
297 static HTAB *pgstat_read_statsfiles(Oid onlydb, bool permanent, bool deep);
298 static void pgstat_read_db_statsfile(Oid databaseid, HTAB *tabhash, HTAB *funchash, bool permanent);
299 static void backend_read_statsfile(void);
300 static void pgstat_read_current_status(void);
301
302 static bool pgstat_write_statsfile_needed(void);
303 static bool pgstat_db_requested(Oid databaseid);
304
305 static void pgstat_send_tabstat(PgStat_MsgTabstat *tsmsg);
306 static void pgstat_send_funcstats(void);
307 static HTAB *pgstat_collect_oids(Oid catalogid);
308
309 static PgStat_TableStatus *get_tabstat_entry(Oid rel_id, bool isshared);
310
311 static void pgstat_setup_memcxt(void);
312
313 static const char *pgstat_get_wait_activity(WaitEventActivity w);
314 static const char *pgstat_get_wait_client(WaitEventClient w);
315 static const char *pgstat_get_wait_ipc(WaitEventIPC w);
316 static const char *pgstat_get_wait_timeout(WaitEventTimeout w);
317 static const char *pgstat_get_wait_io(WaitEventIO w);
318
319 static void pgstat_setheader(PgStat_MsgHdr *hdr, StatMsgType mtype);
320 static void pgstat_send(void *msg, int len);
321
322 static void pgstat_recv_inquiry(PgStat_MsgInquiry *msg, int len);
323 static void pgstat_recv_tabstat(PgStat_MsgTabstat *msg, int len);
324 static void pgstat_recv_tabpurge(PgStat_MsgTabpurge *msg, int len);
325 static void pgstat_recv_dropdb(PgStat_MsgDropdb *msg, int len);
326 static void pgstat_recv_resetcounter(PgStat_MsgResetcounter *msg, int len);
327 static void pgstat_recv_resetsharedcounter(PgStat_MsgResetsharedcounter *msg, int len);
328 static void pgstat_recv_resetsinglecounter(PgStat_MsgResetsinglecounter *msg, int len);
329 static void pgstat_recv_autovac(PgStat_MsgAutovacStart *msg, int len);
330 static void pgstat_recv_vacuum(PgStat_MsgVacuum *msg, int len);
331 static void pgstat_recv_analyze(PgStat_MsgAnalyze *msg, int len);
332 static void pgstat_recv_archiver(PgStat_MsgArchiver *msg, int len);
333 static void pgstat_recv_bgwriter(PgStat_MsgBgWriter *msg, int len);
334 static void pgstat_recv_funcstat(PgStat_MsgFuncstat *msg, int len);
335 static void pgstat_recv_funcpurge(PgStat_MsgFuncpurge *msg, int len);
336 static void pgstat_recv_recoveryconflict(PgStat_MsgRecoveryConflict *msg, int len);
337 static void pgstat_recv_deadlock(PgStat_MsgDeadlock *msg, int len);
338 static void pgstat_recv_tempfile(PgStat_MsgTempFile *msg, int len);
339
340 /* ------------------------------------------------------------
341 * Public functions called from postmaster follow
342 * ------------------------------------------------------------
343 */
344
345 /* ----------
346 * pgstat_init() -
347 *
348 * Called from postmaster at startup. Create the resources required
349 * by the statistics collector process. If unable to do so, do not
350 * fail --- better to let the postmaster start with stats collection
351 * disabled.
352 * ----------
353 */
354 void
pgstat_init(void)355 pgstat_init(void)
356 {
357 ACCEPT_TYPE_ARG3 alen;
358 struct addrinfo *addrs = NULL,
359 *addr,
360 hints;
361 int ret;
362 fd_set rset;
363 struct timeval tv;
364 char test_byte;
365 int sel_res;
366 int tries = 0;
367
368 #define TESTBYTEVAL ((char) 199)
369
370 /*
371 * This static assertion verifies that we didn't mess up the calculations
372 * involved in selecting maximum payload sizes for our UDP messages.
373 * Because the only consequence of overrunning PGSTAT_MAX_MSG_SIZE would
374 * be silent performance loss from fragmentation, it seems worth having a
375 * compile-time cross-check that we didn't.
376 */
377 StaticAssertStmt(sizeof(PgStat_Msg) <= PGSTAT_MAX_MSG_SIZE,
378 "maximum stats message size exceeds PGSTAT_MAX_MSG_SIZE");
379
380 /*
381 * Create the UDP socket for sending and receiving statistic messages
382 */
383 hints.ai_flags = AI_PASSIVE;
384 hints.ai_family = AF_UNSPEC;
385 hints.ai_socktype = SOCK_DGRAM;
386 hints.ai_protocol = 0;
387 hints.ai_addrlen = 0;
388 hints.ai_addr = NULL;
389 hints.ai_canonname = NULL;
390 hints.ai_next = NULL;
391 ret = pg_getaddrinfo_all("localhost", NULL, &hints, &addrs);
392 if (ret || !addrs)
393 {
394 ereport(LOG,
395 (errmsg("could not resolve \"localhost\": %s",
396 gai_strerror(ret))));
397 goto startup_failed;
398 }
399
400 /*
401 * On some platforms, pg_getaddrinfo_all() may return multiple addresses
402 * only one of which will actually work (eg, both IPv6 and IPv4 addresses
403 * when kernel will reject IPv6). Worse, the failure may occur at the
404 * bind() or perhaps even connect() stage. So we must loop through the
405 * results till we find a working combination. We will generate LOG
406 * messages, but no error, for bogus combinations.
407 */
408 for (addr = addrs; addr; addr = addr->ai_next)
409 {
410 #ifdef HAVE_UNIX_SOCKETS
411 /* Ignore AF_UNIX sockets, if any are returned. */
412 if (addr->ai_family == AF_UNIX)
413 continue;
414 #endif
415
416 if (++tries > 1)
417 ereport(LOG,
418 (errmsg("trying another address for the statistics collector")));
419
420 /*
421 * Create the socket.
422 */
423 if ((pgStatSock = socket(addr->ai_family, SOCK_DGRAM, 0)) == PGINVALID_SOCKET)
424 {
425 ereport(LOG,
426 (errcode_for_socket_access(),
427 errmsg("could not create socket for statistics collector: %m")));
428 continue;
429 }
430
431 /*
432 * Bind it to a kernel assigned port on localhost and get the assigned
433 * port via getsockname().
434 */
435 if (bind(pgStatSock, addr->ai_addr, addr->ai_addrlen) < 0)
436 {
437 ereport(LOG,
438 (errcode_for_socket_access(),
439 errmsg("could not bind socket for statistics collector: %m")));
440 closesocket(pgStatSock);
441 pgStatSock = PGINVALID_SOCKET;
442 continue;
443 }
444
445 alen = sizeof(pgStatAddr);
446 if (getsockname(pgStatSock, (struct sockaddr *) &pgStatAddr, &alen) < 0)
447 {
448 ereport(LOG,
449 (errcode_for_socket_access(),
450 errmsg("could not get address of socket for statistics collector: %m")));
451 closesocket(pgStatSock);
452 pgStatSock = PGINVALID_SOCKET;
453 continue;
454 }
455
456 /*
457 * Connect the socket to its own address. This saves a few cycles by
458 * not having to respecify the target address on every send. This also
459 * provides a kernel-level check that only packets from this same
460 * address will be received.
461 */
462 if (connect(pgStatSock, (struct sockaddr *) &pgStatAddr, alen) < 0)
463 {
464 ereport(LOG,
465 (errcode_for_socket_access(),
466 errmsg("could not connect socket for statistics collector: %m")));
467 closesocket(pgStatSock);
468 pgStatSock = PGINVALID_SOCKET;
469 continue;
470 }
471
472 /*
473 * Try to send and receive a one-byte test message on the socket. This
474 * is to catch situations where the socket can be created but will not
475 * actually pass data (for instance, because kernel packet filtering
476 * rules prevent it).
477 */
478 test_byte = TESTBYTEVAL;
479
480 retry1:
481 if (send(pgStatSock, &test_byte, 1, 0) != 1)
482 {
483 if (errno == EINTR)
484 goto retry1; /* if interrupted, just retry */
485 ereport(LOG,
486 (errcode_for_socket_access(),
487 errmsg("could not send test message on socket for statistics collector: %m")));
488 closesocket(pgStatSock);
489 pgStatSock = PGINVALID_SOCKET;
490 continue;
491 }
492
493 /*
494 * There could possibly be a little delay before the message can be
495 * received. We arbitrarily allow up to half a second before deciding
496 * it's broken.
497 */
498 for (;;) /* need a loop to handle EINTR */
499 {
500 FD_ZERO(&rset);
501 FD_SET(pgStatSock, &rset);
502
503 tv.tv_sec = 0;
504 tv.tv_usec = 500000;
505 sel_res = select(pgStatSock + 1, &rset, NULL, NULL, &tv);
506 if (sel_res >= 0 || errno != EINTR)
507 break;
508 }
509 if (sel_res < 0)
510 {
511 ereport(LOG,
512 (errcode_for_socket_access(),
513 errmsg("select() failed in statistics collector: %m")));
514 closesocket(pgStatSock);
515 pgStatSock = PGINVALID_SOCKET;
516 continue;
517 }
518 if (sel_res == 0 || !FD_ISSET(pgStatSock, &rset))
519 {
520 /*
521 * This is the case we actually think is likely, so take pains to
522 * give a specific message for it.
523 *
524 * errno will not be set meaningfully here, so don't use it.
525 */
526 ereport(LOG,
527 (errcode(ERRCODE_CONNECTION_FAILURE),
528 errmsg("test message did not get through on socket for statistics collector")));
529 closesocket(pgStatSock);
530 pgStatSock = PGINVALID_SOCKET;
531 continue;
532 }
533
534 test_byte++; /* just make sure variable is changed */
535
536 retry2:
537 if (recv(pgStatSock, &test_byte, 1, 0) != 1)
538 {
539 if (errno == EINTR)
540 goto retry2; /* if interrupted, just retry */
541 ereport(LOG,
542 (errcode_for_socket_access(),
543 errmsg("could not receive test message on socket for statistics collector: %m")));
544 closesocket(pgStatSock);
545 pgStatSock = PGINVALID_SOCKET;
546 continue;
547 }
548
549 if (test_byte != TESTBYTEVAL) /* strictly paranoia ... */
550 {
551 ereport(LOG,
552 (errcode(ERRCODE_INTERNAL_ERROR),
553 errmsg("incorrect test message transmission on socket for statistics collector")));
554 closesocket(pgStatSock);
555 pgStatSock = PGINVALID_SOCKET;
556 continue;
557 }
558
559 /* If we get here, we have a working socket */
560 break;
561 }
562
563 /* Did we find a working address? */
564 if (!addr || pgStatSock == PGINVALID_SOCKET)
565 goto startup_failed;
566
567 /*
568 * Set the socket to non-blocking IO. This ensures that if the collector
569 * falls behind, statistics messages will be discarded; backends won't
570 * block waiting to send messages to the collector.
571 */
572 if (!pg_set_noblock(pgStatSock))
573 {
574 ereport(LOG,
575 (errcode_for_socket_access(),
576 errmsg("could not set statistics collector socket to nonblocking mode: %m")));
577 goto startup_failed;
578 }
579
580 /*
581 * Try to ensure that the socket's receive buffer is at least
582 * PGSTAT_MIN_RCVBUF bytes, so that it won't easily overflow and lose
583 * data. Use of UDP protocol means that we are willing to lose data under
584 * heavy load, but we don't want it to happen just because of ridiculously
585 * small default buffer sizes (such as 8KB on older Windows versions).
586 */
587 {
588 int old_rcvbuf;
589 int new_rcvbuf;
590 ACCEPT_TYPE_ARG3 rcvbufsize = sizeof(old_rcvbuf);
591
592 if (getsockopt(pgStatSock, SOL_SOCKET, SO_RCVBUF,
593 (char *) &old_rcvbuf, &rcvbufsize) < 0)
594 {
595 elog(LOG, "getsockopt(SO_RCVBUF) failed: %m");
596 /* if we can't get existing size, always try to set it */
597 old_rcvbuf = 0;
598 }
599
600 new_rcvbuf = PGSTAT_MIN_RCVBUF;
601 if (old_rcvbuf < new_rcvbuf)
602 {
603 if (setsockopt(pgStatSock, SOL_SOCKET, SO_RCVBUF,
604 (char *) &new_rcvbuf, sizeof(new_rcvbuf)) < 0)
605 elog(LOG, "setsockopt(SO_RCVBUF) failed: %m");
606 }
607 }
608
609 pg_freeaddrinfo_all(hints.ai_family, addrs);
610
611 return;
612
613 startup_failed:
614 ereport(LOG,
615 (errmsg("disabling statistics collector for lack of working socket")));
616
617 if (addrs)
618 pg_freeaddrinfo_all(hints.ai_family, addrs);
619
620 if (pgStatSock != PGINVALID_SOCKET)
621 closesocket(pgStatSock);
622 pgStatSock = PGINVALID_SOCKET;
623
624 /*
625 * Adjust GUC variables to suppress useless activity, and for debugging
626 * purposes (seeing track_counts off is a clue that we failed here). We
627 * use PGC_S_OVERRIDE because there is no point in trying to turn it back
628 * on from postgresql.conf without a restart.
629 */
630 SetConfigOption("track_counts", "off", PGC_INTERNAL, PGC_S_OVERRIDE);
631 }
632
633 /*
634 * subroutine for pgstat_reset_all
635 */
636 static void
pgstat_reset_remove_files(const char * directory)637 pgstat_reset_remove_files(const char *directory)
638 {
639 DIR *dir;
640 struct dirent *entry;
641 char fname[MAXPGPATH * 2];
642
643 dir = AllocateDir(directory);
644 while ((entry = ReadDir(dir, directory)) != NULL)
645 {
646 int nchars;
647 Oid tmp_oid;
648
649 /*
650 * Skip directory entries that don't match the file names we write.
651 * See get_dbstat_filename for the database-specific pattern.
652 */
653 if (strncmp(entry->d_name, "global.", 7) == 0)
654 nchars = 7;
655 else
656 {
657 nchars = 0;
658 (void) sscanf(entry->d_name, "db_%u.%n",
659 &tmp_oid, &nchars);
660 if (nchars <= 0)
661 continue;
662 /* %u allows leading whitespace, so reject that */
663 if (strchr("0123456789", entry->d_name[3]) == NULL)
664 continue;
665 }
666
667 if (strcmp(entry->d_name + nchars, "tmp") != 0 &&
668 strcmp(entry->d_name + nchars, "stat") != 0)
669 continue;
670
671 snprintf(fname, sizeof(fname), "%s/%s", directory,
672 entry->d_name);
673 unlink(fname);
674 }
675 FreeDir(dir);
676 }
677
678 /*
679 * pgstat_reset_all() -
680 *
681 * Remove the stats files. This is currently used only if WAL
682 * recovery is needed after a crash.
683 */
684 void
pgstat_reset_all(void)685 pgstat_reset_all(void)
686 {
687 pgstat_reset_remove_files(pgstat_stat_directory);
688 pgstat_reset_remove_files(PGSTAT_STAT_PERMANENT_DIRECTORY);
689 }
690
691 #ifdef EXEC_BACKEND
692
693 /*
694 * pgstat_forkexec() -
695 *
696 * Format up the arglist for, then fork and exec, statistics collector process
697 */
698 static pid_t
pgstat_forkexec(void)699 pgstat_forkexec(void)
700 {
701 char *av[10];
702 int ac = 0;
703
704 av[ac++] = "postgres";
705 av[ac++] = "--forkcol";
706 av[ac++] = NULL; /* filled in by postmaster_forkexec */
707
708 av[ac] = NULL;
709 Assert(ac < lengthof(av));
710
711 return postmaster_forkexec(ac, av);
712 }
713 #endif /* EXEC_BACKEND */
714
715
716 /*
717 * pgstat_start() -
718 *
719 * Called from postmaster at startup or after an existing collector
720 * died. Attempt to fire up a fresh statistics collector.
721 *
722 * Returns PID of child process, or 0 if fail.
723 *
724 * Note: if fail, we will be called again from the postmaster main loop.
725 */
726 int
pgstat_start(void)727 pgstat_start(void)
728 {
729 time_t curtime;
730 pid_t pgStatPid;
731
732 /*
733 * Check that the socket is there, else pgstat_init failed and we can do
734 * nothing useful.
735 */
736 if (pgStatSock == PGINVALID_SOCKET)
737 return 0;
738
739 /*
740 * Do nothing if too soon since last collector start. This is a safety
741 * valve to protect against continuous respawn attempts if the collector
742 * is dying immediately at launch. Note that since we will be re-called
743 * from the postmaster main loop, we will get another chance later.
744 */
745 curtime = time(NULL);
746 if ((unsigned int) (curtime - last_pgstat_start_time) <
747 (unsigned int) PGSTAT_RESTART_INTERVAL)
748 return 0;
749 last_pgstat_start_time = curtime;
750
751 /*
752 * Okay, fork off the collector.
753 */
754 #ifdef EXEC_BACKEND
755 switch ((pgStatPid = pgstat_forkexec()))
756 #else
757 switch ((pgStatPid = fork_process()))
758 #endif
759 {
760 case -1:
761 ereport(LOG,
762 (errmsg("could not fork statistics collector: %m")));
763 return 0;
764
765 #ifndef EXEC_BACKEND
766 case 0:
767 /* in postmaster child ... */
768 InitPostmasterChild();
769
770 /* Close the postmaster's sockets */
771 ClosePostmasterPorts(false);
772
773 /* Drop our connection to postmaster's shared memory, as well */
774 dsm_detach_all();
775 PGSharedMemoryDetach();
776
777 PgstatCollectorMain(0, NULL);
778 break;
779 #endif
780
781 default:
782 return (int) pgStatPid;
783 }
784
785 /* shouldn't get here */
786 return 0;
787 }
788
789 void
allow_immediate_pgstat_restart(void)790 allow_immediate_pgstat_restart(void)
791 {
792 last_pgstat_start_time = 0;
793 }
794
795 /* ------------------------------------------------------------
796 * Public functions used by backends follow
797 *------------------------------------------------------------
798 */
799
800
801 /* ----------
802 * pgstat_report_stat() -
803 *
804 * Must be called by processes that performs DML: tcop/postgres.c, logical
805 * receiver processes, SPI worker, etc. to send the so far collected
806 * per-table and function usage statistics to the collector. Note that this
807 * is called only when not within a transaction, so it is fair to use
808 * transaction stop time as an approximation of current time.
809 * ----------
810 */
811 void
pgstat_report_stat(bool force)812 pgstat_report_stat(bool force)
813 {
814 /* we assume this inits to all zeroes: */
815 static const PgStat_TableCounts all_zeroes;
816 static TimestampTz last_report = 0;
817
818 TimestampTz now;
819 PgStat_MsgTabstat regular_msg;
820 PgStat_MsgTabstat shared_msg;
821 TabStatusArray *tsa;
822 int i;
823
824 /* Don't expend a clock check if nothing to do */
825 if ((pgStatTabList == NULL || pgStatTabList->tsa_used == 0) &&
826 pgStatXactCommit == 0 && pgStatXactRollback == 0 &&
827 !have_function_stats)
828 return;
829
830 /*
831 * Don't send a message unless it's been at least PGSTAT_STAT_INTERVAL
832 * msec since we last sent one, or the caller wants to force stats out.
833 */
834 now = GetCurrentTransactionStopTimestamp();
835 if (!force &&
836 !TimestampDifferenceExceeds(last_report, now, PGSTAT_STAT_INTERVAL))
837 return;
838 last_report = now;
839
840 /*
841 * Destroy pgStatTabHash before we start invalidating PgStat_TableEntry
842 * entries it points to. (Should we fail partway through the loop below,
843 * it's okay to have removed the hashtable already --- the only
844 * consequence is we'd get multiple entries for the same table in the
845 * pgStatTabList, and that's safe.)
846 */
847 if (pgStatTabHash)
848 hash_destroy(pgStatTabHash);
849 pgStatTabHash = NULL;
850
851 /*
852 * Scan through the TabStatusArray struct(s) to find tables that actually
853 * have counts, and build messages to send. We have to separate shared
854 * relations from regular ones because the databaseid field in the message
855 * header has to depend on that.
856 */
857 regular_msg.m_databaseid = MyDatabaseId;
858 shared_msg.m_databaseid = InvalidOid;
859 regular_msg.m_nentries = 0;
860 shared_msg.m_nentries = 0;
861
862 for (tsa = pgStatTabList; tsa != NULL; tsa = tsa->tsa_next)
863 {
864 for (i = 0; i < tsa->tsa_used; i++)
865 {
866 PgStat_TableStatus *entry = &tsa->tsa_entries[i];
867 PgStat_MsgTabstat *this_msg;
868 PgStat_TableEntry *this_ent;
869
870 /* Shouldn't have any pending transaction-dependent counts */
871 Assert(entry->trans == NULL);
872
873 /*
874 * Ignore entries that didn't accumulate any actual counts, such
875 * as indexes that were opened by the planner but not used.
876 */
877 if (memcmp(&entry->t_counts, &all_zeroes,
878 sizeof(PgStat_TableCounts)) == 0)
879 continue;
880
881 /*
882 * OK, insert data into the appropriate message, and send if full.
883 */
884 this_msg = entry->t_shared ? &shared_msg : ®ular_msg;
885 this_ent = &this_msg->m_entry[this_msg->m_nentries];
886 this_ent->t_id = entry->t_id;
887 memcpy(&this_ent->t_counts, &entry->t_counts,
888 sizeof(PgStat_TableCounts));
889 if (++this_msg->m_nentries >= PGSTAT_NUM_TABENTRIES)
890 {
891 pgstat_send_tabstat(this_msg);
892 this_msg->m_nentries = 0;
893 }
894 }
895 /* zero out TableStatus structs after use */
896 MemSet(tsa->tsa_entries, 0,
897 tsa->tsa_used * sizeof(PgStat_TableStatus));
898 tsa->tsa_used = 0;
899 }
900
901 /*
902 * Send partial messages. Make sure that any pending xact commit/abort
903 * gets counted, even if there are no table stats to send.
904 */
905 if (regular_msg.m_nentries > 0 ||
906 pgStatXactCommit > 0 || pgStatXactRollback > 0)
907 pgstat_send_tabstat(®ular_msg);
908 if (shared_msg.m_nentries > 0)
909 pgstat_send_tabstat(&shared_msg);
910
911 /* Now, send function statistics */
912 pgstat_send_funcstats();
913 }
914
915 /*
916 * Subroutine for pgstat_report_stat: finish and send a tabstat message
917 */
918 static void
pgstat_send_tabstat(PgStat_MsgTabstat * tsmsg)919 pgstat_send_tabstat(PgStat_MsgTabstat *tsmsg)
920 {
921 int n;
922 int len;
923
924 /* It's unlikely we'd get here with no socket, but maybe not impossible */
925 if (pgStatSock == PGINVALID_SOCKET)
926 return;
927
928 /*
929 * Report and reset accumulated xact commit/rollback and I/O timings
930 * whenever we send a normal tabstat message
931 */
932 if (OidIsValid(tsmsg->m_databaseid))
933 {
934 tsmsg->m_xact_commit = pgStatXactCommit;
935 tsmsg->m_xact_rollback = pgStatXactRollback;
936 tsmsg->m_block_read_time = pgStatBlockReadTime;
937 tsmsg->m_block_write_time = pgStatBlockWriteTime;
938 pgStatXactCommit = 0;
939 pgStatXactRollback = 0;
940 pgStatBlockReadTime = 0;
941 pgStatBlockWriteTime = 0;
942 }
943 else
944 {
945 tsmsg->m_xact_commit = 0;
946 tsmsg->m_xact_rollback = 0;
947 tsmsg->m_block_read_time = 0;
948 tsmsg->m_block_write_time = 0;
949 }
950
951 n = tsmsg->m_nentries;
952 len = offsetof(PgStat_MsgTabstat, m_entry[0]) +
953 n * sizeof(PgStat_TableEntry);
954
955 pgstat_setheader(&tsmsg->m_hdr, PGSTAT_MTYPE_TABSTAT);
956 pgstat_send(tsmsg, len);
957 }
958
959 /*
960 * Subroutine for pgstat_report_stat: populate and send a function stat message
961 */
962 static void
pgstat_send_funcstats(void)963 pgstat_send_funcstats(void)
964 {
965 /* we assume this inits to all zeroes: */
966 static const PgStat_FunctionCounts all_zeroes;
967
968 PgStat_MsgFuncstat msg;
969 PgStat_BackendFunctionEntry *entry;
970 HASH_SEQ_STATUS fstat;
971
972 if (pgStatFunctions == NULL)
973 return;
974
975 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_FUNCSTAT);
976 msg.m_databaseid = MyDatabaseId;
977 msg.m_nentries = 0;
978
979 hash_seq_init(&fstat, pgStatFunctions);
980 while ((entry = (PgStat_BackendFunctionEntry *) hash_seq_search(&fstat)) != NULL)
981 {
982 PgStat_FunctionEntry *m_ent;
983
984 /* Skip it if no counts accumulated since last time */
985 if (memcmp(&entry->f_counts, &all_zeroes,
986 sizeof(PgStat_FunctionCounts)) == 0)
987 continue;
988
989 /* need to convert format of time accumulators */
990 m_ent = &msg.m_entry[msg.m_nentries];
991 m_ent->f_id = entry->f_id;
992 m_ent->f_numcalls = entry->f_counts.f_numcalls;
993 m_ent->f_total_time = INSTR_TIME_GET_MICROSEC(entry->f_counts.f_total_time);
994 m_ent->f_self_time = INSTR_TIME_GET_MICROSEC(entry->f_counts.f_self_time);
995
996 if (++msg.m_nentries >= PGSTAT_NUM_FUNCENTRIES)
997 {
998 pgstat_send(&msg, offsetof(PgStat_MsgFuncstat, m_entry[0]) +
999 msg.m_nentries * sizeof(PgStat_FunctionEntry));
1000 msg.m_nentries = 0;
1001 }
1002
1003 /* reset the entry's counts */
1004 MemSet(&entry->f_counts, 0, sizeof(PgStat_FunctionCounts));
1005 }
1006
1007 if (msg.m_nentries > 0)
1008 pgstat_send(&msg, offsetof(PgStat_MsgFuncstat, m_entry[0]) +
1009 msg.m_nentries * sizeof(PgStat_FunctionEntry));
1010
1011 have_function_stats = false;
1012 }
1013
1014
1015 /* ----------
1016 * pgstat_vacuum_stat() -
1017 *
1018 * Will tell the collector about objects he can get rid of.
1019 * ----------
1020 */
1021 void
pgstat_vacuum_stat(void)1022 pgstat_vacuum_stat(void)
1023 {
1024 HTAB *htab;
1025 PgStat_MsgTabpurge msg;
1026 PgStat_MsgFuncpurge f_msg;
1027 HASH_SEQ_STATUS hstat;
1028 PgStat_StatDBEntry *dbentry;
1029 PgStat_StatTabEntry *tabentry;
1030 PgStat_StatFuncEntry *funcentry;
1031 int len;
1032
1033 if (pgStatSock == PGINVALID_SOCKET)
1034 return;
1035
1036 /*
1037 * If not done for this transaction, read the statistics collector stats
1038 * file into some hash tables.
1039 */
1040 backend_read_statsfile();
1041
1042 /*
1043 * Read pg_database and make a list of OIDs of all existing databases
1044 */
1045 htab = pgstat_collect_oids(DatabaseRelationId);
1046
1047 /*
1048 * Search the database hash table for dead databases and tell the
1049 * collector to drop them.
1050 */
1051 hash_seq_init(&hstat, pgStatDBHash);
1052 while ((dbentry = (PgStat_StatDBEntry *) hash_seq_search(&hstat)) != NULL)
1053 {
1054 Oid dbid = dbentry->databaseid;
1055
1056 CHECK_FOR_INTERRUPTS();
1057
1058 /* the DB entry for shared tables (with InvalidOid) is never dropped */
1059 if (OidIsValid(dbid) &&
1060 hash_search(htab, (void *) &dbid, HASH_FIND, NULL) == NULL)
1061 pgstat_drop_database(dbid);
1062 }
1063
1064 /* Clean up */
1065 hash_destroy(htab);
1066
1067 /*
1068 * Lookup our own database entry; if not found, nothing more to do.
