| 1 | /* ---------- |
| 2 | * pgstat.h |
| 3 | * |
| 4 | * Definitions for the PostgreSQL statistics collector daemon. |
| 5 | * |
| 6 | * Copyright (c) 2001-2019, PostgreSQL Global Development Group |
| 7 | * |
| 8 | * src/include/pgstat.h |
| 9 | * ---------- |
| 10 | */ |
| 11 | #ifndef PGSTAT_H |
| 12 | #define PGSTAT_H |
| 13 | |
| 14 | #include "datatype/timestamp.h" |
| 15 | #include "fmgr.h" |
| 16 | #include "libpq/pqcomm.h" |
| 17 | #include "port/atomics.h" |
| 18 | #include "portability/instr_time.h" |
| 19 | #include "postmaster/pgarch.h" |
| 20 | #include "storage/proc.h" |
| 21 | #include "utils/hsearch.h" |
| 22 | #include "utils/relcache.h" |
| 23 | |
| 24 | |
| 25 | /* ---------- |
| 26 | * Paths for the statistics files (relative to installation's $PGDATA). |
| 27 | * ---------- |
| 28 | */ |
| 29 | #define PGSTAT_STAT_PERMANENT_DIRECTORY "pg_stat" |
| 30 | #define PGSTAT_STAT_PERMANENT_FILENAME "pg_stat/global.stat" |
| 31 | #define PGSTAT_STAT_PERMANENT_TMPFILE "pg_stat/global.tmp" |
| 32 | |
| 33 | /* Default directory to store temporary statistics data in */ |
| 34 | #define PG_STAT_TMP_DIR "pg_stat_tmp" |
| 35 | |
| 36 | /* Values for track_functions GUC variable --- order is significant! */ |
| 37 | typedef enum TrackFunctionsLevel |
| 38 | { |
| 39 | TRACK_FUNC_OFF, |
| 40 | TRACK_FUNC_PL, |
| 41 | TRACK_FUNC_ALL |
| 42 | } TrackFunctionsLevel; |
| 43 | |
| 44 | /* ---------- |
| 45 | * The types of backend -> collector messages |
| 46 | * ---------- |
| 47 | */ |
| 48 | typedef enum StatMsgType |
| 49 | { |
| 50 | PGSTAT_MTYPE_DUMMY, |
| 51 | PGSTAT_MTYPE_INQUIRY, |
| 52 | PGSTAT_MTYPE_TABSTAT, |
| 53 | PGSTAT_MTYPE_TABPURGE, |
| 54 | PGSTAT_MTYPE_DROPDB, |
| 55 | PGSTAT_MTYPE_RESETCOUNTER, |
| 56 | PGSTAT_MTYPE_RESETSHAREDCOUNTER, |
| 57 | PGSTAT_MTYPE_RESETSINGLECOUNTER, |
| 58 | PGSTAT_MTYPE_AUTOVAC_START, |
| 59 | PGSTAT_MTYPE_VACUUM, |
| 60 | PGSTAT_MTYPE_ANALYZE, |
| 61 | PGSTAT_MTYPE_ARCHIVER, |
| 62 | PGSTAT_MTYPE_BGWRITER, |
| 63 | PGSTAT_MTYPE_FUNCSTAT, |
| 64 | PGSTAT_MTYPE_FUNCPURGE, |
| 65 | PGSTAT_MTYPE_RECOVERYCONFLICT, |
| 66 | PGSTAT_MTYPE_TEMPFILE, |
| 67 | PGSTAT_MTYPE_DEADLOCK, |
| 68 | PGSTAT_MTYPE_CHECKSUMFAILURE |
| 69 | } StatMsgType; |
| 70 | |
| 71 | /* ---------- |
| 72 | * The data type used for counters. |
| 73 | * ---------- |
| 74 | */ |
| 75 | typedef int64 PgStat_Counter; |
| 76 | |
| 77 | /* ---------- |
| 78 | * PgStat_TableCounts The actual per-table counts kept by a backend |
| 79 | * |
| 80 | * This struct should contain only actual event counters, because we memcmp |
| 81 | * it against zeroes to detect whether there