1069 */
1070 dbentry = (PgStat_StatDBEntry *) hash_search(pgStatDBHash,
1071 (void *) &MyDatabaseId,
1072 HASH_FIND, NULL);
1073 if (dbentry == NULL || dbentry->tables == NULL)
1074 return;
1075
1076 /*
1077 * Similarly to above, make a list of all known relations in this DB.
1078 */
1079 htab = pgstat_collect_oids(RelationRelationId);
1080
1081 /*
1082 * Initialize our messages table counter to zero
1083 */
1084 msg.m_nentries = 0;
1085
1086 /*
1087 * Check for all tables listed in stats hashtable if they still exist.
1088 */
1089 hash_seq_init(&hstat, dbentry->tables);
1090 while ((tabentry = (PgStat_StatTabEntry *) hash_seq_search(&hstat)) != NULL)
1091 {
1092 Oid tabid = tabentry->tableid;
1093
1094 CHECK_FOR_INTERRUPTS();
1095
1096 if (hash_search(htab, (void *) &tabid, HASH_FIND, NULL) != NULL)
1097 continue;
1098
1099 /*
1100 * Not there, so add this table's Oid to the message
1101 */
1102 msg.m_tableid[msg.m_nentries++] = tabid;
1103
1104 /*
1105 * If the message is full, send it out and reinitialize to empty
1106 */
1107 if (msg.m_nentries >= PGSTAT_NUM_TABPURGE)
1108 {
1109 len = offsetof(PgStat_MsgTabpurge, m_tableid[0])
1110 + msg.m_nentries * sizeof(Oid);
1111
1112 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_TABPURGE);
1113 msg.m_databaseid = MyDatabaseId;
1114 pgstat_send(&msg, len);
1115
1116 msg.m_nentries = 0;
1117 }
1118 }
1119
1120 /*
1121 * Send the rest
1122 */
1123 if (msg.m_nentries > 0)
1124 {
1125 len = offsetof(PgStat_MsgTabpurge, m_tableid[0])
1126 + msg.m_nentries * sizeof(Oid);
1127
1128 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_TABPURGE);
1129 msg.m_databaseid = MyDatabaseId;
1130 pgstat_send(&msg, len);
1131 }
1132
1133 /* Clean up */
1134 hash_destroy(htab);
1135
1136 /*
1137 * Now repeat the above steps for functions. However, we needn't bother
1138 * in the common case where no function stats are being collected.
1139 */
1140 if (dbentry->functions != NULL &&
1141 hash_get_num_entries(dbentry->functions) > 0)
1142 {
1143 htab = pgstat_collect_oids(ProcedureRelationId);
1144
1145 pgstat_setheader(&f_msg.m_hdr, PGSTAT_MTYPE_FUNCPURGE);
1146 f_msg.m_databaseid = MyDatabaseId;
1147 f_msg.m_nentries = 0;
1148
1149 hash_seq_init(&hstat, dbentry->functions);
1150 while ((funcentry = (PgStat_StatFuncEntry *) hash_seq_search(&hstat)) != NULL)
1151 {
1152 Oid funcid = funcentry->functionid;
1153
1154 CHECK_FOR_INTERRUPTS();
1155
1156 if (hash_search(htab, (void *) &funcid, HASH_FIND, NULL) != NULL)
1157 continue;
1158
1159 /*
1160 * Not there, so add this function's Oid to the message
1161 */
1162 f_msg.m_functionid[f_msg.m_nentries++] = funcid;
1163
1164 /*
1165 * If the message is full, send it out and reinitialize to empty
1166 */
1167 if (f_msg.m_nentries >= PGSTAT_NUM_FUNCPURGE)
1168 {
1169 len = offsetof(PgStat_MsgFuncpurge, m_functionid[0])
1170 + f_msg.m_nentries * sizeof(Oid);
1171
1172 pgstat_send(&f_msg, len);
1173
1174 f_msg.m_nentries = 0;
1175 }
1176 }
1177
1178 /*
1179 * Send the rest
1180 */
1181 if (f_msg.m_nentries > 0)
1182 {
1183 len = offsetof(PgStat_MsgFuncpurge, m_functionid[0])
1184 + f_msg.m_nentries * sizeof(Oid);
1185
1186 pgstat_send(&f_msg, len);
1187 }
1188
1189 hash_destroy(htab);
1190 }
1191 }
1192
1193
1194 /* ----------
1195 * pgstat_collect_oids() -
1196 *
1197 * Collect the OIDs of all objects listed in the specified system catalog
1198 * into a temporary hash table. Caller should hash_destroy the result
1199 * when done with it. (However, we make the table in CurrentMemoryContext
1200 * so that it will be freed properly in event of an error.)
1201 * ----------
1202 */
1203 static HTAB *
pgstat_collect_oids(Oid catalogid)1204 pgstat_collect_oids(Oid catalogid)
1205 {
1206 HTAB *htab;
1207 HASHCTL hash_ctl;
1208 Relation rel;
1209 HeapScanDesc scan;
1210 HeapTuple tup;
1211 Snapshot snapshot;
1212
1213 memset(&hash_ctl, 0, sizeof(hash_ctl));
1214 hash_ctl.keysize = sizeof(Oid);
1215 hash_ctl.entrysize = sizeof(Oid);
1216 hash_ctl.hcxt = CurrentMemoryContext;
1217 htab = hash_create("Temporary table of OIDs",
1218 PGSTAT_TAB_HASH_SIZE,
1219 &hash_ctl,
1220 HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
1221
1222 rel = heap_open(catalogid, AccessShareLock);
1223 snapshot = RegisterSnapshot(GetLatestSnapshot());
1224 scan = heap_beginscan(rel, snapshot, 0, NULL);
1225 while ((tup = heap_getnext(scan, ForwardScanDirection)) != NULL)
1226 {
1227 Oid thisoid = HeapTupleGetOid(tup);
1228
1229 CHECK_FOR_INTERRUPTS();
1230
1231 (void) hash_search(htab, (void *) &thisoid, HASH_ENTER, NULL);
1232 }
1233 heap_endscan(scan);
1234 UnregisterSnapshot(snapshot);
1235 heap_close(rel, AccessShareLock);
1236
1237 return htab;
1238 }
1239
1240
1241 /* ----------
1242 * pgstat_drop_database() -
1243 *
1244 * Tell the collector that we just dropped a database.
1245 * (If the message gets lost, we will still clean the dead DB eventually
1246 * via future invocations of pgstat_vacuum_stat().)
1247 * ----------
1248 */
1249 void
pgstat_drop_database(Oid databaseid)1250 pgstat_drop_database(Oid databaseid)
1251 {
1252 PgStat_MsgDropdb msg;
1253
1254 if (pgStatSock == PGINVALID_SOCKET)
1255 return;
1256
1257 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_DROPDB);
1258 msg.m_databaseid = databaseid;
1259 pgstat_send(&msg, sizeof(msg));
1260 }
1261
1262
1263 /* ----------
1264 * pgstat_drop_relation() -
1265 *
1266 * Tell the collector that we just dropped a relation.
1267 * (If the message gets lost, we will still clean the dead entry eventually
1268 * via future invocations of pgstat_vacuum_stat().)
1269 *
1270 * Currently not used for lack of any good place to call it; we rely
1271 * entirely on pgstat_vacuum_stat() to clean out stats for dead rels.
1272 * ----------
1273 */
1274 #ifdef NOT_USED
1275 void
pgstat_drop_relation(Oid relid)1276 pgstat_drop_relation(Oid relid)
1277 {
1278 PgStat_MsgTabpurge msg;
1279 int len;
1280
1281 if (pgStatSock == PGINVALID_SOCKET)
1282 return;
1283
1284 msg.m_tableid[0] = relid;
1285 msg.m_nentries = 1;
1286
1287 len = offsetof(PgStat_MsgTabpurge, m_tableid[0]) + sizeof(Oid);
1288
1289 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_TABPURGE);
1290 msg.m_databaseid = MyDatabaseId;
1291 pgstat_send(&msg, len);
1292 }
1293 #endif /* NOT_USED */
1294
1295
1296 /* ----------
1297 * pgstat_reset_counters() -
1298 *
1299 * Tell the statistics collector to reset counters for our database.
1300 *
1301 * Permission checking for this function is managed through the normal
1302 * GRANT system.
1303 * ----------
1304 */
1305 void
pgstat_reset_counters(void)1306 pgstat_reset_counters(void)
1307 {
1308 PgStat_MsgResetcounter msg;
1309
1310 if (pgStatSock == PGINVALID_SOCKET)
1311 return;
1312
1313 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_RESETCOUNTER);
1314 msg.m_databaseid = MyDatabaseId;
1315 pgstat_send(&msg, sizeof(msg));
1316 }
1317
1318 /* ----------
1319 * pgstat_reset_shared_counters() -
1320 *
1321 * Tell the statistics collector to reset cluster-wide shared counters.
1322 *
1323 * Permission checking for this function is managed through the normal
1324 * GRANT system.
1325 * ----------
1326 */
1327 void
pgstat_reset_shared_counters(const char * target)1328 pgstat_reset_shared_counters(const char *target)
1329 {
1330 PgStat_MsgResetsharedcounter msg;
1331
1332 if (pgStatSock == PGINVALID_SOCKET)
1333 return;
1334
1335 if (strcmp(target, "archiver") == 0)
1336 msg.m_resettarget = RESET_ARCHIVER;
1337 else if (strcmp(target, "bgwriter") == 0)
1338 msg.m_resettarget = RESET_BGWRITER;
1339 else
1340 ereport(ERROR,
1341 (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1342 errmsg("unrecognized reset target: \"%s\"", target),
1343 errhint("Target must be \"archiver\" or \"bgwriter\".")));
1344
1345 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_RESETSHAREDCOUNTER);
1346 pgstat_send(&msg, sizeof(msg));
1347 }
1348
1349 /* ----------
1350 * pgstat_reset_single_counter() -
1351 *
1352 * Tell the statistics collector to reset a single counter.
1353 *
1354 * Permission checking for this function is managed through the normal
1355 * GRANT system.
1356 * ----------
1357 */
1358 void
pgstat_reset_single_counter(Oid objoid,PgStat_Single_Reset_Type type)1359 pgstat_reset_single_counter(Oid objoid, PgStat_Single_Reset_Type type)
1360 {
1361 PgStat_MsgResetsinglecounter msg;
1362
1363 if (pgStatSock == PGINVALID_SOCKET)
1364 return;
1365
1366 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_RESETSINGLECOUNTER);
1367 msg.m_databaseid = MyDatabaseId;
1368 msg.m_resettype = type;
1369 msg.m_objectid = objoid;
1370
1371 pgstat_send(&msg, sizeof(msg));
1372 }
1373
1374 /* ----------
1375 * pgstat_report_autovac() -
1376 *
1377 * Called from autovacuum.c to report startup of an autovacuum process.
1378 * We are called before InitPostgres is done, so can't rely on MyDatabaseId;
1379 * the db OID must be passed in, instead.
1380 * ----------
1381 */
1382 void
pgstat_report_autovac(Oid dboid)1383 pgstat_report_autovac(Oid dboid)
1384 {
1385 PgStat_MsgAutovacStart msg;
1386
1387 if (pgStatSock == PGINVALID_SOCKET)
1388 return;
1389
1390 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_AUTOVAC_START);
1391 msg.m_databaseid = dboid;
1392 msg.m_start_time = GetCurrentTimestamp();
1393
1394 pgstat_send(&msg, sizeof(msg));
1395 }
1396
1397
1398 /* ---------
1399 * pgstat_report_vacuum() -
1400 *
1401 * Tell the collector about the table we just vacuumed.
1402 * ---------
1403 */
1404 void
pgstat_report_vacuum(Oid tableoid,bool shared,PgStat_Counter livetuples,PgStat_Counter deadtuples)1405 pgstat_report_vacuum(Oid tableoid, bool shared,
1406 PgStat_Counter livetuples, PgStat_Counter deadtuples)
1407 {
1408 PgStat_MsgVacuum msg;
1409
1410 if (pgStatSock == PGINVALID_SOCKET || !pgstat_track_counts)
1411 return;
1412
1413 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_VACUUM);
1414 msg.m_databaseid = shared ? InvalidOid : MyDatabaseId;
1415 msg.m_tableoid = tableoid;
1416 msg.m_autovacuum = IsAutoVacuumWorkerProcess();
1417 msg.m_vacuumtime = GetCurrentTimestamp();
1418 msg.m_live_tuples = livetuples;
1419 msg.m_dead_tuples = deadtuples;
1420 pgstat_send(&msg, sizeof(msg));
1421 }
1422
1423 /* --------
1424 * pgstat_report_analyze() -
1425 *
1426 * Tell the collector about the table we just analyzed.
1427 *
1428 * Caller must provide new live- and dead-tuples estimates, as well as a
1429 * flag indicating whether to reset the changes_since_analyze counter.
1430 * --------
1431 */
1432 void
pgstat_report_analyze(Relation rel,PgStat_Counter livetuples,PgStat_Counter deadtuples,bool resetcounter)1433 pgstat_report_analyze(Relation rel,
1434 PgStat_Counter livetuples, PgStat_Counter deadtuples,
1435 bool resetcounter)
1436 {
1437 PgStat_MsgAnalyze msg;
1438
1439 if (pgStatSock == PGINVALID_SOCKET || !pgstat_track_counts)
1440 return;
1441
1442 /*
1443 * Unlike VACUUM, ANALYZE might be running inside a transaction that has
1444 * already inserted and/or deleted rows in the target table. ANALYZE will
1445 * have counted such rows as live or dead respectively. Because we will
1446 * report our counts of such rows at transaction end, we should subtract
1447 * off these counts from what we send to the collector now, else they'll
1448 * be double-counted after commit. (This approach also ensures that the
1449 * collector ends up with the right numbers if we abort instead of
1450 * committing.)
1451 */
1452 if (rel->pgstat_info != NULL)
1453 {
1454 PgStat_TableXactStatus *trans;
1455
1456 for (trans = rel->pgstat_info->trans; trans; trans = trans->upper)
1457 {
1458 livetuples -= trans->tuples_inserted - trans->tuples_deleted;
1459 deadtuples -= trans->tuples_updated + trans->tuples_deleted;
1460 }
1461 /* count stuff inserted by already-aborted subxacts, too */
1462 deadtuples -= rel->pgstat_info->t_counts.t_delta_dead_tuples;
1463 /* Since ANALYZE's counts are estimates, we could have underflowed */
1464 livetuples = Max(livetuples, 0);
1465 deadtuples = Max(deadtuples, 0);
1466 }
1467
1468 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_ANALYZE);
1469 msg.m_databaseid = rel->rd_rel->relisshared ? InvalidOid : MyDatabaseId;
1470 msg.m_tableoid = RelationGetRelid(rel);
1471 msg.m_autovacuum = IsAutoVacuumWorkerProcess();
1472 msg.m_resetcounter = resetcounter;
1473 msg.m_analyzetime = GetCurrentTimestamp();
1474 msg.m_live_tuples = livetuples;
1475 msg.m_dead_tuples = deadtuples;
1476 pgstat_send(&msg, sizeof(msg));
1477 }
1478
1479 /* --------
1480 * pgstat_report_recovery_conflict() -
1481 *
1482 * Tell the collector about a Hot Standby recovery conflict.
1483 * --------
1484 */
1485 void
pgstat_report_recovery_conflict(int reason)1486 pgstat_report_recovery_conflict(int reason)
1487 {
1488 PgStat_MsgRecoveryConflict msg;
1489
1490 if (pgStatSock == PGINVALID_SOCKET || !pgstat_track_counts)
1491 return;
1492
1493 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_RECOVERYCONFLICT);
1494 msg.m_databaseid = MyDatabaseId;
1495 msg.m_reason = reason;
1496 pgstat_send(&msg, sizeof(msg));
1497 }
1498
1499 /* --------
1500 * pgstat_report_deadlock() -
1501 *
1502 * Tell the collector about a deadlock detected.
1503 * --------
1504 */
1505 void
pgstat_report_deadlock(void)1506 pgstat_report_deadlock(void)
1507 {
1508 PgStat_MsgDeadlock msg;
1509
1510 if (pgStatSock == PGINVALID_SOCKET || !pgstat_track_counts)
1511 return;
1512
1513 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_DEADLOCK);
1514 msg.m_databaseid = MyDatabaseId;
1515 pgstat_send(&msg, sizeof(msg));
1516 }
1517
1518 /* --------
1519 * pgstat_report_tempfile() -
1520 *
1521 * Tell the collector about a temporary file.
1522 * --------
1523 */
1524 void
pgstat_report_tempfile(size_t filesize)1525 pgstat_report_tempfile(size_t filesize)
1526 {
1527 PgStat_MsgTempFile msg;
1528
1529 if (pgStatSock == PGINVALID_SOCKET || !pgstat_track_counts)
1530 return;
1531
1532 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_TEMPFILE);
1533 msg.m_databaseid = MyDatabaseId;
1534 msg.m_filesize = filesize;
1535 pgstat_send(&msg, sizeof(msg));
1536 }
1537
1538
1539 /* ----------
1540 * pgstat_ping() -
1541 *
1542 * Send some junk data to the collector to increase traffic.
1543 * ----------
1544 */
1545 void
pgstat_ping(void)1546 pgstat_ping(void)
1547 {
1548 PgStat_MsgDummy msg;
1549
1550 if (pgStatSock == PGINVALID_SOCKET)
1551 return;
1552
1553 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_DUMMY);
1554 pgstat_send(&msg, sizeof(msg));
1555 }
1556
1557 /* ----------
1558 * pgstat_send_inquiry() -
1559 *
1560 * Notify collector that we need fresh data.
1561 * ----------
1562 */
1563 static void
pgstat_send_inquiry(TimestampTz clock_time,TimestampTz cutoff_time,Oid databaseid)1564 pgstat_send_inquiry(TimestampTz clock_time, TimestampTz cutoff_time, Oid databaseid)
1565 {
1566 PgStat_MsgInquiry msg;
1567
1568 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_INQUIRY);
1569 msg.clock_time = clock_time;
1570 msg.cutoff_time = cutoff_time;
1571 msg.databaseid = databaseid;
1572 pgstat_send(&msg, sizeof(msg));
1573 }
1574
1575
1576 /*
1577 * Initialize function call usage data.
1578 * Called by the executor before invoking a function.
1579 */
1580 void
pgstat_init_function_usage(FunctionCallInfoData * fcinfo,PgStat_FunctionCallUsage * fcu)1581 pgstat_init_function_usage(FunctionCallInfoData *fcinfo,
1582 PgStat_FunctionCallUsage *fcu)
1583 {
1584 PgStat_BackendFunctionEntry *htabent;
1585 bool found;
1586
1587 if (pgstat_track_functions <= fcinfo->flinfo->fn_stats)
1588 {
1589 /* stats not wanted */
1590 fcu->fs = NULL;
1591 return;
1592 }
1593
1594 if (!pgStatFunctions)
1595 {
1596 /* First time through - initialize function stat table */
1597 HASHCTL hash_ctl;
1598
1599 memset(&hash_ctl, 0, sizeof(hash_ctl));
1600 hash_ctl.keysize = sizeof(Oid);
1601 hash_ctl.entrysize = sizeof(PgStat_BackendFunctionEntry);
1602 pgStatFunctions = hash_create("Function stat entries",
1603 PGSTAT_FUNCTION_HASH_SIZE,
1604 &hash_ctl,
1605 HASH_ELEM | HASH_BLOBS);
1606 }
1607
1608 /* Get the stats entry for this function, create if necessary */
1609 htabent = hash_search(pgStatFunctions, &fcinfo->flinfo->fn_oid,
1610 HASH_ENTER, &found);
1611 if (!found)
1612 MemSet(&htabent->f_counts, 0, sizeof(PgStat_FunctionCounts));
1613
1614 fcu->fs = &htabent->f_counts;
1615
1616 /* save stats for this function, later used to compensate for recursion */
1617 fcu->save_f_total_time = htabent->f_counts.f_total_time;
1618
1619 /* save current backend-wide total time */
1620 fcu->save_total = total_func_time;
1621
1622 /* get clock time as of function start */
1623 INSTR_TIME_SET_CURRENT(fcu->f_start);
1624 }
1625
1626 /*
1627 * find_funcstat_entry - find any existing PgStat_BackendFunctionEntry entry
1628 * for specified function
1629 *
1630 * If no entry, return NULL, don't create a new one
1631 */
1632 PgStat_BackendFunctionEntry *
find_funcstat_entry(Oid func_id)1633 find_funcstat_entry(Oid func_id)
1634 {
1635 if (pgStatFunctions == NULL)
1636 return NULL;
1637
1638 return (PgStat_BackendFunctionEntry *) hash_search(pgStatFunctions,
1639 (void *) &func_id,
1640 HASH_FIND, NULL);
1641 }
1642
1643 /*
1644 * Calculate function call usage and update stat counters.
1645 * Called by the executor after invoking a function.
1646 *
1647 * In the case of a set-returning function that runs in value-per-call mode,
1648 * we will see multiple pgstat_init_function_usage/pgstat_end_function_usage
1649 * calls for what the user considers a single call of the function. The
1650 * finalize flag should be TRUE on the last call.
1651 */
1652 void
pgstat_end_function_usage(PgStat_FunctionCallUsage * fcu,bool finalize)1653 pgstat_end_function_usage(PgStat_FunctionCallUsage *fcu, bool finalize)
1654 {
1655 PgStat_FunctionCounts *fs = fcu->fs;
1656 instr_time f_total;
1657 instr_time f_others;
1658 instr_time f_self;
1659
1660 /* stats not wanted? */
1661 if (fs == NULL)
1662 return;
1663
1664 /* total elapsed time in this function call */
1665 INSTR_TIME_SET_CURRENT(f_total);
1666 INSTR_TIME_SUBTRACT(f_total, fcu->f_start);
1667
1668 /* self usage: elapsed minus anything already charged to other calls */
1669 f_others = total_func_time;
1670 INSTR_TIME_SUBTRACT(f_others, fcu->save_total);
1671 f_self = f_total;
1672 INSTR_TIME_SUBTRACT(f_self, f_others);
1673
1674 /* update backend-wide total time */
1675 INSTR_TIME_ADD(total_func_time, f_self);
1676
1677 /*
1678 * Compute the new f_total_time as the total elapsed time added to the
1679 * pre-call value of f_total_time. This is necessary to avoid
1680 * double-counting any time taken by recursive calls of myself. (We do
1681 * not need any similar kluge for self time, since that already excludes
1682 * any recursive calls.)
1683 */
1684 INSTR_TIME_ADD(f_total, fcu->save_f_total_time);
1685
1686 /* update counters in function stats table */
1687 if (finalize)
1688 fs->f_numcalls++;
1689 fs->f_total_time = f_total;
1690 INSTR_TIME_ADD(fs->f_self_time, f_self);
1691
1692 /* indicate that we have something to send */
1693 have_function_stats = true;
1694 }
1695
1696
1697 /* ----------
1698 * pgstat_initstats() -
1699 *
1700 * Initialize a relcache entry to count access statistics.
1701 * Called whenever a relation is opened.
1702 *
1703 * We assume that a relcache entry's pgstat_info field is zeroed by
1704 * relcache.c when the relcache entry is made; thereafter it is long-lived
1705 * data. We can avoid repeated searches of the TabStatus arrays when the
1706 * same relation is touched repeatedly within a transaction.
1707 * ----------
1708 */
1709 void
pgstat_initstats(Relation rel)1710 pgstat_initstats(Relation rel)
1711 {
1712 Oid rel_id = rel->rd_id;
1713 char relkind = rel->rd_rel->relkind;
1714
1715 /* We only count stats for things that have storage */
1716 if (!(relkind == RELKIND_RELATION ||
1717 relkind == RELKIND_MATVIEW ||
1718 relkind == RELKIND_INDEX ||
1719 relkind == RELKIND_TOASTVALUE ||
1720 relkind == RELKIND_SEQUENCE))
1721 {
1722 rel->pgstat_info = NULL;
1723 return;
1724 }
1725
1726 if (pgStatSock == PGINVALID_SOCKET || !pgstat_track_counts)
1727 {
1728 /* We're not counting at all */
1729 rel->pgstat_info = NULL;
1730 return;
1731 }
1732
1733 /*
1734 * If we already set up this relation in the current transaction, nothing
1735 * to do.
1736 */
1737 if (rel->pgstat_info != NULL &&
1738 rel->pgstat_info->t_id == rel_id)
1739 return;
1740
1741 /* Else find or make the PgStat_TableStatus entry, and update link */
1742 rel->pgstat_info = get_tabstat_entry(rel_id, rel->rd_rel->relisshared);
1743 }
1744
1745 /*
1746 * get_tabstat_entry - find or create a PgStat_TableStatus entry for rel
1747 */
1748 static PgStat_TableStatus *
get_tabstat_entry(Oid rel_id,bool isshared)1749 get_tabstat_entry(Oid rel_id, bool isshared)
1750 {
1751 TabStatHashEntry *hash_entry;
1752 PgStat_TableStatus *entry;
1753 TabStatusArray *tsa;
1754 bool found;
1755
1756 /*
1757 * Create hash table if we don't have it already.
1758 */
1759 if (pgStatTabHash == NULL)
1760 {
1761 HASHCTL ctl;
1762
1763 memset(&ctl, 0, sizeof(ctl));
1764 ctl.keysize = sizeof(Oid);
1765 ctl.entrysize = sizeof(TabStatHashEntry);
1766
1767 pgStatTabHash = hash_create("pgstat TabStatusArray lookup hash table",
1768 TABSTAT_QUANTUM,
1769 &ctl,
1770 HASH_ELEM | HASH_BLOBS);
1771 }
1772
1773 /*
1774 * Find an entry or create a new one.
1775 */
1776 hash_entry = hash_search(pgStatTabHash, &rel_id, HASH_ENTER, &found);
1777 if (!found)
1778 {
1779 /* initialize new entry with null pointer */
1780 hash_entry->tsa_entry = NULL;
1781 }
1782
1783 /*
1784 * If entry is already valid, we're done.
1785 */
1786 if (hash_entry->tsa_entry)
1787 return hash_entry->tsa_entry;
1788
1789 /*
1790 * Locate the first pgStatTabList entry with free space, making a new list
1791 * entry if needed. Note that we could get an OOM failure here, but if so
1792 * we have left the hashtable and the list in a consistent state.
1793 */
1794 if (pgStatTabList == NULL)
1795 {
1796 /* Set up first pgStatTabList entry */
1797 pgStatTabList = (TabStatusArray *)
1798 MemoryContextAllocZero(TopMemoryContext,
1799 sizeof(TabStatusArray));
1800 }
1801
1802 tsa = pgStatTabList;
1803 while (tsa->tsa_used >= TABSTAT_QUANTUM)
1804 {
1805 if (tsa->tsa_next == NULL)
1806 tsa->tsa_next = (TabStatusArray *)
1807 MemoryContextAllocZero(TopMemoryContext,
1808 sizeof(TabStatusArray));
1809 tsa = tsa->tsa_next;
1810 }
1811
1812 /*
1813 * Allocate a PgStat_TableStatus entry within this list entry. We assume
1814 * the entry was already zeroed, either at creation or after last use.
1815 */
1816 entry = &tsa->tsa_entries[tsa->tsa_used++];
1817 entry->t_id = rel_id;
1818 entry->t_shared = isshared;
1819
1820 /*
1821 * Now we can fill the entry in pgStatTabHash.
1822 */
1823 hash_entry->tsa_entry = entry;
1824
1825 return entry;
1826 }
1827
1828 /*
1829 * find_tabstat_entry - find any existing PgStat_TableStatus entry for rel
1830 *
1831 * If no entry, return NULL, don't create a new one
1832 *
1833 * Note: if we got an error in the most recent execution of pgstat_report_stat,
1834 * it's possible that an entry exists but there's no hashtable entry for it.
1835 * That's okay, we'll treat this case as "doesn't exist".
1836 */
1837 PgStat_TableStatus *
find_tabstat_entry(Oid rel_id)1838 find_tabstat_entry(Oid rel_id)
1839 {
1840 TabStatHashEntry *hash_entry;
1841
1842 /* If hashtable doesn't exist, there are no entries at all */
1843 if (!pgStatTabHash)
1844 return NULL;
1845
1846 hash_entry = hash_search(pgStatTabHash, &rel_id, HASH_FIND, NULL);
1847 if (!hash_entry)
1848 return NULL;
1849
1850 /* Note that this step could also return NULL, but that's correct */
1851 return hash_entry->tsa_entry;
1852 }
1853
1854 /*
1855 * get_tabstat_stack_level - add a new (sub)transaction stack entry if needed
1856 */
1857 static PgStat_SubXactStatus *
get_tabstat_stack_level(int nest_level)1858 get_tabstat_stack_level(int nest_level)
1859 {
1860 PgStat_SubXactStatus *xact_state;
1861
1862 xact_state = pgStatXactStack;
1863 if (xact_state == NULL || xact_state->nest_level != nest_level)
1864 {
1865 xact_state = (PgStat_SubXactStatus *)
1866 MemoryContextAlloc(TopTransactionContext,
1867 sizeof(PgStat_SubXactStatus));
1868 xact_state->nest_level = nest_level;
1869 xact_state->prev = pgStatXactStack;
1870 xact_state->first = NULL;
1871 pgStatXactStack = xact_state;
1872 }
1873 return xact_state;
1874 }
1875
1876 /*
1877 * add_tabstat_xact_level - add a new (sub)transaction state record
1878 */
1879 static void
add_tabstat_xact_level(PgStat_TableStatus * pgstat_info,int nest_level)1880 add_tabstat_xact_level(PgStat_TableStatus *pgstat_info, int nest_level)
1881 {
1882 PgStat_SubXactStatus *xact_state;
1883 PgStat_TableXactStatus *trans;
1884
1885 /*
1886 * If this is the first rel to be modified at the current nest level, we
1887 * first have to push a transaction stack entry.
1888 */
1889 xact_state = get_tabstat_stack_level(nest_level);
1890
1891 /* Now make a per-table stack entry */
1892 trans = (PgStat_TableXactStatus *)
1893 MemoryContextAllocZero(TopTransactionContext,
1894 sizeof(PgStat_TableXactStatus));
1895 trans->nest_level = nest_level;
1896 trans->upper = pgstat_info->trans;
1897 trans->parent = pgstat_info;
1898 trans->next = xact_state->first;
1899 xact_state->first = trans;
1900 pgstat_info->trans = trans;
1901 }
1902
1903 /*
1904 * pgstat_count_heap_insert - count a tuple insertion of n tuples
1905 */
1906 void
pgstat_count_heap_insert(Relation rel,PgStat_Counter n)1907 pgstat_count_heap_insert(Relation rel, PgStat_Counter n)
1908 {
1909 PgStat_TableStatus *pgstat_info = rel->pgstat_info;
1910
1911 if (pgstat_info != NULL)
1912 {
1913 /* We have to log the effect at the proper transactional level */
1914 int nest_level = GetCurrentTransactionNestLevel();
1915
1916 if (pgstat_info->trans == NULL ||
1917 pgstat_info->trans->nest_level != nest_level)
1918 add_tabstat_xact_level(pgstat_info, nest_level);
1919
1920 pgstat_info->trans->tuples_inserted += n;
1921 }
1922 }
1923
1924 /*
1925 * pgstat_count_heap_update - count a tuple update
1926 */
1927 void
pgstat_count_heap_update(Relation rel,bool hot)1928 pgstat_count_heap_update(Relation rel, bool hot)
1929 {
1930 PgStat_TableStatus *pgstat_info = rel->pgstat_info;
1931
1932 if (pgstat_info != NULL)
1933 {
1934 /* We have to log the effect at the proper transactional level */
1935 int nest_level = GetCurrentTransactionNestLevel();
1936
1937 if (pgstat_info->trans == NULL ||
1938 pgstat_info->trans->nest_level != nest_level)
1939 add_tabstat_xact_level(pgstat_info, nest_level);
1940
1941 pgstat_info->trans->tuples_updated++;
1942
1943 /* t_tuples_hot_updated is nontransactional, so just advance it */
1944 if (hot)
1945 pgstat_info->t_counts.t_tuples_hot_updated++;
1946 }
1947 }
1948
1949 /*
1950 * pgstat_count_heap_delete - count a tuple deletion
1951 */
1952 void
pgstat_count_heap_delete(Relation rel)1953 pgstat_count_heap_delete(Relation rel)
1954 {
1955 PgStat_TableStatus *pgstat_info = rel->pgstat_info;
1956
1957 if (pgstat_info != NULL)
1958 {
1959 /* We have to log the effect at the proper transactional level */
1960 int nest_level = GetCurrentTransactionNestLevel();
1961
1962 if (pgstat_info->trans == NULL ||
1963 pgstat_info->trans->nest_level != nest_level)
1964 add_tabstat_xact_level(pgstat_info, nest_level);
1965
1966 pgstat_info->trans->tuples_deleted++;
1967 }
1968 }
1969
1970 /*
1971 * pgstat_truncate_save_counters
1972 *
1973 * Whenever a table is truncated, we save its i/u/d counters so that they can
1974 * be cleared, and if the (sub)xact that executed the truncate later aborts,
1975 * the counters can be restored to the saved (pre-truncate) values. Note we do
1976 * this on the first truncate in any particular subxact level only.
1977 */
1978 static void
pgstat_truncate_save_counters(PgStat_TableXactStatus * trans)1979 pgstat_truncate_save_counters(PgStat_TableXactStatus *trans)
1980 {
1981 if (!trans->truncated)
1982 {
1983 trans->inserted_pre_trunc = trans->tuples_inserted;
1984 trans->updated_pre_trunc = trans->tuples_updated;
1985 trans->deleted_pre_trunc = trans->tuples_deleted;
1986 trans->truncated = true;
1987 }
1988 }
1989
1990 /*
1991 * pgstat_truncate_restore_counters - restore counters when a truncate aborts
1992 */
1993 static void
pgstat_truncate_restore_counters(PgStat_TableXactStatus * trans)1994 pgstat_truncate_restore_counters(PgStat_TableXactStatus *trans)
1995 {
1996 if (trans->truncated)
1997 {
1998 trans->tuples_inserted = trans->inserted_pre_trunc;
1999 trans->tuples_updated = trans->updated_pre_trunc;
2000 trans->tuples_deleted = trans->deleted_pre_trunc;
2001 }
2002 }
2003
2004 /*
2005 * pgstat_count_truncate - update tuple counters due to truncate
2006 */
2007 void
pgstat_count_truncate(Relation rel)2008 pgstat_count_truncate(Relation rel)
2009 {
2010 PgStat_TableStatus *pgstat_info = rel->pgstat_info;
2011
2012 if (pgstat_info != NULL)
2013 {
2014 /* We have to log the effect at the proper transactional level */
2015 int nest_level = GetCurrentTransactionNestLevel();
2016
2017 if (pgstat_info->trans == NULL ||
2018 pgstat_info->trans->nest_level != nest_level)
2019 add_tabstat_xact_level(pgstat_info, nest_level);
2020
2021 pgstat_truncate_save_counters(pgstat_info->trans);
2022 pgstat_info->trans->tuples_inserted = 0;
2023 pgstat_info->trans->tuples_updated = 0;
2024 pgstat_info->trans->tuples_deleted = 0;
2025 }
2026 }
2027
2028 /*
2029 * pgstat_update_heap_dead_tuples - update dead-tuples count
2030 *
2031 * The semantics of this are that we are reporting the nontransactional
2032 * recovery of "delta" dead tuples; so t_delta_dead_tuples decreases
2033 * rather than increasing, and the change goes straight into the per-table
2034 * counter, not into transactional state.
2035 */
2036 void
pgstat_update_heap_dead_tuples(Relation rel,int delta)2037 pgstat_update_heap_dead_tuples(Relation rel, int delta)
2038 {
2039 PgStat_TableStatus *pgstat_info = rel->pgstat_info;
2040
2041 if (pgstat_info != NULL)
2042 pgstat_info->t_counts.t_delta_dead_tuples -= delta;
2043 }
2044
2045
2046 /* ----------
2047 * AtEOXact_PgStat
2048 *
2049 * Called from access/transam/xact.c at top-level transaction commit/abort.
2050 * ----------
2051 */
2052 void
AtEOXact_PgStat(bool isCommit,bool parallel)2053 AtEOXact_PgStat(bool isCommit, bool parallel)
2054 {
2055 PgStat_SubXactStatus *xact_state;
2056
2057 /* Don't count parallel worker transaction stats */
2058 if (!parallel)
2059 {
2060 /*
2061 * Count transaction commit or abort. (We use counters, not just
2062 * bools, in case the reporting message isn't sent right away.)
2063 */
2064 if (isCommit)
2065 pgStatXactCommit++;
2066 else
2067 pgStatXactRollback++;
2068 }
2069
2070 /*
2071 * Transfer transactional insert/update counts into the base tabstat
2072 * entries. We don't bother to free any of the transactional state, since
2073 * it's all in TopTransactionContext and will go away anyway.
2074 */
2075 xact_state = pgStatXactStack;
2076 if (xact_state != NULL)
2077 {
2078 PgStat_TableXactStatus *trans;
2079
2080 Assert(xact_state->nest_level == 1);
2081 Assert(xact_state->prev == NULL);
2082 for (trans = xact_state->first; trans != NULL; trans = trans->next)
2083 {
2084 PgStat_TableStatus *tabstat;
2085
2086 Assert(trans->nest_level == 1);
2087 Assert(trans->upper == NULL);
2088 tabstat = trans->parent;
2089 Assert(tabstat->trans == trans);
2090 /* restore pre-truncate stats (if any) in case of aborted xact */
2091 if (!isCommit)
2092 pgstat_truncate_restore_counters(trans);
2093 /* count attempted actions regardless of commit/abort */
2094 tabstat->t_counts.t_tuples_inserted += trans->tuples_inserted;
2095 tabstat->t_counts.t_tuples_updated += trans->tuples_updated;
2096 tabstat->t_counts.t_tuples_deleted += trans->tuples_deleted;
2097 if (isCommit)
2098 {
2099 tabstat->t_counts.t_truncated = trans->truncated;
2100 if (trans->truncated)
2101 {
2102 /* forget live/dead stats seen by backend thus far */
2103 tabstat->t_counts.t_delta_live_tuples = 0;
2104 tabstat->t_counts.t_delta_dead_tuples = 0;
2105 }
2106 /* insert adds a live tuple, delete removes one */
2107 tabstat->t_counts.t_delta_live_tuples +=
2108 trans->tuples_inserted - trans->tuples_deleted;
2109 /* update and delete each create a dead tuple */
2110 tabstat->t_counts.t_delta_dead_tuples +=
2111 trans->tuples_updated + trans->tuples_deleted;
2112 /* insert, update, delete each count as one change event */
2113 tabstat->t_counts.t_changed_tuples +=
2114 trans->tuples_inserted + trans->tuples_updated +
2115 trans->tuples_deleted;
2116 }
2117 else
2118 {
2119 /* inserted tuples are dead, deleted tuples are unaffected */
2120 tabstat->t_counts.t_delta_dead_tuples +=
2121 trans->tuples_inserted + trans->tuples_updated;
2122 /* an aborted xact generates no changed_tuple events */
2123 }
2124 tabstat->trans = NULL;
2125 }
2126 }
2127 pgStatXactStack = NULL;
2128
2129 /* Make sure any stats snapshot is thrown away */
2130 pgstat_clear_snapshot();
2131 }
2132
2133 /* ----------
2134 * AtEOSubXact_PgStat
2135 *
2136 * Called from access/transam/xact.c at subtransaction commit/abort.
2137 * ----------
2138 */
2139 void
AtEOSubXact_PgStat(bool isCommit,int nestDepth)2140 AtEOSubXact_PgStat(bool isCommit, int nestDepth)
2141 {
2142 PgStat_SubXactStatus *xact_state;
2143
2144 /*
2145 * Transfer transactional insert/update counts into the next higher
2146 * subtransaction state.
2147 */
2148 xact_state = pgStatXactStack;
2149 if (xact_state != NULL &&
2150 xact_state->nest_level >= nestDepth)
2151 {
2152 PgStat_TableXactStatus *trans;
2153 PgStat_TableXactStatus *next_trans;
2154
2155 /* delink xact_state from stack immediately to simplify reuse case */
2156 pgStatXactStack = xact_state->prev;
2157
2158 for (trans = xact_state->first; trans != NULL; trans = next_trans)
2159 {
2160 PgStat_TableStatus *tabstat;
2161
2162 next_trans = trans->next;
2163 Assert(trans->nest_level == nestDepth);
2164 tabstat = trans->parent;
2165 Assert(tabstat->trans == trans);
2166 if (isCommit)
2167 {
2168 if (trans->upper && trans->upper->nest_level == nestDepth - 1)
2169 {
2170 if (trans->truncated)
2171 {
2172 /* propagate the truncate status one level up */
2173 pgstat_truncate_save_counters(trans->upper);
2174 /* replace upper xact stats with ours */
2175 trans->upper->tuples_inserted = trans->tuples_inserted;
2176 trans->upper->tuples_updated = trans->tuples_updated;
2177 trans->upper->tuples_deleted = trans->tuples_deleted;
2178 }
2179 else
2180 {
2181 trans->upper->tuples_inserted += trans->tuples_inserted;
2182 trans->upper->tuples_updated += trans->tuples_updated;
2183 trans->upper->tuples_deleted += trans->tuples_deleted;
2184 }
2185 tabstat->trans = trans->upper;
2186 pfree(trans);
2187 }
2188 else
2189 {
2190 /*
2191 * When there isn't an immediate parent state, we can just
2192 * reuse the record instead of going through a
2193 * palloc/pfree pushup (this works since it's all in
2194 * TopTransactionContext anyway). We have to re-link it
2195 * into the parent level, though, and that might mean
2196 * pushing a new entry into the pgStatXactStack.
2197 */
2198 PgStat_SubXactStatus *upper_xact_state;
2199
2200 upper_xact_state = get_tabstat_stack_level(nestDepth - 1);
2201 trans->next = upper_xact_state->first;
2202 upper_xact_state->first = trans;
2203 trans->nest_level = nestDepth - 1;
2204 }
2205 }
2206 else
2207 {
2208 /*
2209 * On abort, update top-level tabstat counts, then forget the
2210 * subtransaction
2211 */
2212
2213 /* first restore values obliterated by truncate */
2214 pgstat_truncate_restore_counters(trans);
2215 /* count attempted actions regardless of commit/abort */
2216 tabstat->t_counts.t_tuples_inserted += trans->tuples_inserted;
2217 tabstat->t_counts.t_tuples_updated += trans->tuples_updated;
2218 tabstat->t_counts.t_tuples_deleted += trans->tuples_deleted;
2219 /* inserted tuples are dead, deleted tuples are unaffected */
2220 tabstat->t_counts.t_delta_dead_tuples +=
2221 trans->tuples_inserted + trans->tuples_updated;
2222 tabstat->trans = trans->upper;
2223 pfree(trans);
2224 }
2225 }
2226 pfree(xact_state);
2227 }
2228 }
2229
2230
2231 /*
2232 * AtPrepare_PgStat
2233 * Save the transactional stats state at 2PC transaction prepare.
2234 *
2235 * In this phase we just generate 2PC records for all the pending
2236 * transaction-dependent stats work.
2237 */
2238 void
AtPrepare_PgStat(void)2239 AtPrepare_PgStat(void)
2240 {
2241 PgStat_SubXactStatus *xact_state;
2242
2243 xact_state = pgStatXactStack;
2244 if (xact_state != NULL)
2245 {
2246 PgStat_TableXactStatus *trans;
2247
2248 Assert(xact_state->nest_level == 1);
2249 Assert(xact_state->prev == NULL);
2250 for (trans = xact_state->first; trans != NULL; trans = trans->next)
2251 {
2252 PgStat_TableStatus *tabstat;
2253 TwoPhasePgStatRecord record;
2254
2255 Assert(trans->nest_level == 1);
2256 Assert(trans->upper == NULL);
2257 tabstat = trans->parent;
2258 Assert(tabstat->trans == trans);
2259
2260 record.tuples_inserted = trans->tuples_inserted;
2261 record.tuples_updated = trans->tuples_updated;
2262 record.tuples_deleted = trans->tuples_deleted;
2263 record.inserted_pre_trunc = trans->inserted_pre_trunc;
2264 record.updated_pre_trunc = trans->updated_pre_trunc;
2265 record.deleted_pre_trunc = trans->deleted_pre_trunc;
2266 record.t_id = tabstat->t_id;
2267 record.t_shared = tabstat->t_shared;
2268 record.t_truncated = trans->truncated;
2269
2270 RegisterTwoPhaseRecord(TWOPHASE_RM_PGSTAT_ID, 0,
2271 &record, sizeof(TwoPhasePgStatRecord));
2272 }
2273 }
2274 }
2275
2276 /*
2277 * PostPrepare_PgStat
2278 * Clean up after successful PREPARE.
2279 *
2280 * All we need do here is unlink the transaction stats state from the
2281 * nontransactional state. The nontransactional action counts will be
2282 * reported to the stats collector immediately, while the effects on live
2283 * and dead tuple counts are preserved in the 2PC state file.
2284 *
2285 * Note: AtEOXact_PgStat is not called during PREPARE.
2286 */
2287 void
PostPrepare_PgStat(void)2288 PostPrepare_PgStat(void)
2289 {
2290 PgStat_SubXactStatus *xact_state;
2291
2292 /*
2293 * We don't bother to free any of the transactional state, since it's all
2294 * in TopTransactionContext and will go away anyway.
2295 */
2296 xact_state = pgStatXactStack;
2297 if (xact_state != NULL)
2298 {
2299 PgStat_TableXactStatus *trans;
2300
2301 for (trans = xact_state->first; trans != NULL; trans = trans->next)
2302 {
2303 PgStat_TableStatus *tabstat;
2304
2305 tabstat = trans->parent;
2306 tabstat->trans = NULL;
2307 }
2308 }
2309 pgStatXactStack = NULL;
2310
2311 /* Make sure any stats snapshot is thrown away */
2312 pgstat_clear_snapshot();
2313 }
2314
2315 /*
2316 * 2PC processing routine for COMMIT PREPARED case.
2317 *
2318 * Load the saved counts into our local pgstats state.
2319 */
2320 void
pgstat_twophase_postcommit(TransactionId xid,uint16 info,void * recdata,uint32 len)2321 pgstat_twophase_postcommit(TransactionId xid, uint16 info,
2322 void *recdata, uint32 len)
2323 {
2324 TwoPhasePgStatRecord *rec = (TwoPhasePgStatRecord *) recdata;
2325 PgStat_TableStatus *pgstat_info;
2326
2327 /* Find or create a tabstat entry for the rel */
2328 pgstat_info = get_tabstat_entry(rec->t_id, rec->t_shared);
2329
2330 /* Same math as in AtEOXact_PgStat, commit case */
2331 pgstat_info->t_counts.t_tuples_inserted += rec->tuples_inserted;
2332 pgstat_info->t_counts.t_tuples_updated += rec->tuples_updated;
2333 pgstat_info->t_counts.t_tuples_deleted += rec->tuples_deleted;
2334 pgstat_info->t_counts.t_truncated = rec->t_truncated;
2335 if (rec->t_truncated)
2336 {
2337 /* forget live/dead stats seen by backend thus far */
2338 pgstat_info->t_counts.t_delta_live_tuples = 0;
2339 pgstat_info->t_counts.t_delta_dead_tuples = 0;
2340 }
2341 pgstat_info->t_counts.t_delta_live_tuples +=
2342 rec->tuples_inserted - rec->tuples_deleted;
2343 pgstat_info->t_counts.t_delta_dead_tuples +=
2344 rec->tuples_updated + rec->tuples_deleted;
2345 pgstat_info->t_counts.t_changed_tuples +=
2346 rec->tuples_inserted + rec->tuples_updated +
2347 rec->tuples_deleted;
2348 }
2349
2350 /*
2351 * 2PC processing routine for ROLLBACK PREPARED case.
2352 *
2353 * Load the saved counts into our local pgstats state, but treat them
2354 * as aborted.
2355 */
2356 void
pgstat_twophase_postabort(TransactionId xid,uint16 info,void * recdata,uint32 len)2357 pgstat_twophase_postabort(TransactionId xid, uint16 info,
2358 void *recdata, uint32 len)
2359 {
2360 TwoPhasePgStatRecord *rec = (TwoPhasePgStatRecord *) recdata;
2361 PgStat_TableStatus *pgstat_info;
2362
2363 /* Find or create a tabstat entry for the rel */
2364 pgstat_info = get_tabstat_entry(rec->t_id, rec->t_shared);
2365
2366 /* Same math as in AtEOXact_PgStat, abort case */
2367 if (rec->t_truncated)
2368 {
2369 rec->tuples_inserted = rec->inserted_pre_trunc;
2370 rec->tuples_updated = rec->updated_pre_trunc;
2371 rec->tuples_deleted = rec->deleted_pre_trunc;
2372 }
2373 pgstat_info->t_counts.t_tuples_inserted += rec->tuples_inserted;
2374 pgstat_info->t_counts.t_tuples_updated += rec->tuples_updated;
2375 pgstat_info->t_counts.t_tuples_deleted += rec->tuples_deleted;
2376 pgstat_info->t_counts.t_delta_dead_tuples +=
2377 rec->tuples_inserted + rec->tuples_updated;
2378 }
2379
2380
2381 /* ----------
2382 * pgstat_fetch_stat_dbentry() -
2383 *
2384 * Support function for the SQL-callable pgstat* functions. Returns
2385 * the collected statistics for one database or NULL. NULL doesn't mean
2386 * that the database doesn't exist, it is just not yet known by the
2387 * collector, so the caller is better off to report ZERO instead.
2388 * ----------
2389 */
2390 PgStat_StatDBEntry *
pgstat_fetch_stat_dbentry(Oid dbid)2391 pgstat_fetch_stat_dbentry(Oid dbid)
2392 {
2393 /*
2394 * If not done for this transaction, read the statistics collector stats
2395 * file into some hash tables.
2396 */
2397 backend_read_statsfile();
2398
2399 /*
2400 * Lookup the requested database; return NULL if not found
2401 */
2402 return (PgStat_StatDBEntry *) hash_search(pgStatDBHash,
2403 (void *) &dbid,
2404 HASH_FIND, NULL);
2405 }
2406
2407
2408 /* ----------
2409 * pgstat_fetch_stat_tabentry() -
2410 *
2411 * Support function for the SQL-callable pgstat* functions. Returns
2412 * the collected statistics for one table or NULL. NULL doesn't mean
2413 * that the table doesn't exist, it is just not yet known by the
2414 * collector, so the caller is better off to report ZERO instead.
2415 * ----------
2416 */
2417 PgStat_StatTabEntry *
pgstat_fetch_stat_tabentry(Oid relid)2418 pgstat_fetch_stat_tabentry(Oid relid)
2419 {
2420 Oid dbid;
2421 PgStat_StatDBEntry *dbentry;
2422 PgStat_StatTabEntry *tabentry;
2423
2424 /*
2425 * If not done for this transaction, read the statistics collector stats
2426 * file into some hash tables.
2427 */
2428 backend_read_statsfile();
2429
2430 /*
2431 * Lookup our database, then look in its table hash table.
2432 */
2433 dbid = MyDatabaseId;
2434 dbentry = (PgStat_StatDBEntry *) hash_search(pgStatDBHash,
2435 (void *) &dbid,
2436 HASH_FIND, NULL);
2437 if (dbentry != NULL && dbentry->tables != NULL)
2438 {
2439 tabentry = (PgStat_StatTabEntry *) hash_search(dbentry->tables,
2440 (void *) &relid,
2441 HASH_FIND, NULL);
2442 if (tabentry)
2443 return tabentry;
2444 }
2445
2446 /*
2447 * If we didn't find it, maybe it's a shared table.
2448 */
2449 dbid = InvalidOid;
2450 dbentry = (PgStat_StatDBEntry *) hash_search(pgStatDBHash,
2451 (void *) &dbid,
2452 HASH_FIND, NULL);
2453 if (dbentry != NULL && dbentry->tables != NULL)
2454 {
2455 tabentry = (PgStat_StatTabEntry *) hash_search(dbentry->tables,
2456 (void *) &relid,
2457 HASH_FIND, NULL);
2458 if (tabentry)
2459 return tabentry;
2460 }
2461
2462 return NULL;
2463 }
2464
2465
2466 /* ----------
2467 * pgstat_fetch_stat_funcentry() -
2468 *
2469 * Support function for the SQL-callable pgstat* functions. Returns
2470 * the collected statistics for one function or NULL.
2471 * ----------
2472 */
2473 PgStat_StatFuncEntry *
pgstat_fetch_stat_funcentry(Oid func_id)2474 pgstat_fetch_stat_funcentry(Oid func_id)
2475 {
2476 PgStat_StatDBEntry *dbentry;
2477 PgStat_StatFuncEntry *funcentry = NULL;
2478
2479 /* load the stats file if needed */
2480 backend_read_statsfile();
2481
2482 /* Lookup our database, then find the requested function. */
2483 dbentry = pgstat_fetch_stat_dbentry(MyDatabaseId);
2484 if (dbentry != NULL && dbentry->functions != NULL)
2485 {
2486 funcentry = (PgStat_StatFuncEntry *) hash_search(dbentry->functions,
2487 (void *) &func_id,
2488 HASH_FIND, NULL);
2489 }
2490
2491 return funcentry;
2492 }
2493
2494
2495 /* ----------
2496 * pgstat_fetch_stat_beentry() -
2497 *
2498 * Support function for the SQL-callable pgstat* functions. Returns
2499 * our local copy of the current-activity entry for one backend.
2500 *
2501 * NB: caller is responsible for a check if the user is permitted to see
2502 * this info (especially the querystring).
2503 * ----------
2504 */
2505 PgBackendStatus *
pgstat_fetch_stat_beentry(int beid)2506 pgstat_fetch_stat_beentry(int beid)
2507 {
2508 pgstat_read_current_status();
2509
2510 if (beid < 1 || beid > localNumBackends)
2511 return NULL;
2512
2513 return &localBackendStatusTable[beid - 1].backendStatus;
2514 }
2515
2516
2517 /* ----------
2518 * pgstat_fetch_stat_local_beentry() -
2519 *
2520 * Like pgstat_fetch_stat_beentry() but with locally computed additions (like
2521 * xid and xmin values of the backend)
2522 *
2523 * NB: caller is responsible for a check if the user is permitted to see
2524 * this info (especially the querystring).
2525 * ----------
2526 */
2527 LocalPgBackendStatus *
pgstat_fetch_stat_local_beentry(int beid)2528 pgstat_fetch_stat_local_beentry(int beid)
2529 {
2530 pgstat_read_current_status();
2531
2532 if (beid < 1 || beid > localNumBackends)
2533 return NULL;
2534
2535 return &localBackendStatusTable[beid - 1];
2536 }
2537
2538
2539 /* ----------
2540 * pgstat_fetch_stat_numbackends() -
2541 *
2542 * Support function for the SQL-callable pgstat* functions. Returns
2543 * the maximum current backend id.
2544 * ----------
2545 */
2546 int
pgstat_fetch_stat_numbackends(void)2547 pgstat_fetch_stat_numbackends(void)
2548 {
2549 pgstat_read_current_status();
2550
2551 return localNumBackends;
2552 }
2553
2554 /*
2555 * ---------
2556 * pgstat_fetch_stat_archiver() -
2557 *
2558 * Support function for the SQL-callable pgstat* functions. Returns
2559 * a pointer to the archiver statistics struct.
2560 * ---------
2561 */
2562 PgStat_ArchiverStats *
pgstat_fetch_stat_archiver(void)2563 pgstat_fetch_stat_archiver(void)
2564 {
2565 backend_read_statsfile();
2566
2567 return &archiverStats;
2568 }
2569
2570
2571 /*
2572 * ---------
2573 * pgstat_fetch_global() -
2574 *
2575 * Support function for the SQL-callable pgstat* functions. Returns
2576 * a pointer to the global statistics struct.
2577 * ---------
2578 */
2579 PgStat_GlobalStats *
pgstat_fetch_global(void)2580 pgstat_fetch_global(void)
2581 {
2582 backend_read_statsfile();
2583
2584 return &globalStats;
2585 }
2586
2587
2588 /* ------------------------------------------------------------
2589 * Functions for management of the shared-memory PgBackendStatus array
2590 * ------------------------------------------------------------
2591 */
2592
2593 static PgBackendStatus *BackendStatusArray = NULL;
2594 static PgBackendStatus *MyBEEntry = NULL;
2595 static char *BackendAppnameBuffer = NULL;
2596 static char *BackendClientHostnameBuffer = NULL;
2597 static char *BackendActivityBuffer = NULL;
2598 static Size BackendActivityBufferSize = 0;
2599 #ifdef USE_SSL
2600 static PgBackendSSLStatus *BackendSslStatusBuffer = NULL;
2601 #endif
2602
2603
2604 /*
2605 * Report shared-memory space needed by CreateSharedBackendStatus.
2606 */
2607 Size
BackendStatusShmemSize(void)2608 BackendStatusShmemSize(void)
2609 {
2610 Size size;
2611
2612 /* BackendStatusArray: */
2613 size = mul_size(sizeof(PgBackendStatus), NumBackendStatSlots);
2614 /* BackendAppnameBuffer: */
2615 size = add_size(size,
2616 mul_size(NAMEDATALEN, NumBackendStatSlots));
2617 /* BackendClientHostnameBuffer: */
2618 size = add_size(size,
2619 mul_size(NAMEDATALEN, NumBackendStatSlots));
2620 /* BackendActivityBuffer: */
2621 size = add_size(size,
2622 mul_size(pgstat_track_activity_query_size, NumBackendStatSlots));
2623 #ifdef USE_SSL
2624 /* BackendSslStatusBuffer: */
2625 size = add_size(size,
2626 mul_size(sizeof(PgBackendSSLStatus), NumBackendStatSlots));
2627 #endif
2628 return size;
2629 }
2630
2631 /*
2632 * Initialize the shared status array and several string buffers
2633 * during postmaster startup.
2634 */
2635 void
CreateSharedBackendStatus(void)2636 CreateSharedBackendStatus(void)
2637 {
2638 Size size;
2639 bool found;
2640 int i;
2641 char *buffer;
2642
2643 /* Create or attach to the shared array */
2644 size = mul_size(sizeof(PgBackendStatus), NumBackendStatSlots);
2645 BackendStatusArray = (PgBackendStatus *)
2646 ShmemInitStruct("Backend Status Array", size, &found);
2647
2648 if (!found)
2649 {
2650 /*
2651 * We're the first - initialize.
2652 */
2653 MemSet(BackendStatusArray, 0, size);
2654 }
2655
2656 /* Create or attach to the shared appname buffer */
2657 size = mul_size(NAMEDATALEN, NumBackendStatSlots);
2658 BackendAppnameBuffer = (char *)
2659 ShmemInitStruct("Backend Application Name Buffer", size, &found);
2660
2661 if (!found)
2662 {
2663 MemSet(BackendAppnameBuffer, 0, size);
2664
2665 /* Initialize st_appname pointers. */
2666 buffer = BackendAppnameBuffer;
2667 for (i = 0; i < NumBackendStatSlots; i++)
2668 {
2669 BackendStatusArray[i].st_appname = buffer;
2670 buffer += NAMEDATALEN;
2671 }
2672 }
2673
2674 /* Create or attach to the shared client hostname buffer */
2675 size = mul_size(NAMEDATALEN, NumBackendStatSlots);
2676 BackendClientHostnameBuffer = (char *)
2677 ShmemInitStruct("Backend Client Host Name Buffer", size, &found);
2678
2679 if (!found)
2680 {
2681 MemSet(BackendClientHostnameBuffer, 0, size);
2682
2683 /* Initialize st_clienthostname pointers. */
2684 buffer = BackendClientHostnameBuffer;
2685 for (i = 0; i < NumBackendStatSlots; i++)
2686 {
2687 BackendStatusArray[i].st_clienthostname = buffer;
2688 buffer += NAMEDATALEN;
2689 }
2690 }
2691
2692 /* Create or attach to the shared activity buffer */
2693 BackendActivityBufferSize = mul_size(pgstat_track_activity_query_size,
2694 NumBackendStatSlots);
2695 BackendActivityBuffer = (char *)
2696 ShmemInitStruct("Backend Activity Buffer",
2697 BackendActivityBufferSize,
2698 &found);
2699
2700 if (!found)
2701 {
2702 MemSet(BackendActivityBuffer, 0, BackendActivityBufferSize);
2703
2704 /* Initialize st_activity pointers. */
2705 buffer = BackendActivityBuffer;
2706 for (i = 0; i < NumBackendStatSlots; i++)
2707 {
2708 BackendStatusArray[i].st_activity = buffer;
2709 buffer += pgstat_track_activity_query_size;
2710 }
2711 }
2712
2713 #ifdef USE_SSL
2714 /* Create or attach to the shared SSL status buffer */
2715 size = mul_size(sizeof(PgBackendSSLStatus), NumBackendStatSlots);
2716 BackendSslStatusBuffer = (PgBackendSSLStatus *)
2717 ShmemInitStruct("Backend SSL Status Buffer", size, &found);
2718
2719 if (!found)
2720 {
2721 PgBackendSSLStatus *ptr;
2722
2723 MemSet(BackendSslStatusBuffer, 0, size);
2724
2725 /* Initialize st_sslstatus pointers. */
2726 ptr = BackendSslStatusBuffer;
2727 for (i = 0; i < NumBackendStatSlots; i++)
2728 {
2729 BackendStatusArray[i].st_sslstatus = ptr;
2730 ptr++;
2731 }
2732 }
2733 #endif
2734 }
2735
2736
2737 /* ----------
2738 * pgstat_initialize() -
2739 *
2740 * Initialize pgstats state, and set up our on-proc-exit hook.
2741 * Called from InitPostgres and AuxiliaryProcessMain. For auxiliary process,
2742 * MyBackendId is invalid. Otherwise, MyBackendId must be set,
2743 * but we must not have started any transaction yet (since the
2744 * exit hook must run after the last transaction exit).
2745 * NOTE: MyDatabaseId isn't set yet; so the shutdown hook has to be careful.
2746 * ----------
2747 */
2748 void
pgstat_initialize(void)2749 pgstat_initialize(void)
2750 {
2751 /* Initialize MyBEEntry */
2752 if (MyBackendId != InvalidBackendId)
2753 {
2754 Assert(MyBackendId >= 1 && MyBackendId <= MaxBackends);
2755 MyBEEntry = &BackendStatusArray[MyBackendId - 1];
2756 }
2757 else
2758 {
2759 /* Must be an auxiliary process */
2760 Assert(MyAuxProcType != NotAnAuxProcess);
2761
2762 /*
2763 * Assign the MyBEEntry for an auxiliary process. Since it doesn't
2764 * have a BackendId, the slot is statically allocated based on the
2765 * auxiliary process type (MyAuxProcType). Backends use slots indexed
2766 * in the range from 1 to MaxBackends (inclusive), so we use
2767 * MaxBackends + AuxBackendType + 1 as the index of the slot for an
2768 * auxiliary process.
2769 */
2770 MyBEEntry = &BackendStatusArray[MaxBackends + MyAuxProcType];
2771 }
2772
2773 /* Set up a process-exit hook to clean up */
2774 on_shmem_exit(pgstat_beshutdown_hook, 0);
2775 }
2776
2777 /* ----------
2778 * pgstat_bestart() -
2779 *
2780 * Initialize this backend's entry in the PgBackendStatus array.
2781 * Called from InitPostgres.
2782 *
2783 * Apart from auxiliary processes, MyBackendId, MyDatabaseId,
2784 * session userid, and application_name must be set for a
2785 * backend (hence, this cannot be combined with pgstat_initialize).
2786 * Note also that we must be inside a transaction if this isn't an aux
2787 * process, as we may need to do encoding conversion on some strings.
2788 * ----------
2789 */
2790 void
pgstat_bestart(void)2791 pgstat_bestart(void)
2792 {
2793 volatile PgBackendStatus *vbeentry = MyBEEntry;
2794 PgBackendStatus lbeentry;
2795 #ifdef USE_SSL
2796 PgBackendSSLStatus lsslstatus;
2797 #endif
2798
2799 /* pgstats state must be initialized from pgstat_initialize() */
2800 Assert(vbeentry != NULL);
2801
2802 /*
2803 * To minimize the time spent modifying the PgBackendStatus entry, and
2804 * avoid risk of errors inside the critical section, we first copy the
2805 * shared-memory struct to a local variable, then modify the data in the
2806 * local variable, then copy the local variable back to shared memory.
2807 * Only the last step has to be inside the critical section.
2808 *
2809 * Most of the data we copy from shared memory is just going to be
2810 * overwritten, but the struct's not so large that it's worth the
2811 * maintenance hassle to copy only the needful fields.
2812 */
2813 memcpy(&lbeentry,
2814 (char *) vbeentry,
2815 sizeof(PgBackendStatus));
2816
2817 /* This struct can just start from zeroes each time, though */
2818 #ifdef USE_SSL
2819 memset(&lsslstatus, 0, sizeof(lsslstatus));
2820 #endif
2821
2822 /*
2823 * Now fill in all the fields of lbeentry, except for strings that are
2824 * out-of-line data. Those have to be handled separately, below.
2825 */
2826 lbeentry.st_procpid = MyProcPid;
2827
2828 if (MyBackendId != InvalidBackendId)
2829 {
2830 if (IsAutoVacuumLauncherProcess())
2831 {
2832 /* Autovacuum Launcher */
2833 lbeentry.st_backendType = B_AUTOVAC_LAUNCHER;
2834 }
2835 else if (IsAutoVacuumWorkerProcess())
2836 {
2837 /* Autovacuum Worker */
2838 lbeentry.st_backendType = B_AUTOVAC_WORKER;
2839 }
2840 else if (am_walsender)
2841 {
2842 /* Wal sender */
2843 lbeentry.st_backendType = B_WAL_SENDER;
2844 }
2845 else if (IsBackgroundWorker)
2846 {
2847 /* bgworker */
2848 lbeentry.st_backendType = B_BG_WORKER;
2849 }
2850 else
2851 {
2852 /* client-backend */
2853 lbeentry.st_backendType = B_BACKEND;
2854 }
2855 }
2856 else
2857 {
2858 /* Must be an auxiliary process */
2859 Assert(MyAuxProcType != NotAnAuxProcess);
2860 switch (MyAuxProcType)
2861 {
2862 case StartupProcess:
2863 lbeentry.st_backendType = B_STARTUP;
2864 break;
2865 case BgWriterProcess:
2866 lbeentry.st_backendType = B_BG_WRITER;
2867 break;
2868 case CheckpointerProcess:
2869 lbeentry.st_backendType = B_CHECKPOINTER;
2870 break;
2871 case WalWriterProcess:
2872 lbeentry.st_backendType = B_WAL_WRITER;
2873 break;
2874 case WalReceiverProcess:
2875 lbeentry.st_backendType = B_WAL_RECEIVER;
2876 break;
2877 default:
2878 elog(FATAL, "unrecognized process type: %d",
2879 (int) MyAuxProcType);
2880 }
2881 }
2882
2883 /*
2884 * If we have a MyProcPort, use its session start time (for consistency,
2885 * and to save a kernel call).
2886 */
2887 if (MyProcPort)
2888 lbeentry.st_proc_start_timestamp = MyProcPort->SessionStartTime;
2889 else
2890 lbeentry.st_proc_start_timestamp = GetCurrentTimestamp();
2891
2892 lbeentry.st_activity_start_timestamp = 0;
2893 lbeentry.st_state_start_timestamp = 0;
2894 lbeentry.st_xact_start_timestamp = 0;
2895 lbeentry.st_databaseid = MyDatabaseId;
2896
2897 /* We have userid for client-backends, wal-sender and bgworker processes */
2898 if (lbeentry.st_backendType == B_BACKEND
2899 || lbeentry.st_backendType == B_WAL_SENDER
2900 || lbeentry.st_backendType == B_BG_WORKER)
2901 lbeentry.st_userid = GetSessionUserId();
2902 else
2903 lbeentry.st_userid = InvalidOid;
2904
2905 /*
2906 * We may not have a MyProcPort (eg, if this is the autovacuum process).
2907 * If so, use all-zeroes client address, which is dealt with specially in
2908 * pg_stat_get_backend_client_addr and pg_stat_get_backend_client_port.
2909 */
2910 if (MyProcPort)
2911 memcpy(&lbeentry.st_clientaddr, &MyProcPort->raddr,
2912 sizeof(lbeentry.st_clientaddr));
2913 else
2914 MemSet(&lbeentry.st_clientaddr, 0, sizeof(lbeentry.st_clientaddr));
2915
2916 #ifdef USE_SSL
2917 if (MyProcPort && MyProcPort->ssl != NULL)
2918 {
2919 lbeentry.st_ssl = true;
2920 lsslstatus.ssl_bits = be_tls_get_cipher_bits(MyProcPort);
2921 lsslstatus.ssl_compression = be_tls_get_compression(MyProcPort);
2922 be_tls_get_version(MyProcPort, lsslstatus.ssl_version, NAMEDATALEN);
2923 be_tls_get_cipher(MyProcPort, lsslstatus.ssl_cipher, NAMEDATALEN);
2924 be_tls_get_peerdn_name(MyProcPort, lsslstatus.ssl_clientdn, NAMEDATALEN);
2925 }
2926 else
2927 {
2928 lbeentry.st_ssl = false;
2929 }
2930 #else
2931 lbeentry.st_ssl = false;
2932 #endif
2933
2934 lbeentry.st_state = STATE_UNDEFINED;
2935 lbeentry.st_progress_command = PROGRESS_COMMAND_INVALID;
2936 lbeentry.st_progress_command_target = InvalidOid;
2937
2938 /*
2939 * we don't zero st_progress_param here to save cycles; nobody should
2940 * examine it until st_progress_command has been set to something other
2941 * than PROGRESS_COMMAND_INVALID
2942 */
2943
2944 /*
2945 * We're ready to enter the critical section that fills the shared-memory
2946 * status entry. We follow the protocol of bumping st_changecount before
2947 * and after; and make sure it's even afterwards. We use a volatile
2948 * pointer here to ensure the compiler doesn't try to get cute.
2949 */
2950 PGSTAT_BEGIN_WRITE_ACTIVITY(vbeentry);
2951
2952 /* make sure we'll memcpy the same st_changecount back */
2953 lbeentry.st_changecount = vbeentry->st_changecount;
2954
2955 memcpy((char *) vbeentry,
2956 &lbeentry,
2957 sizeof(PgBackendStatus));
2958
2959 /*
2960 * We can write the out-of-line strings and structs using the pointers
2961 * that are in lbeentry; this saves some de-volatilizing messiness.
2962 */
2963 lbeentry.st_appname[0] = '\0';
2964 if (MyProcPort && MyProcPort->remote_hostname)
2965 strlcpy(lbeentry.st_clienthostname, MyProcPort->remote_hostname,
2966 NAMEDATALEN);
2967 else
2968 lbeentry.st_clienthostname[0] = '\0';
2969 lbeentry.st_activity[0] = '\0';
2970 /* Also make sure the last byte in each string area is always 0 */
2971 lbeentry.st_appname[NAMEDATALEN - 1] = '\0';
2972 lbeentry.st_clienthostname[NAMEDATALEN - 1] = '\0';
2973 lbeentry.st_activity[pgstat_track_activity_query_size - 1] = '\0';
2974
2975 #ifdef USE_SSL
2976 memcpy(lbeentry.st_sslstatus, &lsslstatus, sizeof(PgBackendSSLStatus));
2977 #endif
2978
2979 PGSTAT_END_WRITE_ACTIVITY(vbeentry);
2980
2981 /* Update app name to current GUC setting */
2982 if (application_name)
2983 pgstat_report_appname(application_name);
2984 }
2985
2986 /*
2987 * Shut down a single backend's statistics reporting at process exit.
2988 *
2989 * Flush any remaining statistics counts out to the collector.
2990 * Without this, operations triggered during backend exit (such as
2991 * temp table deletions) won't be counted.
2992 *
2993 * Lastly, clear out our entry in the PgBackendStatus array.
2994 */
2995 static void
pgstat_beshutdown_hook(int code,Datum arg)2996 pgstat_beshutdown_hook(int code, Datum arg)
2997 {
2998 volatile PgBackendStatus *beentry = MyBEEntry;
2999
3000 /*
3001 * If we got as far as discovering our own database ID, we can report what
3002 * we did to the collector. Otherwise, we'd be sending an invalid
3003 * database ID, so forget it. (This means that accesses to pg_database
3004 * during failed backend starts might never get counted.)
3005 */
3006 if (OidIsValid(MyDatabaseId))
3007 pgstat_report_stat(true);
3008
3009 /*
3010 * Clear my status entry, following the protocol of bumping st_changecount
3011 * before and after. We use a volatile pointer here to ensure the
3012 * compiler doesn't try to get cute.
3013 */
3014 PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
3015
3016 beentry->st_procpid = 0; /* mark invalid */
3017
3018 PGSTAT_END_WRITE_ACTIVITY(beentry);
3019 }
3020
3021
3022 /* ----------
3023 * pgstat_report_activity() -
3024 *
3025 * Called from tcop/postgres.c to report what the backend is actually doing
3026 * (but note cmd_str can be NULL for certain cases).
3027 *
3028 * All updates of the status entry follow the protocol of bumping
3029 * st_changecount before and after. We use a volatile pointer here to
3030 * ensure the compiler doesn't try to get cute.
3031 * ----------
3032 */
3033 void
pgstat_report_activity(BackendState state,const char * cmd_str)3034 pgstat_report_activity(BackendState state, const char *cmd_str)
3035 {
3036 volatile PgBackendStatus *beentry = MyBEEntry;
3037 TimestampTz start_timestamp;
3038 TimestampTz current_timestamp;
3039 int len = 0;
3040
3041 TRACE_POSTGRESQL_STATEMENT_STATUS(cmd_str);
3042
3043 if (!beentry)
3044 return;
3045
3046 if (!pgstat_track_activities)
3047 {
3048 if (beentry->st_state != STATE_DISABLED)
3049 {
3050 volatile PGPROC *proc = MyProc;
3051
3052 /*
3053 * track_activities is disabled, but we last reported a
3054 * non-disabled state. As our final update, change the state and
3055 * clear fields we will not be updating anymore.
3056 */
3057 PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
3058 beentry->st_state = STATE_DISABLED;
3059 beentry->st_state_start_timestamp = 0;
3060 beentry->st_activity[0] = '\0';
3061 beentry->st_activity_start_timestamp = 0;
3062 /* st_xact_start_timestamp and wait_event_info are also disabled */
3063 beentry->st_xact_start_timestamp = 0;
3064 proc->wait_event_info = 0;
3065 PGSTAT_END_WRITE_ACTIVITY(beentry);
3066 }
3067 return;
3068 }
3069
3070 /*
3071 * To minimize the time spent modifying the entry, and avoid risk of
3072 * errors inside the critical section, fetch all the needed data first.
3073 */
3074 start_timestamp = GetCurrentStatementStartTimestamp();
3075 if (cmd_str != NULL)
3076 {
3077 len = pg_mbcliplen(cmd_str, strlen(cmd_str),
3078 pgstat_track_activity_query_size - 1);
3079 }
3080 current_timestamp = GetCurrentTimestamp();
3081
3082 /*
3083 * Now update the status entry
3084 */
3085 PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
3086
3087 beentry->st_state = state;
3088 beentry->st_state_start_timestamp = current_timestamp;
3089
3090 if (cmd_str != NULL)
3091 {
3092 memcpy((char *) beentry->st_activity, cmd_str, len);
3093 beentry->st_activity[len] = '\0';
3094 beentry->st_activity_start_timestamp = start_timestamp;
3095 }
3096
3097 PGSTAT_END_WRITE_ACTIVITY(beentry);
3098 }
3099
3100 /*-----------
3101 * pgstat_progress_start_command() -
3102 *
3103 * Set st_progress_command (and st_progress_command_target) in own backend
3104 * entry. Also, zero-initialize st_progress_param array.
3105 *-----------
3106 */
3107 void
pgstat_progress_start_command(ProgressCommandType cmdtype,Oid relid)3108 pgstat_progress_start_command(ProgressCommandType cmdtype, Oid relid)
3109 {
3110 volatile PgBackendStatus *beentry = MyBEEntry;
3111
3112 if (!beentry || !pgstat_track_activities)
3113 return;
3114
3115 PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
3116 beentry->st_progress_command = cmdtype;
3117 beentry->st_progress_command_target = relid;
3118 MemSet(&beentry->st_progress_param, 0, sizeof(beentry->st_progress_param));
3119 PGSTAT_END_WRITE_ACTIVITY(beentry);
3120 }
3121
3122 /*-----------
3123 * pgstat_progress_update_param() -
3124 *
3125 * Update index'th member in st_progress_param[] of own backend entry.
3126 *-----------
3127 */
3128 void
pgstat_progress_update_param(int index,int64 val)3129 pgstat_progress_update_param(int index, int64 val)
3130 {
3131 volatile PgBackendStatus *beentry = MyBEEntry;
3132
3133 Assert(index >= 0 && index < PGSTAT_NUM_PROGRESS_PARAM);
3134
3135 if (!beentry || !pgstat_track_activities)
3136 return;
3137
3138 PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
3139 beentry->st_progress_param[index] = val;
3140 PGSTAT_END_WRITE_ACTIVITY(beentry);
3141 }
3142
3143 /*-----------
3144 * pgstat_progress_update_multi_param() -
3145 *
3146 * Update multiple members in st_progress_param[] of own backend entry.
3147 * This is atomic; readers won't see intermediate states.
3148 *-----------
3149 */
3150 void
pgstat_progress_update_multi_param(int nparam,const int * index,const int64 * val)3151 pgstat_progress_update_multi_param(int nparam, const int *index,
3152 const int64 *val)
3153 {
3154 volatile PgBackendStatus *beentry = MyBEEntry;
3155 int i;
3156
3157 if (!beentry || !pgstat_track_activities || nparam == 0)
3158 return;
3159
3160 PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
3161
3162 for (i = 0; i < nparam; ++i)
3163 {
3164 Assert(index[i] >= 0 && index[i] < PGSTAT_NUM_PROGRESS_PARAM);
3165
3166 beentry->st_progress_param[index[i]] = val[i];
3167 }
3168
3169 PGSTAT_END_WRITE_ACTIVITY(beentry);
3170 }
3171
3172 /*-----------
3173 * pgstat_progress_end_command() -
3174 *
3175 * Reset st_progress_command (and st_progress_command_target) in own backend
3176 * entry. This signals the end of the command.
3177 *-----------
3178 */
3179 void
pgstat_progress_end_command(void)3180 pgstat_progress_end_command(void)
3181 {
3182 volatile PgBackendStatus *beentry = MyBEEntry;
3183
3184 if (!beentry || !pgstat_track_activities)
3185 return;
3186
3187 if (beentry->st_progress_command == PROGRESS_COMMAND_INVALID)
3188 return;
3189
3190 PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
3191 beentry->st_progress_command = PROGRESS_COMMAND_INVALID;
3192 beentry->st_progress_command_target = InvalidOid;
3193 PGSTAT_END_WRITE_ACTIVITY(beentry);
3194 }
3195
3196 /* ----------
3197 * pgstat_report_appname() -
3198 *
3199 * Called to update our application name.
3200 * ----------
3201 */
3202 void
pgstat_report_appname(const char * appname)3203 pgstat_report_appname(const char *appname)
3204 {
3205 volatile PgBackendStatus *beentry = MyBEEntry;
3206 int len;
3207
3208 if (!beentry)
3209 return;
3210
3211 /* This should be unnecessary if GUC did its job, but be safe */
3212 len = pg_mbcliplen(appname, strlen(appname), NAMEDATALEN - 1);
3213
3214 /*
3215 * Update my status entry, following the protocol of bumping
3216 * st_changecount before and after. We use a volatile pointer here to
3217 * ensure the compiler doesn't try to get cute.
3218 */
3219 PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
3220
3221 memcpy((char *) beentry->st_appname, appname, len);
3222 beentry->st_appname[len] = '\0';
3223
3224 PGSTAT_END_WRITE_ACTIVITY(beentry);
3225 }
3226
3227 /*
3228 * Report current transaction start timestamp as the specified value.
3229 * Zero means there is no active transaction.
3230 */
3231 void
pgstat_report_xact_timestamp(TimestampTz tstamp)3232 pgstat_report_xact_timestamp(TimestampTz tstamp)
3233 {
3234 volatile PgBackendStatus *beentry = MyBEEntry;
3235
3236 if (!pgstat_track_activities || !beentry)
3237 return;
3238
3239 /*
3240 * Update my status entry, following the protocol of bumping
3241 * st_changecount before and after. We use a volatile pointer here to
3242 * ensure the compiler doesn't try to get cute.
3243 */
3244 PGSTAT_BEGIN_WRITE_ACTIVITY(beentry);
3245
3246 beentry->st_xact_start_timestamp = tstamp;
3247
3248 PGSTAT_END_WRITE_ACTIVITY(beentry);
3249 }
3250
3251 /* ----------
3252 * pgstat_read_current_status() -
3253 *
3254 * Copy the current contents of the PgBackendStatus array to local memory,
3255 * if not already done in this transaction.
3256 * ----------
3257 */
3258 static void
pgstat_read_current_status(void)3259 pgstat_read_current_status(void)
3260 {
3261 volatile PgBackendStatus *beentry;
3262 LocalPgBackendStatus *localtable;
3263 LocalPgBackendStatus *localentry;
3264 char *localappname,
3265 *localclienthostname,
3266 *localactivity;
3267 #ifdef USE_SSL
3268 PgBackendSSLStatus *localsslstatus;
3269 #endif
3270 int i;
3271
3272 Assert(!pgStatRunningInCollector);
3273 if (localBackendStatusTable)
3274 return; /* already done */
3275
3276 pgstat_setup_memcxt();
3277
3278 localtable = (LocalPgBackendStatus *)
3279 MemoryContextAlloc(pgStatLocalContext,
3280 sizeof(LocalPgBackendStatus) * NumBackendStatSlots);
3281 localappname = (char *)
3282 MemoryContextAlloc(pgStatLocalContext,
3283 NAMEDATALEN * NumBackendStatSlots);
3284 localclienthostname = (char *)
3285 MemoryContextAlloc(pgStatLocalContext,
3286 NAMEDATALEN * NumBackendStatSlots);
3287 localactivity = (char *)
3288 MemoryContextAlloc(pgStatLocalContext,
3289 pgstat_track_activity_query_size * NumBackendStatSlots);
3290 #ifdef USE_SSL
3291 localsslstatus = (PgBackendSSLStatus *)
3292 MemoryContextAlloc(pgStatLocalContext,
3293 sizeof(PgBackendSSLStatus) * NumBackendStatSlots);
3294 #endif
3295
3296 localNumBackends = 0;
3297
3298 beentry = BackendStatusArray;
3299 localentry = localtable;
3300 for (i = 1; i <= NumBackendStatSlots; i++)
3301 {
3302 /*
3303 * Follow the protocol of retrying if st_changecount changes while we
3304 * copy the entry, or if it's odd. (The check for odd is needed to
3305 * cover the case where we are able to completely copy the entry while
3306 * the source backend is between increment steps.) We use a volatile
3307 * pointer here to ensure the compiler doesn't try to get cute.
3308 */
3309 for (;;)
3310 {
3311 int before_changecount;
3312 int after_changecount;
3313
3314 pgstat_begin_read_activity(beentry, before_changecount);
3315
3316 localentry->backendStatus.st_procpid = beentry->st_procpid;
3317 /* Skip all the data-copying work if entry is not in use */
3318 if (localentry->backendStatus.st_procpid > 0)
3319 {
3320 memcpy(&localentry->backendStatus, (char *) beentry, sizeof(PgBackendStatus));
3321
3322 /*
3323 * For each PgBackendStatus field that is a pointer, copy the
3324 * pointed-to data, then adjust the local copy of the pointer
3325 * field to point at the local copy of the data.
3326 *
3327 * strcpy is safe even if the string is modified concurrently,
3328 * because there's always a \0 at the end of the buffer.
3329 */
3330 strcpy(localappname, (char *) beentry->st_appname);
3331 localentry->backendStatus.st_appname = localappname;
3332 strcpy(localclienthostname, (char *) beentry->st_clienthostname);
3333 localentry->backendStatus.st_clienthostname = localclienthostname;
3334 strcpy(localactivity, (char *) beentry->st_activity);
3335 localentry->backendStatus.st_activity = localactivity;
3336 #ifdef USE_SSL
3337 if (beentry->st_ssl)
3338 {
3339 memcpy(localsslstatus, beentry->st_sslstatus, sizeof(PgBackendSSLStatus));
3340 localentry->backendStatus.st_sslstatus = localsslstatus;
3341 }
3342 #endif
3343 }
3344
3345 pgstat_end_read_activity(beentry, after_changecount);
3346
3347 if (pgstat_read_activity_complete(before_changecount,
3348 after_changecount))
3349 break;
3350
3351 /* Make sure we can break out of loop if stuck... */
3352 CHECK_FOR_INTERRUPTS();
3353 }
3354
3355 beentry++;
3356 /* Only valid entries get included into the local array */
3357 if (localentry->backendStatus.st_procpid > 0)
3358 {
3359 BackendIdGetTransactionIds(i,
3360 &localentry->backend_xid,
3361 &localentry->backend_xmin);
3362
3363 localentry++;
3364 localappname += NAMEDATALEN;
3365 localclienthostname += NAMEDATALEN;
3366 localactivity += pgstat_track_activity_query_size;
3367 #ifdef USE_SSL
3368 localsslstatus++;
3369 #endif
3370 localNumBackends++;
3371 }
3372 }
3373
3374 /* Set the pointer only after completion of a valid table */
3375 localBackendStatusTable = localtable;
3376 }
3377
3378 /* ----------
3379 * pgstat_get_wait_event_type() -
3380 *
3381 * Return a string representing the current wait event type, backend is
3382 * waiting on.
3383 */
3384 const char *
pgstat_get_wait_event_type(uint32 wait_event_info)3385 pgstat_get_wait_event_type(uint32 wait_event_info)
3386 {
3387 uint32 classId;
3388 const char *event_type;
3389
3390 /* report process as not waiting. */
3391 if (wait_event_info == 0)
3392 return NULL;
3393
3394 classId = wait_event_info & 0xFF000000;
3395
3396 switch (classId)
3397 {
3398 case PG_WAIT_LWLOCK:
3399 event_type = "LWLock";
3400 break;
3401 case PG_WAIT_LOCK:
3402 event_type = "Lock";
3403 break;
3404 case PG_WAIT_BUFFER_PIN:
3405 event_type = "BufferPin";
3406 break;
3407 case PG_WAIT_ACTIVITY:
3408 event_type = "Activity";
3409 break;
3410 case PG_WAIT_CLIENT:
3411 event_type = "Client";
3412 break;
3413 case PG_WAIT_EXTENSION:
3414 event_type = "Extension";
3415 break;
3416 case PG_WAIT_IPC:
3417 event_type = "IPC";
3418 break;
3419 case PG_WAIT_TIMEOUT:
3420 event_type = "Timeout";
3421 break;
3422 case PG_WAIT_IO:
3423 event_type = "IO";
3424 break;
3425 default:
3426 event_type = "???";
3427 break;
3428 }
3429
3430 return event_type;
3431 }
3432
3433 /* ----------
3434 * pgstat_get_wait_event() -
3435 *
3436 * Return a string representing the current wait event, backend is
3437 * waiting on.
3438 */
3439 const char *
pgstat_get_wait_event(uint32 wait_event_info)3440 pgstat_get_wait_event(uint32 wait_event_info)
3441 {
3442 uint32 classId;
3443 uint16 eventId;
3444 const char *event_name;
3445
3446 /* report process as not waiting. */
3447 if (wait_event_info == 0)
3448 return NULL;
3449
3450 classId = wait_event_info & 0xFF000000;
3451 eventId = wait_event_info & 0x0000FFFF;
3452
3453 switch (classId)
3454 {
3455 case PG_WAIT_LWLOCK:
3456 event_name = GetLWLockIdentifier(classId, eventId);
3457 break;
3458 case PG_WAIT_LOCK:
3459 event_name = GetLockNameFromTagType(eventId);
3460 break;
3461 case PG_WAIT_BUFFER_PIN:
3462 event_name = "BufferPin";
3463 break;
3464 case PG_WAIT_ACTIVITY:
3465 {
3466 WaitEventActivity w = (WaitEventActivity) wait_event_info;
3467
3468 event_name = pgstat_get_wait_activity(w);
3469 break;
3470 }
3471 case PG_WAIT_CLIENT:
3472 {
3473 WaitEventClient w = (WaitEventClient) wait_event_info;
3474
3475 event_name = pgstat_get_wait_client(w);
3476 break;
3477 }
3478 case PG_WAIT_EXTENSION:
3479 event_name = "Extension";
3480 break;
3481 case PG_WAIT_IPC:
3482 {
3483 WaitEventIPC w = (WaitEventIPC) wait_event_info;
3484
3485 event_name = pgstat_get_wait_ipc(w);
3486 break;
3487 }
3488 case PG_WAIT_TIMEOUT:
3489 {
3490 WaitEventTimeout w = (WaitEventTimeout) wait_event_info;
3491
3492 event_name = pgstat_get_wait_timeout(w);
3493 break;
3494 }
3495 case PG_WAIT_IO:
3496 {
3497 WaitEventIO w = (WaitEventIO) wait_event_info;
3498
3499 event_name = pgstat_get_wait_io(w);
3500 break;
3501 }
3502 default:
3503 event_name = "unknown wait event";
3504 break;
3505 }
3506
3507 return event_name;
3508 }
3509
3510 /* ----------
3511 * pgstat_get_wait_activity() -
3512 *
3513 * Convert WaitEventActivity to string.
3514 * ----------
3515 */
3516 static const char *
pgstat_get_wait_activity(WaitEventActivity w)3517 pgstat_get_wait_activity(WaitEventActivity w)
3518 {
3519 const char *event_name = "unknown wait event";
3520
3521 switch (w)
3522 {
3523 case WAIT_EVENT_ARCHIVER_MAIN:
3524 event_name = "ArchiverMain";
3525 break;
3526 case WAIT_EVENT_AUTOVACUUM_MAIN:
3527 event_name = "AutoVacuumMain";
3528 break;
3529 case WAIT_EVENT_BGWRITER_HIBERNATE:
3530 event_name = "BgWriterHibernate";
3531 break;
3532 case WAIT_EVENT_BGWRITER_MAIN:
3533 event_name = "BgWriterMain";
3534 break;
3535 case WAIT_EVENT_CHECKPOINTER_MAIN:
3536 event_name = "CheckpointerMain";
3537 break;
3538 case WAIT_EVENT_LOGICAL_LAUNCHER_MAIN:
3539 event_name = "LogicalLauncherMain";
3540 break;
3541 case WAIT_EVENT_LOGICAL_APPLY_MAIN:
3542 event_name = "LogicalApplyMain";
3543 break;
3544 case WAIT_EVENT_PGSTAT_MAIN:
3545 event_name = "PgStatMain";
3546 break;
3547 case WAIT_EVENT_RECOVERY_WAL_ALL:
3548 event_name = "RecoveryWalAll";
3549 break;
3550 case WAIT_EVENT_RECOVERY_WAL_STREAM:
3551 event_name = "RecoveryWalStream";
3552 break;
3553 case WAIT_EVENT_SYSLOGGER_MAIN:
3554 event_name = "SysLoggerMain";
3555 break;
3556 case WAIT_EVENT_WAL_RECEIVER_MAIN:
3557 event_name = "WalReceiverMain";
3558 break;
3559 case WAIT_EVENT_WAL_SENDER_MAIN:
3560 event_name = "WalSenderMain";
3561 break;
3562 case WAIT_EVENT_WAL_WRITER_MAIN:
3563 event_name = "WalWriterMain";
3564 break;
3565 /* no default case, so that compiler will warn */
3566 }
3567
3568 return event_name;
3569 }
3570
3571 /* ----------
3572 * pgstat_get_wait_client() -
3573 *
3574 * Convert WaitEventClient to string.
3575 * ----------
3576 */
3577 static const char *
pgstat_get_wait_client(WaitEventClient w)3578 pgstat_get_wait_client(WaitEventClient w)
3579 {
3580 const char *event_name = "unknown wait event";
3581
3582 switch (w)
3583 {
3584 case WAIT_EVENT_CLIENT_READ:
3585 event_name = "ClientRead";
3586 break;
3587 case WAIT_EVENT_CLIENT_WRITE:
3588 event_name = "ClientWrite";
3589 break;
3590 case WAIT_EVENT_LIBPQWALRECEIVER_CONNECT:
3591 event_name = "LibPQWalReceiverConnect";
3592 break;
3593 case WAIT_EVENT_LIBPQWALRECEIVER_RECEIVE:
3594 event_name = "LibPQWalReceiverReceive";
3595 break;
3596 case WAIT_EVENT_SSL_OPEN_SERVER:
3597 event_name = "SSLOpenServer";
3598 break;
3599 case WAIT_EVENT_WAL_RECEIVER_WAIT_START:
3600 event_name = "WalReceiverWaitStart";
3601 break;
3602 case WAIT_EVENT_WAL_SENDER_WAIT_WAL:
3603 event_name = "WalSenderWaitForWAL";
3604 break;
3605 case WAIT_EVENT_WAL_SENDER_WRITE_DATA:
3606 event_name = "WalSenderWriteData";
3607 break;
3608 /* no default case, so that compiler will warn */
3609 }
3610
3611 return event_name;
3612 }
3613
3614 /* ----------
3615 * pgstat_get_wait_ipc() -
3616 *
3617 * Convert WaitEventIPC to string.
3618 * ----------
3619 */
3620 static const char *
pgstat_get_wait_ipc(WaitEventIPC w)3621 pgstat_get_wait_ipc(WaitEventIPC w)
3622 {
3623 const char *event_name = "unknown wait event";
3624
3625 switch (w)
3626 {
3627 case WAIT_EVENT_BGWORKER_SHUTDOWN:
3628 event_name = "BgWorkerShutdown";
3629 break;
3630 case WAIT_EVENT_BGWORKER_STARTUP:
3631 event_name = "BgWorkerStartup";
3632 break;
3633 case WAIT_EVENT_BTREE_PAGE:
3634 event_name = "BtreePage";
3635 break;
3636 case WAIT_EVENT_EXECUTE_GATHER:
3637 event_name = "ExecuteGather";
3638 break;
3639 case WAIT_EVENT_LOGICAL_SYNC_DATA:
3640 event_name = "LogicalSyncData";
3641 break;
3642 case WAIT_EVENT_LOGICAL_SYNC_STATE_CHANGE:
3643 event_name = "LogicalSyncStateChange";
3644 break;
3645 case WAIT_EVENT_MQ_INTERNAL:
3646 event_name = "MessageQueueInternal";
3647 break;
3648 case WAIT_EVENT_MQ_PUT_MESSAGE:
3649 event_name = "MessageQueuePutMessage";
3650 break;
3651 case WAIT_EVENT_MQ_RECEIVE:
3652 event_name = "MessageQueueReceive";
3653 break;
3654 case WAIT_EVENT_MQ_SEND:
3655 event_name = "MessageQueueSend";
3656 break;
3657 case WAIT_EVENT_PARALLEL_FINISH:
3658 event_name = "ParallelFinish";
3659 break;
3660 case WAIT_EVENT_PARALLEL_BITMAP_SCAN:
3661 event_name = "ParallelBitmapScan";
3662 break;
3663 case WAIT_EVENT_PROCARRAY_GROUP_UPDATE:
3664 event_name = "ProcArrayGroupUpdate";
3665 break;
3666 case WAIT_EVENT_REPLICATION_ORIGIN_DROP:
3667 event_name = "ReplicationOriginDrop";
3668 break;
3669 case WAIT_EVENT_REPLICATION_SLOT_DROP:
3670 event_name = "ReplicationSlotDrop";
3671 break;
3672 case WAIT_EVENT_SAFE_SNAPSHOT:
3673 event_name = "SafeSnapshot";
3674 break;
3675 case WAIT_EVENT_SYNC_REP:
3676 event_name = "SyncRep";
3677 break;
3678 /* no default case, so that compiler will warn */
3679 }
3680
3681 return event_name;
3682 }
3683
3684 /* ----------
3685 * pgstat_get_wait_timeout() -
3686 *
3687 * Convert WaitEventTimeout to string.
3688 * ----------
3689 */
3690 static const char *
pgstat_get_wait_timeout(WaitEventTimeout w)3691 pgstat_get_wait_timeout(WaitEventTimeout w)
3692 {
3693 const char *event_name = "unknown wait event";
3694
3695 switch (w)
3696 {
3697 case WAIT_EVENT_BASE_BACKUP_THROTTLE:
3698 event_name = "BaseBackupThrottle";
3699 break;
3700 case WAIT_EVENT_PG_SLEEP:
3701 event_name = "PgSleep";
3702 break;
3703 case WAIT_EVENT_RECOVERY_APPLY_DELAY:
3704 event_name = "RecoveryApplyDelay";
3705 break;
3706 /* no default case, so that compiler will warn */
3707 }
3708
3709 return event_name;
3710 }
3711
3712 /* ----------
3713 * pgstat_get_wait_io() -
3714 *
3715 * Convert WaitEventIO to string.
3716 * ----------
3717 */
3718 static const char *
pgstat_get_wait_io(WaitEventIO w)3719 pgstat_get_wait_io(WaitEventIO w)
3720 {
3721 const char *event_name = "unknown wait event";
3722
3723 switch (w)
3724 {
3725 case WAIT_EVENT_BUFFILE_READ:
3726 event_name = "BufFileRead";
3727 break;
3728 case WAIT_EVENT_BUFFILE_WRITE:
3729 event_name = "BufFileWrite";
3730 break;
3731 case WAIT_EVENT_CONTROL_FILE_READ:
3732 event_name = "ControlFileRead";
3733 break;
3734 case WAIT_EVENT_CONTROL_FILE_SYNC:
3735 event_name = "ControlFileSync";
3736 break;
3737 case WAIT_EVENT_CONTROL_FILE_SYNC_UPDATE:
3738 event_name = "ControlFileSyncUpdate";
3739 break;
3740 case WAIT_EVENT_CONTROL_FILE_WRITE:
3741 event_name = "ControlFileWrite";
3742 break;
3743 case WAIT_EVENT_CONTROL_FILE_WRITE_UPDATE:
3744 event_name = "ControlFileWriteUpdate";
3745 break;
3746 case WAIT_EVENT_COPY_FILE_READ:
3747 event_name = "CopyFileRead";
3748 break;
3749 case WAIT_EVENT_COPY_FILE_WRITE:
3750 event_name = "CopyFileWrite";
3751 break;
3752 case WAIT_EVENT_DATA_FILE_EXTEND:
3753 event_name = "DataFileExtend";
3754 break;
3755 case WAIT_EVENT_DATA_FILE_FLUSH:
3756 event_name = "DataFileFlush";
3757 break;
3758 case WAIT_EVENT_DATA_FILE_IMMEDIATE_SYNC:
3759 event_name = "DataFileImmediateSync";
3760 break;
3761 case WAIT_EVENT_DATA_FILE_PREFETCH:
3762 event_name = "DataFilePrefetch";
3763 break;
3764 case WAIT_EVENT_DATA_FILE_READ:
3765 event_name = "DataFileRead";
3766 break;
3767 case WAIT_EVENT_DATA_FILE_SYNC:
3768 event_name = "DataFileSync";
3769 break;
3770 case WAIT_EVENT_DATA_FILE_TRUNCATE:
3771 event_name = "DataFileTruncate";
3772 break;
3773 case WAIT_EVENT_DATA_FILE_WRITE:
3774 event_name = "DataFileWrite";
3775 break;
3776 case WAIT_EVENT_DSM_FILL_ZERO_WRITE:
3777 event_name = "DSMFillZeroWrite";
3778 break;
3779 case WAIT_EVENT_LOCK_FILE_ADDTODATADIR_READ:
3780 event_name = "LockFileAddToDataDirRead";
3781 break;
3782 case WAIT_EVENT_LOCK_FILE_ADDTODATADIR_SYNC:
3783 event_name = "LockFileAddToDataDirSync";
3784 break;
3785 case WAIT_EVENT_LOCK_FILE_ADDTODATADIR_WRITE:
3786 event_name = "LockFileAddToDataDirWrite";
3787 break;
3788 case WAIT_EVENT_LOCK_FILE_CREATE_READ:
3789 event_name = "LockFileCreateRead";
3790 break;
3791 case WAIT_EVENT_LOCK_FILE_CREATE_SYNC:
3792 event_name = "LockFileCreateSync";
3793 break;
3794 case WAIT_EVENT_LOCK_FILE_CREATE_WRITE:
3795 event_name = "LockFileCreateWrite";
3796 break;
3797 case WAIT_EVENT_LOCK_FILE_RECHECKDATADIR_READ:
3798 event_name = "LockFileReCheckDataDirRead";
3799 break;
3800 case WAIT_EVENT_LOGICAL_REWRITE_CHECKPOINT_SYNC:
3801 event_name = "LogicalRewriteCheckpointSync";
3802 break;
3803 case WAIT_EVENT_LOGICAL_REWRITE_MAPPING_SYNC:
3804 event_name = "LogicalRewriteMappingSync";
3805 break;
3806 case WAIT_EVENT_LOGICAL_REWRITE_MAPPING_WRITE:
3807 event_name = "LogicalRewriteMappingWrite";
3808 break;
3809 case WAIT_EVENT_LOGICAL_REWRITE_SYNC:
3810 event_name = "LogicalRewriteSync";
3811 break;
3812 case WAIT_EVENT_LOGICAL_REWRITE_TRUNCATE:
3813 event_name = "LogicalRewriteTruncate";
3814 break;
3815 case WAIT_EVENT_LOGICAL_REWRITE_WRITE:
3816 event_name = "LogicalRewriteWrite";
3817 break;
3818 case WAIT_EVENT_RELATION_MAP_READ:
3819 event_name = "RelationMapRead";
3820 break;
3821 case WAIT_EVENT_RELATION_MAP_SYNC:
3822 event_name = "RelationMapSync";
3823 break;
3824 case WAIT_EVENT_RELATION_MAP_WRITE:
3825 event_name = "RelationMapWrite";
3826 break;
3827 case WAIT_EVENT_REORDER_BUFFER_READ:
3828 event_name = "ReorderBufferRead";
3829 break;
3830 case WAIT_EVENT_REORDER_BUFFER_WRITE:
3831 event_name = "ReorderBufferWrite";
3832 break;
3833 case WAIT_EVENT_REORDER_LOGICAL_MAPPING_READ:
3834 event_name = "ReorderLogicalMappingRead";
3835 break;
3836 case WAIT_EVENT_REPLICATION_SLOT_READ:
3837 event_name = "ReplicationSlotRead";
3838 break;
3839 case WAIT_EVENT_REPLICATION_SLOT_RESTORE_SYNC:
3840 event_name = "ReplicationSlotRestoreSync";
3841 break;
3842 case WAIT_EVENT_REPLICATION_SLOT_SYNC:
3843 event_name = "ReplicationSlotSync";
3844 break;
3845 case WAIT_EVENT_REPLICATION_SLOT_WRITE:
3846 event_name = "ReplicationSlotWrite";
3847 break;
3848 case WAIT_EVENT_SLRU_FLUSH_SYNC:
3849 event_name = "SLRUFlushSync";
3850 break;
3851 case WAIT_EVENT_SLRU_READ:
3852 event_name = "SLRURead";
3853 break;
3854 case WAIT_EVENT_SLRU_SYNC:
3855 event_name = "SLRUSync";
3856 break;
3857 case WAIT_EVENT_SLRU_WRITE:
3858 event_name = "SLRUWrite";
3859 break;
3860 case WAIT_EVENT_SNAPBUILD_READ:
3861 event_name = "SnapbuildRead";
3862 break;
3863 case WAIT_EVENT_SNAPBUILD_SYNC:
3864 event_name = "SnapbuildSync";
3865 break;
3866 case WAIT_EVENT_SNAPBUILD_WRITE:
3867 event_name = "SnapbuildWrite";
3868 break;
3869 case WAIT_EVENT_TIMELINE_HISTORY_FILE_SYNC:
3870 event_name = "TimelineHistoryFileSync";
3871 break;
3872 case WAIT_EVENT_TIMELINE_HISTORY_FILE_WRITE:
3873 event_name = "TimelineHistoryFileWrite";
3874 break;
3875 case WAIT_EVENT_TIMELINE_HISTORY_READ:
3876 event_name = "TimelineHistoryRead";
3877 break;
3878 case WAIT_EVENT_TIMELINE_HISTORY_SYNC:
3879 event_name = "TimelineHistorySync";
3880 break;
3881 case WAIT_EVENT_TIMELINE_HISTORY_WRITE:
3882 event_name = "TimelineHistoryWrite";
3883 break;
3884 case WAIT_EVENT_TWOPHASE_FILE_READ:
3885 event_name = "TwophaseFileRead";
3886 break;
3887 case WAIT_EVENT_TWOPHASE_FILE_SYNC:
3888 event_name = "TwophaseFileSync";
3889 break;
3890 case WAIT_EVENT_TWOPHASE_FILE_WRITE:
3891 event_name = "TwophaseFileWrite";
3892 break;
3893 case WAIT_EVENT_WALSENDER_TIMELINE_HISTORY_READ:
3894 event_name = "WALSenderTimelineHistoryRead";
3895 break;
3896 case WAIT_EVENT_WAL_BOOTSTRAP_SYNC:
3897 event_name = "WALBootstrapSync";
3898 break;
3899 case WAIT_EVENT_WAL_BOOTSTRAP_WRITE:
3900 event_name = "WALBootstrapWrite";
3901 break;
3902 case WAIT_EVENT_WAL_COPY_READ:
3903 event_name = "WALCopyRead";
3904 break;
3905 case WAIT_EVENT_WAL_COPY_SYNC:
3906 event_name = "WALCopySync";
3907 break;
3908 case WAIT_EVENT_WAL_COPY_WRITE:
3909 event_name = "WALCopyWrite";
3910 break;
3911 case WAIT_EVENT_WAL_INIT_SYNC:
3912 event_name = "WALInitSync";
3913 break;
3914 case WAIT_EVENT_WAL_INIT_WRITE:
3915 event_name = "WALInitWrite";
3916 break;
3917 case WAIT_EVENT_WAL_READ:
3918 event_name = "WALRead";
3919 break;
3920 case WAIT_EVENT_WAL_SYNC_METHOD_ASSIGN:
3921 event_name = "WALSyncMethodAssign";
3922 break;
3923 case WAIT_EVENT_WAL_WRITE:
3924 event_name = "WALWrite";
3925 break;
3926
3927 /* no default case, so that compiler will warn */
3928 }
3929
3930 return event_name;
3931 }
3932
3933
3934 /* ----------
3935 * pgstat_get_backend_current_activity() -
3936 *
3937 * Return a string representing the current activity of the backend with
3938 * the specified PID. This looks directly at the BackendStatusArray,
3939 * and so will provide current information regardless of the age of our
3940 * transaction's snapshot of the status array.
3941 *
3942 * It is the caller's responsibility to invoke this only for backends whose
3943 * state is expected to remain stable while the result is in use. The
3944 * only current use is in deadlock reporting, where we can expect that
3945 * the target backend is blocked on a lock. (There are corner cases
3946 * where the target's wait could get aborted while we are looking at it,
3947 * but the very worst consequence is to return a pointer to a string
3948 * that's been changed, so we won't worry too much.)
3949 *
3950 * Note: return strings for special cases match pg_stat_get_backend_activity.
3951 * ----------
3952 */
3953 const char *
pgstat_get_backend_current_activity(int pid,bool checkUser)3954 pgstat_get_backend_current_activity(int pid, bool checkUser)
3955 {
3956 PgBackendStatus *beentry;
3957 int i;
3958
3959 beentry = BackendStatusArray;
3960 for (i = 1; i <= MaxBackends; i++)
3961 {
3962 /*
3963 * Although we expect the target backend's entry to be stable, that
3964 * doesn't imply that anyone else's is. To avoid identifying the
3965 * wrong backend, while we check for a match to the desired PID we
3966 * must follow the protocol of retrying if st_changecount changes
3967 * while we examine the entry, or if it's odd. (This might be
3968 * unnecessary, since fetching or storing an int is almost certainly
3969 * atomic, but let's play it safe.) We use a volatile pointer here to
3970 * ensure the compiler doesn't try to get cute.
3971 */
3972 volatile PgBackendStatus *vbeentry = beentry;
3973 bool found;
3974
3975 for (;;)
3976 {
3977 int before_changecount;
3978 int after_changecount;
3979
3980 pgstat_begin_read_activity(vbeentry, before_changecount);
3981
3982 found = (vbeentry->st_procpid == pid);
3983
3984 pgstat_end_read_activity(vbeentry, after_changecount);
3985
3986 if (pgstat_read_activity_complete(before_changecount,
3987 after_changecount))
3988 break;
3989
3990 /* Make sure we can break out of loop if stuck... */
3991 CHECK_FOR_INTERRUPTS();
3992 }
3993
3994 if (found)
3995 {
3996 /* Now it is safe to use the non-volatile pointer */
3997 if (checkUser && !superuser() && beentry->st_userid != GetUserId())
3998 return "<insufficient privilege>";
3999 else if (*(beentry->st_activity) == '\0')
4000 return "<command string not enabled>";
4001 else
4002 return beentry->st_activity;
4003 }
4004
4005 beentry++;
4006 }
4007
4008 /* If we get here, caller is in error ... */
4009 return "<backend information not available>";
4010 }
4011
4012 /* ----------
4013 * pgstat_get_crashed_backend_activity() -
4014 *
4015 * Return a string representing the current activity of the backend with
4016 * the specified PID. Like the function above, but reads shared memory with
4017 * the expectation that it may be corrupt. On success, copy the string
4018 * into the "buffer" argument and return that pointer. On failure,
4019 * return NULL.
4020 *
4021 * This function is only intended to be used by the postmaster to report the
4022 * query that crashed a backend. In particular, no attempt is made to
4023 * follow the correct concurrency protocol when accessing the
4024 * BackendStatusArray. But that's OK, in the worst case we'll return a
4025 * corrupted message. We also must take care not to trip on ereport(ERROR).
4026 * ----------
4027 */
4028 const char *
pgstat_get_crashed_backend_activity(int pid,char * buffer,int buflen)4029 pgstat_get_crashed_backend_activity(int pid, char *buffer, int buflen)
4030 {
4031 volatile PgBackendStatus *beentry;
4032 int i;
4033
4034 beentry = BackendStatusArray;
4035
4036 /*
4037 * We probably shouldn't get here before shared memory has been set up,
4038 * but be safe.
4039 */
4040 if (beentry == NULL || BackendActivityBuffer == NULL)
4041 return NULL;
4042
4043 for (i = 1; i <= MaxBackends; i++)
4044 {
4045 if (beentry->st_procpid == pid)
4046 {
4047 /* Read pointer just once, so it can't change after validation */
4048 const char *activity = beentry->st_activity;
4049 const char *activity_last;
4050
4051 /*
4052 * We mustn't access activity string before we verify that it
4053 * falls within the BackendActivityBuffer. To make sure that the
4054 * entire string including its ending is contained within the
4055 * buffer, subtract one activity length from the buffer size.
4056 */
4057 activity_last = BackendActivityBuffer + BackendActivityBufferSize
4058 - pgstat_track_activity_query_size;
4059
4060 if (activity < BackendActivityBuffer ||
4061 activity > activity_last)
4062 return NULL;
4063
4064 /* If no string available, no point in a report */
4065 if (activity[0] == '\0')
4066 return NULL;
4067
4068 /*
4069 * Copy only ASCII-safe characters so we don't run into encoding
4070 * problems when reporting the message; and be sure not to run off
4071 * the end of memory.
4072 */
4073 ascii_safe_strlcpy(buffer, activity,
4074 Min(buflen, pgstat_track_activity_query_size));
4075
4076 return buffer;
4077 }
4078
4079 beentry++;
4080 }
4081
4082 /* PID not found */
4083 return NULL;
4084 }
4085
4086 const char *
pgstat_get_backend_desc(BackendType backendType)4087 pgstat_get_backend_desc(BackendType backendType)
4088 {
4089 const char *backendDesc = "unknown process type";
4090
4091 switch (backendType)
4092 {
4093 case B_AUTOVAC_LAUNCHER:
4094 backendDesc = "autovacuum launcher";
4095 break;
4096 case B_AUTOVAC_WORKER:
4097 backendDesc = "autovacuum worker";
4098 break;
4099 case B_BACKEND:
4100 backendDesc = "client backend";
4101 break;
4102 case B_BG_WORKER:
4103 backendDesc = "background worker";
4104 break;
4105 case B_BG_WRITER:
4106 backendDesc = "background writer";
4107 break;
4108 case B_CHECKPOINTER:
4109 backendDesc = "checkpointer";
4110 break;
4111 case B_STARTUP:
4112 backendDesc = "startup";
4113 break;
4114 case B_WAL_RECEIVER:
4115 backendDesc = "walreceiver";
4116 break;
4117 case B_WAL_SENDER:
4118 backendDesc = "walsender";
4119 break;
4120 case B_WAL_WRITER:
4121 backendDesc = "walwriter";
4122 break;
4123 }
4124
4125 return backendDesc;
4126 }
4127
4128 /* ------------------------------------------------------------
4129 * Local support functions follow
4130 * ------------------------------------------------------------
4131 */
4132
4133
4134 /* ----------
4135 * pgstat_setheader() -
4136 *
4137 * Set common header fields in a statistics message
4138 * ----------
4139 */
4140 static void
pgstat_setheader(PgStat_MsgHdr * hdr,StatMsgType mtype)4141 pgstat_setheader(PgStat_MsgHdr *hdr, StatMsgType mtype)
4142 {
4143 hdr->m_type = mtype;
4144 }
4145
4146
4147 /* ----------
4148 * pgstat_send() -
4149 *
4150 * Send out one statistics message to the collector
4151 * ----------
4152 */
4153 static void
pgstat_send(void * msg,int len)4154 pgstat_send(void *msg, int len)
4155 {
4156 int rc;
4157
4158 if (pgStatSock == PGINVALID_SOCKET)
4159 return;
4160
4161 ((PgStat_MsgHdr *) msg)->m_size = len;
4162
4163 /* We'll retry after EINTR, but ignore all other failures */
4164 do
4165 {
4166 rc = send(pgStatSock, msg, len, 0);
4167 } while (rc < 0 && errno == EINTR);
4168
4169 #ifdef USE_ASSERT_CHECKING
4170 /* In debug builds, log send failures ... */
4171 if (rc < 0)
4172 elog(LOG, "could not send to statistics collector: %m");
4173 #endif
4174 }
4175
4176 /* ----------
4177 * pgstat_send_archiver() -
4178 *
4179 * Tell the collector about the WAL file that we successfully
4180 * archived or failed to archive.
4181 * ----------
4182 */
4183 void
pgstat_send_archiver(const char * xlog,bool failed)4184 pgstat_send_archiver(const char *xlog, bool failed)
4185 {
4186 PgStat_MsgArchiver msg;
4187
4188 /*
4189 * Prepare and send the message
4190 */
4191 pgstat_setheader(&msg.m_hdr, PGSTAT_MTYPE_ARCHIVER);
4192 msg.m_failed = failed;
4193 StrNCpy(msg.m_xlog, xlog, sizeof(msg.m_xlog));
4194 msg.m_timestamp = GetCurrentTimestamp();
4195 pgstat_send(&msg, sizeof(msg));
4196 }
4197
4198 /* ----------
4199 * pgstat_send_bgwriter() -
4200 *
4201 * Send bgwriter statistics to the collector
4202 * ----------
4203 */
4204 void
pgstat_send_bgwriter(void)4205 pgstat_send_bgwriter(void)
4206 {
4207 /* We assume this initializes to zeroes */
4208 static const PgStat_MsgBgWriter all_zeroes;
4209
4210 /*
4211 * This function can be called even if nothing at all has happened. In
4212 * this case, avoid sending a completely empty message to the stats
4213 * collector.
4214 */
4215 if (memcmp(&BgWriterStats, &all_zeroes, sizeof(PgStat_MsgBgWriter)) == 0)
4216 return;
4217
4218 /*
4219 * Prepare and send the message
4220 */
4221 pgstat_setheader(&BgWriterStats.m_hdr, PGSTAT_MTYPE_BGWRITER);
4222 pgstat_send(&BgWriterStats, sizeof(BgWriterStats));
4223
4224 /*
4225 * Clear out the statistics buffer, so it can be re-used.
4226 */
4227 MemSet(&BgWriterStats, 0, sizeof(BgWriterStats));
4228 }
4229
4230
4231 /* ----------
4232 * PgstatCollectorMain() -
4233 *
4234 * Start up the statistics collector process. This is the body of the
4235 * postmaster child process.
4236 *
4237 * The argc/argv parameters are valid only in EXEC_BACKEND case.
4238 * ----------
4239 */
4240 NON_EXEC_STATIC void
PgstatCollectorMain(int argc,char * argv[])4241 PgstatCollectorMain(int argc, char *argv[])
4242 {
4243 int len;
4244 PgStat_Msg msg;
4245 int wr;
4246
4247 /*
4248 * Ignore all signals usually bound to some action in the postmaster,
4249 * except SIGHUP and SIGQUIT. Note we don't need a SIGUSR1 handler to
4250 * support latch operations, because we only use a local latch.
4251 */
4252 pqsignal(SIGHUP, pgstat_sighup_handler);
4253 pqsignal(SIGINT, SIG_IGN);
4254 pqsignal(SIGTERM, SIG_IGN);
4255 pqsignal(SIGQUIT, pgstat_exit);
4256 pqsignal(SIGALRM, SIG_IGN);
4257 pqsignal(SIGPIPE, SIG_IGN);
4258 pqsignal(SIGUSR1, SIG_IGN);
4259 pqsignal(SIGUSR2, SIG_IGN);
4260 pqsignal(SIGCHLD, SIG_DFL);
4261 pqsignal(SIGTTIN, SIG_DFL);
4262 pqsignal(SIGTTOU, SIG_DFL);
4263 pqsignal(SIGCONT, SIG_DFL);
4264 pqsignal(SIGWINCH, SIG_DFL);
4265 PG_SETMASK(&UnBlockSig);
4266
4267 /*
4268 * Identify myself via ps
4269 */
4270 init_ps_display("stats collector process", "", "", "");
4271
4272 /*
4273 * Read in existing stats files or initialize the stats to zero.
4274 */
4275 pgStatRunningInCollector = true;
4276 pgStatDBHash = pgstat_read_statsfiles(InvalidOid, true, true);
4277
4278 /*
4279 * Loop to process messages until we get SIGQUIT or detect ungraceful
4280 * death of our parent postmaster.
4281 *
4282 * For performance reasons, we don't want to do ResetLatch/WaitLatch after
4283 * every message; instead, do that only after a recv() fails to obtain a
4284 * message. (This effectively means that if backends are sending us stuff
4285 * like mad, we won't notice postmaster death until things slack off a
4286 * bit; which seems fine.) To do that, we have an inner loop that
4287 * iterates as long as recv() succeeds. We do recognize got_SIGHUP inside
4288 * the inner loop, which means that such interrupts will get serviced but
4289 * the latch won't get cleared until next time there is a break in the
4290 * action.
4291 */
4292 for (;;)
4293 {
4294 /* Clear any already-pending wakeups */
4295 ResetLatch(MyLatch);
4296
4297 /*
4298 * Quit if we get SIGQUIT from the postmaster.
4299 */
4300 if (need_exit)
4301 break;
4302
4303 /*
4304 * Inner loop iterates as long as we keep getting messages, or until
4305 * need_exit becomes set.
4306 */
4307 while (!need_exit)
4308 {
4309 /*
4310 * Reload configuration if we got SIGHUP from the postmaster.
4311 */
4312 if (got_SIGHUP)
4313 {
4314 got_SIGHUP = false;
4315 ProcessConfigFile(PGC_SIGHUP);
4316 }
4317
4318 /*
4319 * Write the stats file(s) if a new request has arrived that is
4320 * not satisfied by existing file(s).
4321 */
4322 if (pgstat_write_statsfile_needed())
4323 pgstat_write_statsfiles(false, false);
4324
4325 /*
4326 * Try to receive and process a message. This will not block,
4327 * since the socket is set to non-blocking mode.
4328 *
4329 * XXX On Windows, we have to force pgwin32_recv to cooperate,
4330 * despite the previous use of pg_set_noblock() on the socket.
4331 * This is extremely broken and should be fixed someday.
4332 */
4333 #ifdef WIN32
4334 pgwin32_noblock = 1;
4335 #endif
4336
4337 len = recv(pgStatSock, (char *) &msg,
4338 sizeof(PgStat_Msg), 0);
4339
4340 #ifdef WIN32
4341 pgwin32_noblock = 0;
4342 #endif
4343
4344 if (len < 0)
4345 {
4346 if (errno == EAGAIN || errno == EWOULDBLOCK || errno == EINTR)
4347 break; /* out of inner loop */
4348 ereport(ERROR,
4349 (errcode_for_socket_access(),
4350 errmsg("could not read statistics message: %m")));
4351 }
4352
4353 /*
4354 * We ignore messages that are smaller than our common header
4355 */
4356 if (len < sizeof(PgStat_MsgHdr))
4357 continue;
4358
4359 /*
4360 * The received length must match the length in the header
4361 */
4362 if (msg.msg_hdr.m_size != len)
4363 continue;
4364
4365 /*
4366 * O.K. - we accept this message. Process it.
4367 */
4368 switch (msg.msg_hdr.m_type)
4369 {
4370 case PGSTAT_MTYPE_DUMMY:
4371 break;
4372
4373 case PGSTAT_MTYPE_INQUIRY:
4374 pgstat_recv_inquiry((PgStat_MsgInquiry *) &msg, len);
4375 break;
4376
4377 case PGSTAT_MTYPE_TABSTAT:
4378 pgstat_recv_tabstat((PgStat_MsgTabstat *) &msg, len);
4379 break;
4380
4381 case PGSTAT_MTYPE_TABPURGE:
4382 pgstat_recv_tabpurge((PgStat_MsgTabpurge *) &msg, len);
4383 break;
4384
4385 case PGSTAT_MTYPE_DROPDB:
4386 pgstat_recv_dropdb((PgStat_MsgDropdb *) &msg, len);
4387 break;
4388
4389 case PGSTAT_MTYPE_RESETCOUNTER:
4390 pgstat_recv_resetcounter((PgStat_MsgResetcounter *) &msg,
4391 len);
4392 break;
4393
4394 case PGSTAT_MTYPE_RESETSHAREDCOUNTER:
4395 pgstat_recv_resetsharedcounter(
4396 (PgStat_MsgResetsharedcounter *) &msg,
4397 len);
4398 break;
4399
4400 case PGSTAT_MTYPE_RESETSINGLECOUNTER:
4401 pgstat_recv_resetsinglecounter(
4402 (PgStat_MsgResetsinglecounter *) &msg,
4403 len);
4404 break;
4405
4406 case PGSTAT_MTYPE_AUTOVAC_START:
4407 pgstat_recv_autovac((PgStat_MsgAutovacStart *) &msg, len);
4408 break;
4409
4410 case PGSTAT_MTYPE_VACUUM:
4411 pgstat_recv_vacuum((PgStat_MsgVacuum *) &msg, len);
4412 break;
4413
4414 case PGSTAT_MTYPE_ANALYZE:
4415 pgstat_recv_analyze((PgStat_MsgAnalyze *) &msg, len);
4416 break;
4417
4418 case PGSTAT_MTYPE_ARCHIVER:
4419 pgstat_recv_archiver((PgStat_MsgArchiver *) &msg, len);
4420 break;
4421
4422 case PGSTAT_MTYPE_BGWRITER:
4423 pgstat_recv_bgwriter((PgStat_MsgBgWriter *) &msg, len);
4424 break;
4425
4426 case PGSTAT_MTYPE_FUNCSTAT:
4427 pgstat_recv_funcstat((PgStat_MsgFuncstat *) &msg, len);
4428 break;
4429
4430 case PGSTAT_MTYPE_FUNCPURGE:
4431 pgstat_recv_funcpurge((PgStat_MsgFuncpurge *) &msg, len);
4432 break;
4433
4434 case PGSTAT_MTYPE_RECOVERYCONFLICT:
4435 pgstat_recv_recoveryconflict((PgStat_MsgRecoveryConflict *) &msg, len);
4436 break;
4437
4438 case PGSTAT_MTYPE_DEADLOCK:
4439 pgstat_recv_deadlock((PgStat_MsgDeadlock *) &msg, len);
4440 break;
4441
4442 case PGSTAT_MTYPE_TEMPFILE:
4443 pgstat_recv_tempfile((PgStat_MsgTempFile *) &msg, len);
4444 break;
4445
4446 default:
4447 break;
4448 }
4449 } /* end of inner message-processing loop */
4450
4451 /* Sleep until there's something to do */
4452 #ifndef WIN32
4453 wr = WaitLatchOrSocket(MyLatch,
4454 WL_LATCH_SET | WL_POSTMASTER_DEATH | WL_SOCKET_READABLE,
4455 pgStatSock, -1L,
4456 WAIT_EVENT_PGSTAT_MAIN);
4457 #else
4458
4459 /*
4460 * Windows, at least in its Windows Server 2003 R2 incarnation,
4461 * sometimes loses FD_READ events. Waking up and retrying the recv()
4462 * fixes that, so don't sleep indefinitely. This is a crock of the
4463 * first water, but until somebody wants to debug exactly what's
4464 * happening there, this is the best we can do. The two-second
4465 * timeout matches our pre-9.2 behavior, and needs to be short enough
4466 * to not provoke "using stale statistics" complaints from
4467 * backend_read_statsfile.
4468 */
4469 wr = WaitLatchOrSocket(MyLatch,
4470 WL_LATCH_SET | WL_POSTMASTER_DEATH | WL_SOCKET_READABLE | WL_TIMEOUT,
4471 pgStatSock,
4472 2 * 1000L /* msec */ ,
4473 WAIT_EVENT_PGSTAT_MAIN);
4474 #endif
4475
4476 /*
4477 * Emergency bailout if postmaster has died. This is to avoid the
4478 * necessity for manual cleanup of all postmaster children.
4479 */
4480 if (wr & WL_POSTMASTER_DEATH)
4481 break;
4482 } /* end of outer loop */
4483
4484 /*
4485 * Save the final stats to reuse at next startup.
4486 */
4487 pgstat_write_statsfiles(true, true);
4488
4489 exit(0);
4490 }
4491
4492
4493 /* SIGQUIT signal handler for collector process */
4494 static void
pgstat_exit(SIGNAL_ARGS)4495 pgstat_exit(SIGNAL_ARGS)
4496 {
4497 int save_errno = errno;
4498
4499 need_exit = true;
4500 SetLatch(MyLatch);
4501
4502 errno = save_errno;
4503 }
4504
4505 /* SIGHUP handler for collector process */
4506 static void
pgstat_sighup_handler(SIGNAL_ARGS)4507 pgstat_sighup_handler(SIGNAL_ARGS)
4508 {
4509 int save_errno = errno;
4510
4511 got_SIGHUP = true;
4512 SetLatch(MyLatch);
4513
4514 errno = save_errno;
4515 }
4516
4517 /*
4518 * Subroutine to clear stats in a database entry
4519 *
4520 * Tables and functions hashes are initialized to empty.
4521 */
4522 static void
reset_dbentry_counters(PgStat_StatDBEntry * dbentry)4523 reset_dbentry_counters(PgStat_StatDBEntry *dbentry)
4524 {
4525 HASHCTL hash_ctl;
4526
4527 dbentry->n_xact_commit = 0;
4528 dbentry->n_xact_rollback = 0;
4529 dbentry->n_blocks_fetched = 0;
4530 dbentry->n_blocks_hit = 0;
4531 dbentry->n_tuples_returned = 0;
4532 dbentry->n_tuples_fetched = 0;
4533 dbentry->n_tuples_inserted = 0;
4534 dbentry->n_tuples_updated = 0;
4535 dbentry->n_tuples_deleted = 0;
4536 dbentry->last_autovac_time = 0;
4537 dbentry->n_conflict_tablespace = 0;
4538 dbentry->n_conflict_lock = 0;
4539 dbentry->n_conflict_snapshot = 0;
4540 dbentry->n_conflict_bufferpin = 0;
4541 dbentry->n_conflict_startup_deadlock = 0;
4542 dbentry->n_temp_files = 0;
4543 dbentry->n_temp_bytes = 0;
4544 dbentry->n_deadlocks = 0;
4545 dbentry->n_block_read_time = 0;
4546 dbentry->n_block_write_time = 0;
4547
4548 dbentry->stat_reset_timestamp = GetCurrentTimestamp();
4549 dbentry->stats_timestamp = 0;
4550
4551 memset(&hash_ctl, 0, sizeof(hash_ctl));
4552 hash_ctl.keysize = sizeof(Oid);
4553 hash_ctl.entrysize = sizeof(PgStat_StatTabEntry);
4554 dbentry->tables = hash_create("Per-database table",
4555 PGSTAT_TAB_HASH_SIZE,
4556 &hash_ctl,
4557 HASH_ELEM | HASH_BLOBS);
4558
4559 hash_ctl.keysize = sizeof(Oid);
4560 hash_ctl.entrysize = sizeof(PgStat_StatFuncEntry);
4561 dbentry->functions = hash_create("Per-database function",
4562 PGSTAT_FUNCTION_HASH_SIZE,
4563 &hash_ctl,
4564 HASH_ELEM | HASH_BLOBS);
4565 }
4566
4567 /*
4568 * Lookup the hash table entry for the specified database. If no hash
4569 * table entry exists, initialize it, if the create parameter is true.
4570 * Else, return NULL.
4571 */
4572 static PgStat_StatDBEntry *
pgstat_get_db_entry(Oid databaseid,bool create)4573 pgstat_get_db_entry(Oid databaseid, bool create)
4574 {
4575 PgStat_StatDBEntry *result;
4576 bool found;
4577 HASHACTION action = (create ? HASH_ENTER : HASH_FIND);
4578
4579 /* Lookup or create the hash table entry for this database */
4580 result = (PgStat_StatDBEntry *) hash_search(pgStatDBHash,
4581 &databaseid,
4582 action, &found);
4583
4584 if (!create && !found)
4585 return NULL;
4586
4587 /*
4588 * If not found, initialize the new one. This creates empty hash tables
4589 * for tables and functions, too.
4590 */
4591 if (!found)
4592 reset_dbentry_counters(result);
4593
4594 return result;
4595 }
4596
4597
4598 /*
4599 * Lookup the hash table entry for the specified table. If no hash
4600 * table entry exists, initialize it, if the create parameter is true.
4601 * Else, return NULL.
4602 */
4603 static PgStat_StatTabEntry *
pgstat_get_tab_entry(PgStat_StatDBEntry * dbentry,Oid tableoid,bool create)4604 pgstat_get_tab_entry(PgStat_StatDBEntry *dbentry, Oid tableoid, bool create)
4605 {
4606 PgStat_StatTabEntry *result;
4607 bool found;
4608 HASHACTION action = (create ? HASH_ENTER : HASH_FIND);
4609
4610 /* Lookup or create the hash table entry for this table */
4611 result = (PgStat_StatTabEntry *) hash_search(dbentry->tables,
4612 &tableoid,
4613 action, &found);
4614
4615 if (!create && !found)
4616 return NULL;
4617
4618 /* If not found, initialize the new one. */
4619 if (!found)
4620 {
4621 result->numscans = 0;
4622 result->tuples_returned = 0;
4623 result->tuples_fetched = 0;
4624 result->tuples_inserted = 0;
4625 result->tuples_updated = 0;
4626 result->tuples_deleted = 0;
4627 result->tuples_hot_updated = 0;
4628 result->n_live_tuples = 0;
4629 result->n_dead_tuples = 0;
4630 result->changes_since_analyze = 0;
4631 result->blocks_fetched = 0;
4632 result->blocks_hit = 0;
4633 result->vacuum_timestamp = 0;
4634 result->vacuum_count = 0;
4635 result->autovac_vacuum_timestamp = 0;
4636 result->autovac_vacuum_count = 0;
4637 result->analyze_timestamp = 0;
4638 result->analyze_count = 0;
4639 result->autovac_analyze_timestamp = 0;
4640 result->autovac_analyze_count = 0;
4641 }
4642
4643 return result;
4644 }
4645
4646
4647 /* ----------
4648 * pgstat_write_statsfiles() -
4649 * Write the global statistics file, as well as requested DB files.
4650 *
4651 * 'permanent' specifies writing to the permanent files not temporary ones.
4652 * When true (happens only when the collector is shutting down), also remove
4653 * the temporary files so that backends starting up under a new postmaster
4654 * can't read old data before the new collector is ready.
4655 *
4656 * When 'allDbs' is false, only the requested databases (listed in
4657 * pending_write_requests) will be written; otherwise, all databases
4658 * will be written.
4659 * ----------
4660 */
4661 static void
pgstat_write_statsfiles(bool permanent,bool allDbs)4662 pgstat_write_statsfiles(bool permanent, bool allDbs)
4663 {
4664 HASH_SEQ_STATUS hstat;
4665 PgStat_StatDBEntry *dbentry;
4666 FILE *fpout;
4667 int32 format_id;
4668 const char *tmpfile = permanent ? PGSTAT_STAT_PERMANENT_TMPFILE : pgstat_stat_tmpname;
4669 const char *statfile = permanent ? PGSTAT_STAT_PERMANENT_FILENAME : pgstat_stat_filename;
4670 int rc;
4671
4672 elog(DEBUG2, "writing stats file \"%s\"", statfile);
4673
4674 /*
4675 * Open the statistics temp file to write out the current values.
4676 */
4677 fpout = AllocateFile(tmpfile, PG_BINARY_W);
4678 if (fpout == NULL)
4679 {
4680 ereport(LOG,
4681 (errcode_for_file_access(),
4682 errmsg("could not open temporary statistics file \"%s\": %m",
4683 tmpfile)));
4684 return;
4685 }
4686
4687 /*
4688 * Set the timestamp of the stats file.
4689 */
4690 globalStats.stats_timestamp = GetCurrentTimestamp();
4691
4692 /*
4693 * Write the file header --- currently just a format ID.
4694 */
4695 format_id = PGSTAT_FILE_FORMAT_ID;
4696 rc = fwrite(&format_id, sizeof(format_id), 1, fpout);
4697 (void) rc; /* we'll check for error with ferror */
4698
4699 /*
4700 * Write global stats struct
4701 */
4702 rc = fwrite(&globalStats, sizeof(globalStats), 1, fpout);
4703 (void) rc; /* we'll check for error with ferror */
4704
4705 /*
4706 * Write archiver stats struct
4707 */
4708 rc = fwrite(&archiverStats, sizeof(archiverStats), 1, fpout);
4709 (void) rc; /* we'll check for error with ferror */
4710
4711 /*
4712 * Walk through the database table.
4713 */
4714 hash_seq_init(&hstat, pgStatDBHash);
4715 while ((dbentry = (PgStat_StatDBEntry *) hash_seq_search(&hstat)) != NULL)
4716 {
4717 /*
4718 * Write out the table and function stats for this DB into the
4719 * appropriate per-DB stat file, if required.
4720 */
4721 if (allDbs || pgstat_db_requested(dbentry->databaseid))
4722 {
4723 /* Make DB's timestamp consistent with the global stats */
4724 dbentry->stats_timestamp = globalStats.stats_timestamp;
4725
4726 pgstat_write_db_statsfile(dbentry, permanent);
4727 }
4728
4729 /*
4730 * Write out the DB entry. We don't write the tables or functions
4731 * pointers, since they're of no use to any other process.
4732 */
4733 fputc('D', fpout);
4734 rc = fwrite(dbentry, offsetof(PgStat_StatDBEntry, tables), 1, fpout);
4735 (void) rc; /* we'll check for error with ferror */
4736 }
4737
4738 /*
4739 * No more output to be done. Close the temp file and replace the old
4740 * pgstat.stat with it. The ferror() check replaces testing for error
4741 * after each individual fputc or fwrite above.
4742 */
4743 fputc('E', fpout);
4744
4745 if (ferror(fpout))
4746 {
4747 ereport(LOG,
4748 (errcode_for_file_access(),
4749 errmsg("could not write temporary statistics file \"%s\": %m",
4750 tmpfile)));
4751 FreeFile(fpout);
4752 unlink(tmpfile);
4753 }
4754 else if (FreeFile(fpout) < 0)
4755 {
4756 ereport(LOG,
4757 (errcode_for_file_access(),
4758 errmsg("could not close temporary statistics file \"%s\": %m",
4759 tmpfile)));
4760 unlink(tmpfile);
4761 }
4762 else if (rename(tmpfile, statfile) < 0)
4763 {
4764 ereport(LOG,
4765 (errcode_for_file_access(),
4766 errmsg("could not rename temporary statistics file \"%s\" to \"%s\": %m",
4767 tmpfile, statfile)));
4768 unlink(tmpfile);
4769 }
4770
4771 if (permanent)
4772 unlink(pgstat_stat_filename);
4773
4774 /*
4775 * Now throw away the list of requests. Note that requests sent after we
4776 * started the write are still waiting on the network socket.
4777 */
4778 list_free(pending_write_requests);
4779 pending_write_requests = NIL;
4780 }
4781
4782 /*
4783 * return the filename for a DB stat file; filename is the output buffer,
4784 * of length len.
4785 */
4786 static void
get_dbstat_filename(bool permanent,bool tempname,Oid databaseid,char * filename,int len)4787 get_dbstat_filename(bool permanent, bool tempname, Oid databaseid,
4788 char *filename, int len)
4789 {
4790 int printed;
4791
4792 /* NB -- pgstat_reset_remove_files knows about the pattern this uses */
4793 printed = snprintf(filename, len, "%s/db_%u.%s",
4794 permanent ? PGSTAT_STAT_PERMANENT_DIRECTORY :
4795 pgstat_stat_directory,
4796 databaseid,
4797 tempname ? "tmp" : "stat");
4798 if (printed >= len)
4799 elog(ERROR, "overlength pgstat path");
4800 }
4801
4802 /* ----------
4803 * pgstat_write_db_statsfile() -
4804 * Write the stat file for a single database.
4805 *
4806 * If writing to the permanent file (happens when the collector is
4807 * shutting down only), remove the temporary file so that backends
4808 * starting up under a new postmaster can't read the old data before
4809 * the new collector is ready.
4810 * ----------
4811 */
4812 static void
pgstat_write_db_statsfile(PgStat_StatDBEntry * dbentry,bool permanent)4813 pgstat_write_db_statsfile(PgStat_StatDBEntry *dbentry, bool permanent)
4814 {
4815 HASH_SEQ_STATUS tstat;
4816 HASH_SEQ_STATUS fstat;
4817 PgStat_StatTabEntry *tabentry;
4818 PgStat_StatFuncEntry *funcentry;
4819 FILE *fpout;
4820 int32 format_id;
4821 Oid dbid = dbentry->databaseid;
4822 int rc;
4823 char tmpfile[MAXPGPATH];
4824 char statfile[MAXPGPATH];
4825
4826 get_dbstat_filename(permanent, true, dbid, tmpfile, MAXPGPATH);
4827 get_dbstat_filename(permanent, false, dbid, statfile, MAXPGPATH);
4828
4829 elog(DEBUG2, "writing stats file \"%s\"", statfile);
4830
4831 /*
4832 * Open the statistics temp file to write out the current values.
4833 */
4834 fpout = AllocateFile(tmpfile, PG_BINARY_W);
4835 if (fpout == NULL)
4836 {
4837 ereport(LOG,
4838 (errcode_for_file_access(),
4839 errmsg("could not open temporary statistics file \"%s\": %m",
4840 tmpfile)));
4841 return;
4842 }
4843
4844 /*
4845 * Write the file header --- currently just a format ID.
4846 */
4847 format_id = PGSTAT_FILE_FORMAT_ID;
4848 rc = fwrite(&format_id, sizeof(format_id), 1, fpout);
4849 (void) rc; /* we'll check for error with ferror */
4850
4851 /*
4852 * Walk through the database's access stats per table.
4853 */
4854 hash_seq_init(&tstat, dbentry->tables);
4855 while ((tabentry = (PgStat_StatTabEntry *) hash_seq_search(&tstat)) != NULL)
4856 {
4857 fputc('T', fpout);
4858 rc = fwrite(tabentry, sizeof(PgStat_StatTabEntry), 1, fpout);
4859 (void) rc; /* we'll check for error with ferror */
4860 }
4861
4862 /*
4863 * Walk through the database's function stats table.
4864 */
4865 hash_seq_init(&fstat, dbentry->functions);
4866 while ((funcentry = (PgStat_StatFuncEntry *) hash_seq_search(&fstat)) != NULL)
4867 {
4868 fputc('F', fpout);
4869 rc = fwrite(funcentry, sizeof(PgStat_StatFuncEntry), 1, fpout);
4870 (void) rc; /* we'll check for error with ferror */
4871 }
4872
4873 /*
4874 * No more output to be done. Close the temp file and replace the old
4875 * pgstat.stat with it. The ferror() check replaces testing for error
4876 * after each individual fputc or fwrite above.
4877 */
4878 fputc('E', fpout);
4879
4880 if (ferror(fpout))
4881 {
4882 ereport(LOG,
4883 (errcode_for_file_access(),
4884 errmsg("could not write temporary statistics file \"%s\": %m",
4885 tmpfile)));
4886 FreeFile(fpout);
4887 unlink(tmpfile);
4888 }
4889 else if (FreeFile(fpout) < 0)
4890 {
4891 ereport(LOG,
4892 (errcode_for_file_access(),
4893 errmsg("could not close temporary statistics file \"%s\": %m",
4894 tmpfile)));
4895 unlink(tmpfile);
4896 }
4897 else if (rename(tmpfile, statfile) < 0)
4898 {
4899 ereport(LOG,
4900 (errcode_for_file_access(),
4901 errmsg("could not rename temporary statistics file \"%s\" to \"%s\": %m",
4902 tmpfile, statfile)));
4903 unlink(tmpfile);
4904 }
4905
4906 if (permanent)
4907 {
4908 get_dbstat_filename(false, false, dbid, statfile, MAXPGPATH);
4909
4910 elog(DEBUG2, "removing temporary stats file \"%s\"", statfile);
4911 unlink(statfile);
4912 }
4913 }
4914
4915 /* ----------
4916 * pgstat_read_statsfiles() -
4917 *
4918 * Reads in some existing statistics collector files and returns the
4919 * databases hash table that is the top level of the data.
4920 *
4921 * If 'onlydb' is not InvalidOid, it means we only want data for that DB
4922 * plus the shared catalogs ("DB 0"). We'll still populate the DB hash
4923 * table for all databases, but we don't bother even creating table/function
4924 * hash tables for other databases.
4925 *
4926 * 'permanent' specifies reading from the permanent files not temporary ones.
4927 * When true (happens only when the collector is starting up), remove the
4928 * files after reading; the in-memory status is now authoritative, and the
4929 * files would be out of date in case somebody else reads them.
4930 *
4931 * If a 'deep' read is requested, table/function stats are read, otherwise
4932 * the table/function hash tables remain empty.
4933 * ----------
4934 */
4935 static HTAB *
pgstat_read_statsfiles(Oid onlydb,bool permanent,bool deep)4936 pgstat_read_statsfiles(Oid onlydb, bool permanent, bool deep)
4937 {
4938 PgStat_StatDBEntry *dbentry;
4939 PgStat_StatDBEntry dbbuf;
4940 HASHCTL hash_ctl;
4941 HTAB *dbhash;
4942 FILE *fpin;
4943 int32 format_id;
4944 bool found;
4945 const char *statfile = permanent ? PGSTAT_STAT_PERMANENT_FILENAME : pgstat_stat_filename;
4946
4947 /*
4948 * The tables will live in pgStatLocalContext.
4949 */
4950 pgstat_setup_memcxt();
4951
4952 /*
4953 * Create the DB hashtable
4954 */
4955 memset(&hash_ctl, 0, sizeof(hash_ctl));
4956 hash_ctl.keysize = sizeof(Oid);
4957 hash_ctl.entrysize = sizeof(PgStat_StatDBEntry);
4958 hash_ctl.hcxt = pgStatLocalContext;
4959 dbhash = hash_create("Databases hash", PGSTAT_DB_HASH_SIZE, &hash_ctl,
4960 HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
4961
4962 /*
4963 * Clear out global and archiver statistics so they start from zero in
4964 * case we can't load an existing statsfile.
4965 */
4966 memset(&globalStats, 0, sizeof(globalStats));
4967 memset(&archiverStats, 0, sizeof(archiverStats));
4968
4969 /*
4970 * Set the current timestamp (will be kept only in case we can't load an
4971 * existing statsfile).
4972 */
4973 globalStats.stat_reset_timestamp = GetCurrentTimestamp();
4974 archiverStats.stat_reset_timestamp = globalStats.stat_reset_timestamp;
4975
4976 /*
4977 * Try to open the stats file. If it doesn't exist, the backends simply
4978 * return zero for anything and the collector simply starts from scratch
4979 * with empty counters.
4980 *
4981 * ENOENT is a possibility if the stats collector is not running or has
4982 * not yet written the stats file the first time. Any other failure
4983 * condition is suspicious.
4984 */
4985 if ((fpin = AllocateFile(statfile, PG_BINARY_R)) == NULL)
4986 {
4987 if (errno != ENOENT)
4988 ereport(pgStatRunningInCollector ? LOG : WARNING,
4989 (errcode_for_file_access(),
4990 errmsg("could not open statistics file \"%s\": %m",
4991 statfile)));
4992 return dbhash;
4993 }
4994
4995 /*
4996 * Verify it's of the expected format.
4997 */
4998 if (fread(&format_id, 1, sizeof(format_id), fpin) != sizeof(format_id) ||
4999 format_id != PGSTAT_FILE_FORMAT_ID)
5000 {
5001 ereport(pgStatRunningInCollector ? LOG : WARNING,
5002 (errmsg("corrupted statistics file \"%s\"", statfile)));
5003 goto done;
5004 }
5005
5006 /*
5007 * Read global stats struct
5008 */
5009 if (fread(&globalStats, 1, sizeof(globalStats), fpin) != sizeof(globalStats))
5010 {
5011 ereport(pgStatRunningInCollector ? LOG : WARNING,
5012 (errmsg("corrupted statistics file \"%s\"", statfile)));
5013 memset(&globalStats, 0, sizeof(globalStats));
5014 goto done;
5015 }
5016
5017 /*
5018 * In the collector, disregard the timestamp we read from the permanent
5019 * stats file; we should be willing to write a temp stats file immediately
5020 * upon the first request from any backend. This only matters if the old
5021 * file's timestamp is less than PGSTAT_STAT_INTERVAL ago, but that's not
5022 * an unusual scenario.
5023 */
5024 if (pgStatRunningInCollector)
5025 globalStats.stats_timestamp = 0;
5026
5027 /*
5028 * Read archiver stats struct
5029 */
5030 if (fread(&archiverStats, 1, sizeof(archiverStats), fpin) != sizeof(archiverStats))
5031 {
5032 ereport(pgStatRunningInCollector ? LOG : WARNING,
5033 (errmsg("corrupted statistics file \"%s\"", statfile)));
5034 memset(&archiverStats, 0, sizeof(archiverStats));
5035 goto done;
5036 }
5037
5038 /*
5039 * We found an existing collector stats file. Read it and put all the
5040 * hashtable entries into place.
5041 */
5042 for (;;)
5043 {
5044 switch (fgetc(fpin))
5045 {
5046 /*
5047 * 'D' A PgStat_StatDBEntry struct describing a database
5048 * follows.
5049 */
5050 case 'D':
5051 if (fread(&dbbuf, 1, offsetof(PgStat_StatDBEntry, tables),
5052 fpin) != offsetof(PgStat_StatDBEntry, tables))
5053 {
5054 ereport(pgStatRunningInCollector ? LOG : WARNING,
5055 (errmsg("corrupted statistics file \"%s\"",
5056 statfile)));
5057 goto done;
5058 }
5059
5060 /*
5061 * Add to the DB hash
5062 */
5063 dbentry = (PgStat_StatDBEntry *) hash_search(dbhash,
5064 (void *) &dbbuf.databaseid,
5065 HASH_ENTER,
5066 &found);
5067 if (found)
5068 {
5069 ereport(pgStatRunningInCollector ? LOG : WARNING,
5070 (errmsg("corrupted statistics file \"%s\"",
5071 statfile)));
5072 goto done;
5073 }
5074
5075 memcpy(dbentry, &dbbuf, sizeof(PgStat_StatDBEntry));
5076 dbentry->tables = NULL;
5077 dbentry->functions = NULL;
5078
5079 /*
5080 * In the collector, disregard the timestamp we read from the
5081 * permanent stats file; we should be willing to write a temp
5082 * stats file immediately upon the first request from any
5083 * backend.
5084 */
5085 if (pgStatRunningInCollector)
5086 dbentry->stats_timestamp = 0;
5087
5088 /*
5089 * Don't create tables/functions hashtables for uninteresting
5090 * databases.
5091 */
5092 if (onlydb != InvalidOid)
5093 {
5094 if (dbbuf.databaseid != onlydb &&
5095 dbbuf.databaseid != InvalidOid)
5096 break;
5097 }
5098
5099 memset(&hash_ctl, 0, sizeof(hash_ctl));
5100 hash_ctl.keysize = sizeof(Oid);
5101 hash_ctl.entrysize = sizeof(PgStat_StatTabEntry);
5102 hash_ctl.hcxt = pgStatLocalContext;
5103 dbentry->tables = hash_create("Per-database table",
5104 PGSTAT_TAB_HASH_SIZE,
5105 &hash_ctl,
5106 HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
5107
5108 hash_ctl.keysize = sizeof(Oid);
5109 hash_ctl.entrysize = sizeof(PgStat_StatFuncEntry);
5110 hash_ctl.hcxt = pgStatLocalContext;
5111 dbentry->functions = hash_create("Per-database function",
5112 PGSTAT_FUNCTION_HASH_SIZE,
5113 &hash_ctl,
5114 HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
5115
5116 /*
5117 * If requested, read the data from the database-specific
5118 * file. Otherwise we just leave the hashtables empty.
5119 */
5120 if (deep)
5121 pgstat_read_db_statsfile(dbentry->databaseid,
5122 dbentry->tables,
5123 dbentry->functions,
5124 permanent);
5125
5126 break;
5127
5128 case 'E':
5129 goto done;
5130
5131 default:
5132 ereport(pgStatRunningInCollector ? LOG : WARNING,
5133 (errmsg("corrupted statistics file \"%s\"",
5134 statfile)));
5135 goto done;
5136 }
5137 }
5138
5139 done:
5140 FreeFile(fpin);
5141
5142 /* If requested to read the permanent file, also get rid of it. */
5143 if (permanent)
5144 {
5145 elog(DEBUG2, "removing permanent stats file \"%s\"", statfile);
5146 unlink(statfile);
5147 }
5148
5149 return dbhash;
5150 }
5151
5152
5153 /* ----------
5154 * pgstat_read_db_statsfile() -
5155 *
5156 * Reads in the existing statistics collector file for the given database,
5157 * filling the passed-in tables and functions hash tables.
5158 *
5159 * As in pgstat_read_statsfiles, if the permanent file is requested, it is
5160 * removed after reading.
5161 *
5162 * Note: this code has the ability to skip storing per-table or per-function
5163 * data, if NULL is passed for the corresponding hashtable. That's not used
5164 * at the moment though.
5165 * ----------
5166 */
5167 static void
pgstat_read_db_statsfile(Oid databaseid,HTAB * tabhash,HTAB * funchash,bool permanent)5168 pgstat_read_db_statsfile(Oid databaseid, HTAB *tabhash, HTAB *funchash,
5169 bool permanent)
5170 {
5171 PgStat_StatTabEntry *tabentry;
5172 PgStat_StatTabEntry tabbuf;
5173 PgStat_StatFuncEntry funcbuf;
5174 PgStat_StatFuncEntry *funcentry;
5175 FILE *fpin;
5176 int32 format_id;
5177 bool found;
5178 char statfile[MAXPGPATH];
5179
5180 get_dbstat_filename(permanent, false, databaseid, statfile, MAXPGPATH);
5181
5182 /*
5183 * Try to open the stats file. If it doesn't exist, the backends simply
5184 * return zero for anything and the collector simply starts from scratch
5185 * with empty counters.
5186 *
5187 * ENOENT is a possibility if the stats collector is not running or has
5188 * not yet written the stats file the first time. Any other failure
5189 * condition is suspicious.
5190 */
5191 if ((fpin = AllocateFile(statfile, PG_BINARY_R)) == NULL)
5192 {
5193 if (errno != ENOENT)
5194 ereport(pgStatRunningInCollector ? LOG : WARNING,
5195 (errcode_for_file_access(),
5196 errmsg("could not open statistics file \"%s\": %m",
5197 statfile)));
5198 return;
5199 }
5200
5201 /*
5202 * Verify it's of the expected format.
5203 */
5204 if (fread(&format_id, 1, sizeof(format_id), fpin) != sizeof(format_id) ||
5205 format_id != PGSTAT_FILE_FORMAT_ID)
5206 {
5207 ereport(pgStatRunningInCollector ? LOG : WARNING,
5208 (errmsg("corrupted statistics file \"%s\"", statfile)));
5209 goto done;
5210 }
5211
5212 /*
5213 * We found an existing collector stats file. Read it and put all the
5214 * hashtable entries into place.
5215 */
5216 for (;;)
5217 {
5218 switch (fgetc(fpin))
5219 {
5220 /*
5221 * 'T' A PgStat_StatTabEntry follows.
5222 */
5223 case 'T':
5224 if (fread(&tabbuf, 1, sizeof(PgStat_StatTabEntry),
5225 fpin) != sizeof(PgStat_StatTabEntry))
5226 {
5227 ereport(pgStatRunningInCollector ? LOG : WARNING,
5228 (errmsg("corrupted statistics file \"%s\"",
5229 statfile)));
5230 goto done;
5231 }
5232
5233 /*
5234 * Skip if table data not wanted.
5235 */
5236 if (tabhash == NULL)
5237 break;
5238
5239 tabentry = (PgStat_StatTabEntry *) hash_search(tabhash,
5240 (void *) &tabbuf.tableid,
5241 HASH_ENTER, &found);
5242
5243 if (found)
5244 {
5245 ereport(pgStatRunningInCollector ? LOG : WARNING,
5246 (errmsg("corrupted statistics file \"%s\"",
5247 statfile)));
5248 goto done;
5249 }
5250
5251 memcpy(tabentry, &tabbuf, sizeof(tabbuf));
5252 break;
5253
5254 /*
5255 * 'F' A PgStat_StatFuncEntry follows.
5256 */
5257 case 'F':
5258 if (fread(&funcbuf, 1, sizeof(PgStat_StatFuncEntry),
5259 fpin) != sizeof(PgStat_StatFuncEntry))
5260 {
5261 ereport(pgStatRunningInCollector ? LOG : WARNING,
5262 (errmsg("corrupted statistics file \"%s\"",
5263 statfile)));
5264 goto done;
5265 }
5266
5267 /*
5268 * Skip if function data not wanted.
5269 */
5270 if (funchash == NULL)
5271 break;
5272
5273 funcentry = (PgStat_StatFuncEntry *) hash_search(funchash,
5274 (void *) &funcbuf.functionid,
5275 HASH_ENTER, &found);
5276
5277 if (found)
5278 {
5279 ereport(pgStatRunningInCollector ? LOG : WARNING,
5280 (errmsg("corrupted statistics file \"%s\"",
5281 statfile)));
5282 goto done;
5283 }
5284
5285 memcpy(funcentry, &funcbuf, sizeof(funcbuf));
5286 break;
5287
5288 /*
5289 * 'E' The EOF marker of a complete stats file.
5290 */
5291 case 'E':
5292 goto done;
5293
5294 default:
5295 ereport(pgStatRunningInCollector ? LOG : WARNING,
5296 (errmsg("corrupted statistics file \"%s\"",
5297 statfile)));
5298 goto done;
5299 }
5300 }
5301
5302 done:
5303 FreeFile(fpin);
5304
5305 if (permanent)
5306 {
5307 elog(DEBUG2, "removing permanent stats file \"%s\"", statfile);
5308 unlink(statfile);
5309 }
5310 }
5311
5312 /* ----------
5313 * pgstat_read_db_statsfile_timestamp() -
5314 *
5315 * Attempt to determine the timestamp of the last db statfile write.
5316 * Returns TRUE if successful; the timestamp is stored in *ts.
5317 *
5318 * This needs to be careful about handling databases for which no stats file
5319 * exists, such as databases without a stat entry or those not yet written:
5320 *
5321 * - if there's a database entry in the global file, return the corresponding
5322 * stats_timestamp value.
5323 *
5324 * - if there's no db stat entry (e.g. for a new or inactive database),
5325 * there's no stats_timestamp value, but also nothing to write so we return
5326 * the timestamp of the global statfile.
5327 * ----------
5328 */
5329 static bool
pgstat_read_db_statsfile_timestamp(Oid databaseid,bool permanent,TimestampTz * ts)5330 pgstat_read_db_statsfile_timestamp(Oid databaseid, bool permanent,
5331 TimestampTz *ts)
5332 {
5333 PgStat_StatDBEntry dbentry;
5334 PgStat_GlobalStats myGlobalStats;
5335 PgStat_ArchiverStats myArchiverStats;
5336 FILE *fpin;
5337 int32 format_id;
5338 const char *statfile = permanent ? PGSTAT_STAT_PERMANENT_FILENAME : pgstat_stat_filename;
5339
5340 /*
5341 * Try to open the stats file. As above, anything but ENOENT is worthy of
5342 * complaining about.
5343 */
5344 if ((fpin = AllocateFile(statfile, PG_BINARY_R)) == NULL)
5345 {
5346 if (errno != ENOENT)
5347 ereport(pgStatRunningInCollector ? LOG : WARNING,
5348 (errcode_for_file_access(),
5349 errmsg("could not open statistics file \"%s\": %m",
5350 statfile)));
5351 return false;
5352 }
5353
5354 /*
5355 * Verify it's of the expected format.
5356 */
5357 if (fread(&format_id, 1, sizeof(format_id), fpin) != sizeof(format_id) ||
5358 format_id != PGSTAT_FILE_FORMAT_ID)
5359 {
5360 ereport(pgStatRunningInCollector ? LOG : WARNING,
5361 (errmsg("corrupted statistics file \"%s\"", statfile)));
5362 FreeFile(fpin);
5363 return false;
5364 }
5365
5366 /*
5367 * Read global stats struct
5368 */
5369 if (fread(&myGlobalStats, 1, sizeof(myGlobalStats),
5370 fpin) != sizeof(myGlobalStats))
5371 {
5372 ereport(pgStatRunningInCollector ? LOG : WARNING,
5373 (errmsg("corrupted statistics file \"%s\"", statfile)));
5374 FreeFile(fpin);
5375 return false;
5376 }
5377
5378 /*
5379 * Read archiver stats struct
5380 */
5381 if (fread(&myArchiverStats, 1, sizeof(myArchiverStats),
5382 fpin) != sizeof(myArchiverStats))
5383 {
5384 ereport(pgStatRunningInCollector ? LOG : WARNING,
5385 (errmsg("corrupted statistics file \"%s\"", statfile)));
5386 FreeFile(fpin);
5387 return false;
5388 }
5389
5390 /* By default, we're going to return the timestamp of the global file. */
5391 *ts = myGlobalStats.stats_timestamp;
5392
5393 /*
5394 * We found an existing collector stats file. Read it and look for a
5395 * record for the requested database. If found, use its timestamp.
5396 */
5397 for (;;)
5398 {
5399 switch (fgetc(fpin))
5400 {
5401 /*
5402 * 'D' A PgStat_StatDBEntry struct describing a database
5403 * follows.
5404 */
5405 case 'D':
5406 if (fread(&dbentry, 1, offsetof(PgStat_StatDBEntry, tables),
5407 fpin) != offsetof(PgStat_StatDBEntry, tables))
5408 {
5409 ereport(pgStatRunningInCollector ? LOG : WARNING,
5410 (errmsg("corrupted statistics file \"%s\"",
5411 statfile)));
5412 goto done;
5413 }
5414
5415 /*
5416 * If this is the DB we're looking for, save its timestamp and
5417 * we're done.
5418 */
5419 if (dbentry.databaseid == databaseid)
5420 {
5421 *ts = dbentry.stats_timestamp;
5422 goto done;
5423 }
5424
5425 break;
5426
5427 case 'E':
5428 goto done;
5429
5430 default:
5431 ereport(pgStatRunningInCollector ? LOG : WARNING,
5432 (errmsg("corrupted statistics file \"%s\"",
5433 statfile)));
5434 goto done;
5435 }
5436 }
5437
5438 done:
5439 FreeFile(fpin);
5440 return true;
5441 }
5442
5443 /*
5444 * If not already done, read the statistics collector stats file into
5445 * some hash tables. The results will be kept until pgstat_clear_snapshot()
5446 * is called (typically, at end of transaction).
5447 */
5448 static void
backend_read_statsfile(void)5449 backend_read_statsfile(void)
5450 {
5451 TimestampTz min_ts = 0;
5452 TimestampTz ref_ts = 0;
5453 Oid inquiry_db;
5454 int count;
5455
5456 /* already read it? */
5457 if (pgStatDBHash)
5458 return;
5459 Assert(!pgStatRunningInCollector);
5460
5461 /*
5462 * In a normal backend, we check staleness of the data for our own DB, and
5463 * so we send MyDatabaseId in inquiry messages. In the autovac launcher,
5464 * check staleness of the shared-catalog data, and send InvalidOid in
5465 * inquiry messages so as not to force writing unnecessary data.
5466 */
5467 if (IsAutoVacuumLauncherProcess())
5468 inquiry_db = InvalidOid;
5469 else
5470 inquiry_db = MyDatabaseId;
5471
5472 /*
5473 * Loop until fresh enough stats file is available or we ran out of time.
5474 * The stats inquiry message is sent repeatedly in case collector drops
5475 * it; but not every single time, as that just swamps the collector.
5476 */
5477 for (count = 0; count < PGSTAT_POLL_LOOP_COUNT; count++)
5478 {
5479 bool ok;
5480 TimestampTz file_ts = 0;
5481 TimestampTz cur_ts;
5482
5483 CHECK_FOR_INTERRUPTS();
5484
5485 ok = pgstat_read_db_statsfile_timestamp(inquiry_db, false, &file_ts);
5486
5487 cur_ts = GetCurrentTimestamp();
5488 /* Calculate min acceptable timestamp, if we didn't already */
5489 if (count == 0 || cur_ts < ref_ts)
5490 {
5491 /*
5492 * We set the minimum acceptable timestamp to PGSTAT_STAT_INTERVAL
5493 * msec before now. This indirectly ensures that the collector
5494 * needn't write the file more often than PGSTAT_STAT_INTERVAL. In
5495 * an autovacuum worker, however, we want a lower delay to avoid
5496 * using stale data, so we use PGSTAT_RETRY_DELAY (since the
5497 * number of workers is low, this shouldn't be a problem).
5498 *
5499 * We don't recompute min_ts after sleeping, except in the
5500 * unlikely case that cur_ts went backwards. So we might end up
5501 * accepting a file a bit older than PGSTAT_STAT_INTERVAL. In
5502 * practice that shouldn't happen, though, as long as the sleep
5503 * time is less than PGSTAT_STAT_INTERVAL; and we don't want to
5504 * tell the collector that our cutoff time is less than what we'd
5505 * actually accept.
5506 */
5507 ref_ts = cur_ts;
5508 if (IsAutoVacuumWorkerProcess())
5509 min_ts = TimestampTzPlusMilliseconds(ref_ts,
5510 -PGSTAT_RETRY_DELAY);
5511 else
5512 min_ts = TimestampTzPlusMilliseconds(ref_ts,
5513 -PGSTAT_STAT_INTERVAL);
5514 }
5515
5516 /*
5517 * If the file timestamp is actually newer than cur_ts, we must have
5518 * had a clock glitch (system time went backwards) or there is clock
5519 * skew between our processor and the stats collector's processor.
5520 * Accept the file, but send an inquiry message anyway to make
5521 * pgstat_recv_inquiry do a sanity check on the collector's time.
5522 */
5523 if (ok && file_ts > cur_ts)
5524 {
5525 /*
5526 * A small amount of clock skew between processors isn't terribly
5527 * surprising, but a large difference is worth logging. We
5528 * arbitrarily define "large" as 1000 msec.
5529 */
5530 if (file_ts >= TimestampTzPlusMilliseconds(cur_ts, 1000))
5531 {
5532 char *filetime;
5533 char *mytime;
5534
5535 /* Copy because timestamptz_to_str returns a static buffer */
5536 filetime = pstrdup(timestamptz_to_str(file_ts));
5537 mytime = pstrdup(timestamptz_to_str(cur_ts));
5538 elog(LOG, "stats collector's time %s is later than backend local time %s",
5539 filetime, mytime);
5540 pfree(filetime);
5541 pfree(mytime);
5542 }
5543
5544 pgstat_send_inquiry(cur_ts, min_ts, inquiry_db);
5545 break;
5546 }
5547
5548 /* Normal acceptance case: file is not older than cutoff time */
5549 if (ok && file_ts >= min_ts)
5550 break;
5551
5552 /* Not there or too old, so kick the collector and wait a bit */
5553 if ((count % PGSTAT_INQ_LOOP_COUNT) == 0)
5554 pgstat_send_inquiry(cur_ts, min_ts, inquiry_db);
5555
5556 pg_usleep(PGSTAT_RETRY_DELAY * 1000L);
5557 }
5558
5559 if (count >= PGSTAT_POLL_LOOP_COUNT)
5560 ereport(LOG,
5561 (errmsg("using stale statistics instead of current ones "
5562 "because stats collector is not responding")));
5563
5564 /*
5565 * Autovacuum launcher wants stats about all databases, but a shallow read
5566 * is sufficient. Regular backends want a deep read for just the tables
5567 * they can see (MyDatabaseId + shared catalogs).
5568 */
5569 if (IsAutoVacuumLauncherProcess())
5570 pgStatDBHash = pgstat_read_statsfiles(InvalidOid, false, false);
5571 else
5572 pgStatDBHash = pgstat_read_statsfiles(MyDatabaseId, false, true);
5573 }
5574
5575
5576 /* ----------
5577 * pgstat_setup_memcxt() -
5578 *
5579 * Create pgStatLocalContext, if not already done.
5580 * ----------
5581 */
5582 static void
pgstat_setup_memcxt(void)5583 pgstat_setup_memcxt(void)
5584 {
5585 if (!pgStatLocalContext)
5586 pgStatLocalContext = AllocSetContextCreate(TopMemoryContext,
5587 "Statistics snapshot",
5588 ALLOCSET_SMALL_SIZES);
5589 }
5590
5591
5592 /* ----------
5593 * pgstat_clear_snapshot() -
5594 *
5595 * Discard any data collected in the current transaction. Any subsequent
5596 * request will cause new snapshots to be read.
5597 *
5598 * This is also invoked during transaction commit or abort to discard
5599 * the no-longer-wanted snapshot.
5600 * ----------
5601 */
5602 void
pgstat_clear_snapshot(void)5603 pgstat_clear_snapshot(void)
5604 {
5605 /* Release memory, if any was allocated */
5606 if (pgStatLocalContext)
5607 MemoryContextDelete(pgStatLocalContext);
5608
5609 /* Reset variables */
5610 pgStatLocalContext = NULL;
5611 pgStatDBHash = NULL;
5612 localBackendStatusTable = NULL;
5613 localNumBackends = 0;
5614 }
5615
5616
5617 /* ----------
5618 * pgstat_recv_inquiry() -
5619 *
5620 * Process stat inquiry requests.
5621 * ----------
5622 */
5623 static void
pgstat_recv_inquiry(PgStat_MsgInquiry * msg,int len)5624 pgstat_recv_inquiry(PgStat_MsgInquiry *msg, int len)
5625 {
5626 PgStat_StatDBEntry *dbentry;
5627
5628 elog(DEBUG2, "received inquiry for database %u", msg->databaseid);
5629
5630 /*
5631 * If there's already a write request for this DB, there's nothing to do.
5632 *
5633 * Note that if a request is found, we return early and skip the below
5634 * check for clock skew. This is okay, since the only way for a DB
5635 * request to be present in the list is that we have been here since the
5636 * last write round. It seems sufficient to check for clock skew once per
5637 * write round.
5638 */
5639 if (list_member_oid(pending_write_requests, msg->databaseid))
5640 return;
5641
5642 /*
5643 * Check to see if we last wrote this database at a time >= the requested
5644 * cutoff time. If so, this is a stale request that was generated before
5645 * we updated the DB file, and we don't need to do so again.
5646 *
5647 * If the requestor's local clock time is older than stats_timestamp, we
5648 * should suspect a clock glitch, ie system time going backwards; though
5649 * the more likely explanation is just delayed message receipt. It is
5650 * worth expending a GetCurrentTimestamp call to be sure, since a large
5651 * retreat in the system clock reading could otherwise cause us to neglect
5652 * to update the stats file for a long time.
5653 */
5654 dbentry = pgstat_get_db_entry(msg->databaseid, false);
5655 if (dbentry == NULL)
5656 {
5657 /*
5658 * We have no data for this DB. Enter a write request anyway so that
5659 * the global stats will get updated. This is needed to prevent
5660 * backend_read_statsfile from waiting for data that we cannot supply,
5661 * in the case of a new DB that nobody has yet reported any stats for.
5662 * See the behavior of pgstat_read_db_statsfile_timestamp.
5663 */
5664 }
5665 else if (msg->clock_time < dbentry->stats_timestamp)
5666 {
5667 TimestampTz cur_ts = GetCurrentTimestamp();
5668
5669 if (cur_ts < dbentry->stats_timestamp)
5670 {
5671 /*
5672 * Sure enough, time went backwards. Force a new stats file write
5673 * to get back in sync; but first, log a complaint.
5674 */
5675 char *writetime;
5676 char *mytime;
5677
5678 /* Copy because timestamptz_to_str returns a static buffer */
5679 writetime = pstrdup(timestamptz_to_str(dbentry->stats_timestamp));
5680 mytime = pstrdup(timestamptz_to_str(cur_ts));
5681 elog(LOG,
5682 "stats_timestamp %s is later than collector's time %s for database %u",
5683 writetime, mytime, dbentry->databaseid);
5684 pfree(writetime);
5685 pfree(mytime);
5686 }
5687 else
5688 {
5689 /*
5690 * Nope, it's just an old request. Assuming msg's clock_time is
5691 * >= its cutoff_time, it must be stale, so we can ignore it.
5692 */
5693 return;
5694 }
5695 }
5696 else if (msg->cutoff_time <= dbentry->stats_timestamp)
5697 {
5698 /* Stale request, ignore it */
5699 return;
5700 }
5701
5702 /*
5703 * We need to write this DB, so create a request.
5704 */
5705 pending_write_requests = lappend_oid(pending_write_requests,
5706 msg->databaseid);
5707 }
5708
5709
5710 /* ----------
5711 * pgstat_recv_tabstat() -
5712 *
5713 * Count what the backend has done.
5714 * ----------
5715 */
5716 static void
pgstat_recv_tabstat(PgStat_MsgTabstat * msg,int len)5717 pgstat_recv_tabstat(PgStat_MsgTabstat *msg, int len)
5718 {
5719 PgStat_StatDBEntry *dbentry;
5720 PgStat_StatTabEntry *tabentry;
5721 int i;
5722 bool found;
5723
5724 dbentry = pgstat_get_db_entry(msg->m_databaseid, true);
5725
5726 /*
5727 * Update database-wide stats.
5728 */
5729 dbentry->n_xact_commit += (PgStat_Counter) (msg->m_xact_commit);
5730 dbentry->n_xact_rollback += (PgStat_Counter) (msg->m_xact_rollback);
5731 dbentry->n_block_read_time += msg->m_block_read_time;
5732 dbentry->n_block_write_time += msg->m_block_write_time;
5733
5734 /*
5735 * Process all table entries in the message.
5736 */
5737 for (i = 0; i < msg->m_nentries; i++)
5738 {
5739 PgStat_TableEntry *tabmsg = &(msg->m_entry[i]);
5740
5741 tabentry = (PgStat_StatTabEntry *) hash_search(dbentry->tables,
5742 (void *) &(tabmsg->t_id),
5743 HASH_ENTER, &found);
5744
5745 if (!found)
5746 {
5747 /*
5748 * If it's a new table entry, initialize counters to the values we
5749 * just got.
5750 */
5751 tabentry->numscans = tabmsg->t_counts.t_numscans;
5752 tabentry->tuples_returned = tabmsg->t_counts.t_tuples_returned;
5753 tabentry->tuples_fetched = tabmsg->t_counts.t_tuples_fetched;
5754 tabentry->tuples_inserted = tabmsg->t_counts.t_tuples_inserted;
5755 tabentry->tuples_updated = tabmsg->t_counts.t_tuples_updated;
5756 tabentry->tuples_deleted = tabmsg->t_counts.t_tuples_deleted;
5757 tabentry->tuples_hot_updated = tabmsg->t_counts.t_tuples_hot_updated;
5758 tabentry->n_live_tuples = tabmsg->t_counts.t_delta_live_tuples;
5759 tabentry->n_dead_tuples = tabmsg->t_counts.t_delta_dead_tuples;
5760 tabentry->changes_since_analyze = tabmsg->t_counts.t_changed_tuples;
5761 tabentry->blocks_fetched = tabmsg->t_counts.t_blocks_fetched;
5762 tabentry->blocks_hit = tabmsg->t_counts.t_blocks_hit;
5763
5764 tabentry->vacuum_timestamp = 0;
5765 tabentry->vacuum_count = 0;
5766 tabentry->autovac_vacuum_timestamp = 0;
5767 tabentry->autovac_vacuum_count = 0;
5768 tabentry->analyze_timestamp = 0;
5769 tabentry->analyze_count = 0;
5770 tabentry->autovac_analyze_timestamp = 0;
5771 tabentry->autovac_analyze_count = 0;
5772 }
5773 else
5774 {
5775 /*
5776 * Otherwise add the values to the existing entry.
5777 */
5778 tabentry->numscans += tabmsg->t_counts.t_numscans;
5779 tabentry->tuples_returned += tabmsg->t_counts.t_tuples_returned;
5780 tabentry->tuples_fetched += tabmsg->t_counts.t_tuples_fetched;
5781 tabentry->tuples_inserted += tabmsg->t_counts.t_tuples_inserted;
5782 tabentry->tuples_updated += tabmsg->t_counts.t_tuples_updated;
5783 tabentry->tuples_deleted += tabmsg->t_counts.t_tuples_deleted;
5784 tabentry->tuples_hot_updated += tabmsg->t_counts.t_tuples_hot_updated;
5785 /* If table was truncated, first reset the live/dead counters */
5786 if (tabmsg->t_counts.t_truncated)
5787 {
5788 tabentry->n_live_tuples = 0;
5789 tabentry->n_dead_tuples = 0;
5790 }
5791 tabentry->n_live_tuples += tabmsg->t_counts.t_delta_live_tuples;
5792 tabentry->n_dead_tuples += tabmsg->t_counts.t_delta_dead_tuples;
5793 tabentry->changes_since_analyze += tabmsg->t_counts.t_changed_tuples;
5794 tabentry->blocks_fetched += tabmsg->t_counts.t_blocks_fetched;
5795 tabentry->blocks_hit += tabmsg->t_counts.t_blocks_hit;
5796 }
5797
5798 /* Clamp n_live_tuples in case of negative delta_live_tuples */
5799 tabentry->n_live_tuples = Max(tabentry->n_live_tuples, 0);
5800 /* Likewise for n_dead_tuples */
5801 tabentry->n_dead_tuples = Max(tabentry->n_dead_tuples, 0);
5802
5803 /*
5804 * Add per-table stats to the per-database entry, too.
5805 */
5806 dbentry->n_tuples_returned += tabmsg->t_counts.t_tuples_returned;
5807 dbentry->n_tuples_fetched += tabmsg->t_counts.t_tuples_fetched;
5808 dbentry->n_tuples_inserted += tabmsg->t_counts.t_tuples_inserted;
5809 dbentry->n_tuples_updated += tabmsg->t_counts.t_tuples_updated;
5810 dbentry->n_tuples_deleted += tabmsg->t_counts.t_tuples_deleted;
5811 dbentry->n_blocks_fetched += tabmsg->t_counts.t_blocks_fetched;
5812 dbentry->n_blocks_hit += tabmsg->t_counts.t_blocks_hit;
5813 }
5814 }
5815
5816
5817 /* ----------
5818 * pgstat_recv_tabpurge() -
5819 *
5820 * Arrange for dead table removal.
5821 * ----------
5822 */
5823 static void
pgstat_recv_tabpurge(PgStat_MsgTabpurge * msg,int len)5824 pgstat_recv_tabpurge(PgStat_MsgTabpurge *msg, int len)
5825 {
5826 PgStat_StatDBEntry *dbentry;
5827 int i;
5828
5829 dbentry = pgstat_get_db_entry(msg->m_databaseid, false);
5830
5831 /*
5832 * No need to purge if we don't even know the database.
5833 */
5834 if (!dbentry || !dbentry->tables)
5835 return;
5836
5837 /*
5838 * Process all table entries in the message.
5839 */
5840 for (i = 0; i < msg->m_nentries; i++)
5841 {
5842 /* Remove from hashtable if present; we don't care if it's not. */
5843 (void) hash_search(dbentry->tables,
5844 (void *) &(msg->m_tableid[i]),
5845 HASH_REMOVE, NULL);
5846 }
5847 }
5848
5849
5850 /* ----------
5851 * pgstat_recv_dropdb() -
5852 *
5853 * Arrange for dead database removal
5854 * ----------
5855 */
5856 static void
pgstat_recv_dropdb(PgStat_MsgDropdb * msg,int len)5857 pgstat_recv_dropdb(PgStat_MsgDropdb *msg, int len)
5858 {
5859 Oid dbid = msg->m_databaseid;
5860 PgStat_StatDBEntry *dbentry;
5861
5862 /*
5863 * Lookup the database in the hashtable.
5864 */
5865 dbentry = pgstat_get_db_entry(dbid, false);
5866
5867 /*
5868 * If found, remove it (along with the db statfile).
5869 */
5870 if (dbentry)
5871 {
5872 char statfile[MAXPGPATH];
5873
5874 get_dbstat_filename(false, false, dbid, statfile, MAXPGPATH);
5875
5876 elog(DEBUG2, "removing stats file \"%s\"", statfile);
5877 unlink(statfile);
5878
5879 if (dbentry->tables != NULL)
5880 hash_destroy(dbentry->tables);
5881 if (dbentry->functions != NULL)
5882 hash_destroy(dbentry->functions);
5883
5884 if (hash_search(pgStatDBHash,
5885 (void *) &dbid,
5886 HASH_REMOVE, NULL) == NULL)
5887 ereport(ERROR,
5888 (errmsg("database hash table corrupted during cleanup --- abort")));
5889 }
5890 }
5891
5892
5893 /* ----------
5894 * pgstat_recv_resetcounter() -
5895 *
5896 * Reset the statistics for the specified database.
5897 * ----------
5898 */
5899 static void
pgstat_recv_resetcounter(PgStat_MsgResetcounter * msg,int len)5900 pgstat_recv_resetcounter(PgStat_MsgResetcounter *msg, int len)
5901 {
5902 PgStat_StatDBEntry *dbentry;
5903
5904 /*
5905 * Lookup the database in the hashtable. Nothing to do if not there.
5906 */
5907 dbentry = pgstat_get_db_entry(msg->m_databaseid, false);
5908
5909 if (!dbentry)
5910 return;
5911
5912 /*
5913 * We simply throw away all the database's table entries by recreating a
5914 * new hash table for them.
5915 */
5916 if (dbentry->tables != NULL)
5917 hash_destroy(dbentry->tables);
5918 if (dbentry->functions != NULL)
5919 hash_destroy(dbentry->functions);
5920
5921 dbentry->tables = NULL;
5922 dbentry->functions = NULL;
5923
5924 /*
5925 * Reset database-level stats, too. This creates empty hash tables for
5926 * tables and functions.
5927 */
5928 reset_dbentry_counters(dbentry);
5929 }
5930
5931 /* ----------
5932 * pgstat_recv_resetshared() -
5933 *
5934 * Reset some shared statistics of the cluster.
5935 * ----------
5936 */
5937 static void
pgstat_recv_resetsharedcounter(PgStat_MsgResetsharedcounter * msg,int len)5938 pgstat_recv_resetsharedcounter(PgStat_MsgResetsharedcounter *msg, int len)
5939 {
5940 if (msg->m_resettarget == RESET_BGWRITER)
5941 {
5942 /* Reset the global background writer statistics for the cluster. */
5943 memset(&globalStats, 0, sizeof(globalStats));
5944 globalStats.stat_reset_timestamp = GetCurrentTimestamp();
5945 }
5946 else if (msg->m_resettarget == RESET_ARCHIVER)
5947 {
5948 /* Reset the archiver statistics for the cluster. */
5949 memset(&archiverStats, 0, sizeof(archiverStats));
5950 archiverStats.stat_reset_timestamp = GetCurrentTimestamp();
5951 }
5952
5953 /*
5954 * Presumably the sender of this message validated the target, don't
5955 * complain here if it's not valid
5956 */
5957 }
5958
5959 /* ----------
5960 * pgstat_recv_resetsinglecounter() -
5961 *
5962 * Reset a statistics for a single object
5963 * ----------
5964 */
5965 static void
pgstat_recv_resetsinglecounter(PgStat_MsgResetsinglecounter * msg,int len)5966 pgstat_recv_resetsinglecounter(PgStat_MsgResetsinglecounter *msg, int len)
5967 {
5968 PgStat_StatDBEntry *dbentry;
5969
5970 dbentry = pgstat_get_db_entry(msg->m_databaseid, false);
5971
5972 if (!dbentry)
5973 return;
5974
5975 /* Set the reset timestamp for the whole database */
5976 dbentry->stat_reset_timestamp = GetCurrentTimestamp();
5977
5978 /* Remove object if it exists, ignore it if not */
5979 if (msg->m_resettype == RESET_TABLE)
5980 (void) hash_search(dbentry->tables, (void *) &(msg->m_objectid),
5981 HASH_REMOVE, NULL);
5982 else if (msg->m_resettype == RESET_FUNCTION)
5983 (void) hash_search(dbentry->functions, (void *) &(msg->m_objectid),
5984 HASH_REMOVE, NULL);
5985 }
5986
5987 /* ----------
5988 * pgstat_recv_autovac() -
5989 *
5990 * Process an autovacuum signalling message.
5991 * ----------
5992 */
5993 static void
pgstat_recv_autovac(PgStat_MsgAutovacStart * msg,int len)5994 pgstat_recv_autovac(PgStat_MsgAutovacStart *msg, int len)
5995 {
5996 PgStat_StatDBEntry *dbentry;
5997
5998 /*
5999 * Store the last autovacuum time in the database's hashtable entry.
6000 */
6001 dbentry = pgstat_get_db_entry(msg->m_databaseid, true);
6002
6003 dbentry->last_autovac_time = msg->m_start_time;
6004 }
6005
6006 /* ----------
6007 * pgstat_recv_vacuum() -
6008 *
6009 * Process a VACUUM message.
6010 * ----------
6011 */
6012 static void
pgstat_recv_vacuum(PgStat_MsgVacuum * msg,int len)6013 pgstat_recv_vacuum(PgStat_MsgVacuum *msg, int len)
6014 {
6015 PgStat_StatDBEntry *dbentry;
6016 PgStat_StatTabEntry *tabentry;
6017
6018 /*
6019 * Store the data in the table's hashtable entry.
6020 */
6021 dbentry = pgstat_get_db_entry(msg->m_databaseid, true);
6022
6023 tabentry = pgstat_get_tab_entry(dbentry, msg->m_tableoid, true);
6024
6025 tabentry->n_live_tuples = msg->m_live_tuples;
6026 tabentry->n_dead_tuples = msg->m_dead_tuples;
6027
6028 if (msg->m_autovacuum)
6029 {
6030 tabentry->autovac_vacuum_timestamp = msg->m_vacuumtime;
6031 tabentry->autovac_vacuum_count++;
6032 }
6033 else
6034 {
6035 tabentry->vacuum_timestamp = msg->m_vacuumtime;
6036 tabentry->vacuum_count++;
6037 }
6038 }
6039
6040 /* ----------
6041 * pgstat_recv_analyze() -
6042 *
6043 * Process an ANALYZE message.
6044 * ----------
6045 */
6046 static void
pgstat_recv_analyze(PgStat_MsgAnalyze * msg,int len)6047 pgstat_recv_analyze(PgStat_MsgAnalyze *msg, int len)
6048 {
6049 PgStat_StatDBEntry *dbentry;
6050 PgStat_StatTabEntry *tabentry;
6051
6052 /*
6053 * Store the data in the table's hashtable entry.
6054 */
6055 dbentry = pgstat_get_db_entry(msg->m_databaseid, true);
6056
6057 tabentry = pgstat_get_tab_entry(dbentry, msg->m_tableoid, true);
6058
6059 tabentry->n_live_tuples = msg->m_live_tuples;
6060 tabentry->n_dead_tuples = msg->m_dead_tuples;
6061
6062 /*
6063 * If commanded, reset changes_since_analyze to zero. This forgets any
6064 * changes that were committed while the ANALYZE was in progress, but we
6065 * have no good way to estimate how many of those there were.
6066 */
6067 if (msg->m_resetcounter)
6068 tabentry->changes_since_analyze = 0;
6069
6070 if (msg->m_autovacuum)
6071 {
6072 tabentry->autovac_analyze_timestamp = msg->m_analyzetime;
6073 tabentry->autovac_analyze_count++;
6074 }
6075 else
6076 {
6077 tabentry->analyze_timestamp = msg->m_analyzetime;
6078 tabentry->analyze_count++;
6079 }
6080 }
6081
6082
6083 /* ----------
6084 * pgstat_recv_archiver() -
6085 *
6086 * Process a ARCHIVER message.
6087 * ----------
6088 */
6089 static void
pgstat_recv_archiver(PgStat_MsgArchiver * msg,int len)6090 pgstat_recv_archiver(PgStat_MsgArchiver *msg, int len)
6091 {
6092 if (msg->m_failed)
6093 {
6094 /* Failed archival attempt */
6095 ++archiverStats.failed_count;
6096 memcpy(archiverStats.last_failed_wal, msg->m_xlog,
6097 sizeof(archiverStats.last_failed_wal));
6098 archiverStats.last_failed_timestamp = msg->m_timestamp;
6099 }
6100 else
6101 {
6102 /* Successful archival operation */
6103 ++archiverStats.archived_count;
6104 memcpy(archiverStats.last_archived_wal, msg->m_xlog,
6105 sizeof(archiverStats.last_archived_wal));
6106 archiverStats.last_archived_timestamp = msg->m_timestamp;
6107 }
6108 }
6109
6110 /* ----------
6111 * pgstat_recv_bgwriter() -
6112 *
6113 * Process a BGWRITER message.
6114 * ----------
6115 */
6116 static void
pgstat_recv_bgwriter(PgStat_MsgBgWriter * msg,int len)6117 pgstat_recv_bgwriter(PgStat_MsgBgWriter *msg, int len)
6118 {
6119 globalStats.timed_checkpoints += msg->m_timed_checkpoints;
6120 globalStats.requested_checkpoints += msg->m_requested_checkpoints;
6121 globalStats.checkpoint_write_time += msg->m_checkpoint_write_time;
6122 globalStats.checkpoint_sync_time += msg->m_checkpoint_sync_time;
6123 globalStats.buf_written_checkpoints += msg->m_buf_written_checkpoints;
6124 globalStats.buf_written_clean += msg->m_buf_written_clean;
6125 globalStats.maxwritten_clean += msg->m_maxwritten_clean;
6126 globalStats.buf_written_backend += msg->m_buf_written_backend;
6127 globalStats.buf_fsync_backend += msg->m_buf_fsync_backend;
6128 globalStats.buf_alloc += msg->m_buf_alloc;
6129 }
6130
6131 /* ----------
6132 * pgstat_recv_recoveryconflict() -
6133 *
6134 * Process a RECOVERYCONFLICT message.
6135 * ----------
6136 */
6137 static void
pgstat_recv_recoveryconflict(PgStat_MsgRecoveryConflict * msg,int len)6138 pgstat_recv_recoveryconflict(PgStat_MsgRecoveryConflict *msg, int len)
6139 {
6140 PgStat_StatDBEntry *dbentry;
6141
6142 dbentry = pgstat_get_db_entry(msg->m_databaseid, true);
6143
6144 switch (msg->m_reason)
6145 {
6146 case PROCSIG_RECOVERY_CONFLICT_DATABASE:
6147
6148 /*
6149 * Since we drop the information about the database as soon as it
6150 * replicates, there is no point in counting these conflicts.
6151 */
6152 break;
6153 case PROCSIG_RECOVERY_CONFLICT_TABLESPACE:
6154 dbentry->n_conflict_tablespace++;
6155 break;
6156 case PROCSIG_RECOVERY_CONFLICT_LOCK:
6157 dbentry->n_conflict_lock++;
6158 break;
6159 case PROCSIG_RECOVERY_CONFLICT_SNAPSHOT:
6160 dbentry->n_conflict_snapshot++;
6161 break;
6162 case PROCSIG_RECOVERY_CONFLICT_BUFFERPIN:
6163 dbentry->n_conflict_bufferpin++;
6164 break;
6165 case PROCSIG_RECOVERY_CONFLICT_STARTUP_DEADLOCK:
6166 dbentry->n_conflict_startup_deadlock++;
6167 break;
6168 }
6169 }
6170
6171 /* ----------
6172 * pgstat_recv_deadlock() -
6173 *
6174 * Process a DEADLOCK message.
6175 * ----------
6176 */
6177 static void
pgstat_recv_deadlock(PgStat_MsgDeadlock * msg,int len)6178 pgstat_recv_deadlock(PgStat_MsgDeadlock *msg, int len)
6179 {
6180 PgStat_StatDBEntry *dbentry;
6181
6182 dbentry = pgstat_get_db_entry(msg->m_databaseid, true);
6183
6184 dbentry->n_deadlocks++;
6185 }
6186
6187 /* ----------
6188 * pgstat_recv_tempfile() -
6189 *
6190 * Process a TEMPFILE message.
6191 * ----------
6192 */
6193 static void
pgstat_recv_tempfile(PgStat_MsgTempFile * msg,int len)6194 pgstat_recv_tempfile(PgStat_MsgTempFile *msg, int len)
6195 {
6196 PgStat_StatDBEntry *dbentry;
6197
6198 dbentry = pgstat_get_db_entry(msg->m_databaseid, true);
6199
6200 dbentry->n_temp_bytes += msg->m_filesize;
6201 dbentry->n_temp_files += 1;
6202 }
6203
6204 /* ----------
6205 * pgstat_recv_funcstat() -
6206 *
6207 * Count what the backend has done.
6208 * ----------
6209 */
6210 static void
pgstat_recv_funcstat(PgStat_MsgFuncstat * msg,int len)6211 pgstat_recv_funcstat(PgStat_MsgFuncstat *msg, int len)
6212 {
6213 PgStat_FunctionEntry *funcmsg = &(msg->m_entry[0]);
6214 PgStat_StatDBEntry *dbentry;
6215 PgStat_StatFuncEntry *funcentry;
6216 int i;
6217 bool found;
6218
6219 dbentry = pgstat_get_db_entry(msg->m_databaseid, true);
6220
6221 /*
6222 * Process all function entries in the message.
6223 */
6224 for (i = 0; i < msg->m_nentries; i++, funcmsg++)
6225 {
6226 funcentry = (PgStat_StatFuncEntry *) hash_search(dbentry->functions,
6227 (void *) &(funcmsg->f_id),
6228 HASH_ENTER, &found);
6229
6230 if (!found)
6231 {
6232 /*
6233 * If it's a new function entry, initialize counters to the values
6234 * we just got.
6235 */
6236 funcentry->f_numcalls = funcmsg->f_numcalls;
6237 funcentry->f_total_time = funcmsg->f_total_time;
6238 funcentry->f_self_time = funcmsg->f_self_time;
6239 }
6240 else
6241 {
6242 /*
6243 * Otherwise add the values to the existing entry.
6244 */
6245 funcentry->f_numcalls += funcmsg->f_numcalls;
6246 funcentry->f_total_time += funcmsg->f_total_time;
6247 funcentry->f_self_time += funcmsg->f_self_time;
6248 }
6249 }
6250 }
6251
6252 /* ----------
6253 * pgstat_recv_funcpurge() -
6254 *
6255 * Arrange for dead function removal.
6256 * ----------
6257 */
6258 static void
pgstat_recv_funcpurge(PgStat_MsgFuncpurge * msg,int len)6259 pgstat_recv_funcpurge(PgStat_MsgFuncpurge *msg, int len)
6260 {
6261 PgStat_StatDBEntry *dbentry;
6262 int i;
6263
6264 dbentry = pgstat_get_db_entry(msg->m_databaseid, false);
6265
6266 /*
6267 * No need to purge if we don't even know the database.
6268 */
6269 if (!dbentry || !dbentry->functions)
6270 return;
6271
6272 /*
6273 * Process all function entries in the message.
6274 */
6275 for (i = 0; i < msg->m_nentries; i++)
6276 {
6277 /* Remove from hashtable if present; we don't care if it's not. */
6278 (void) hash_search(dbentry->functions,
6279 (void *) &(msg->m_functionid[i]),
6280 HASH_REMOVE, NULL);
6281 }
6282 }
6283
6284 /* ----------
6285 * pgstat_write_statsfile_needed() -
6286 *
6287 * Do we need to write out any stats files?
6288 * ----------
6289 */
6290 static bool
pgstat_write_statsfile_needed(void)6291 pgstat_write_statsfile_needed(void)
6292 {
6293 if (pending_write_requests != NIL)
6294 return true;
6295
6296 /* Everything was written recently */
6297 return false;
6298 }
6299
6300 /* ----------
6301 * pgstat_db_requested() -
6302 *
6303 * Checks whether stats for a particular DB need to be written to a file.
6304 * ----------
6305 */
6306 static bool
pgstat_db_requested(Oid databaseid)6307 pgstat_db_requested(Oid databaseid)
6308 {
6309 /*
6310 * If any requests are outstanding at all, we should write the stats for
6311 * shared catalogs (the "database" with OID 0). This ensures that
6312 * backends will see up-to-date stats for shared catalogs, even though
6313 * they send inquiry messages mentioning only their own DB.
6314 */
6315 if (databaseid == InvalidOid && pending_write_requests != NIL)
6316 return true;
6317
6318 /* Search to see if there's an open request to write this database. */
6319 if (list_member_oid(pending_write_requests, databaseid))
6320 return true;
6321
6322 return false;
6323 }
6324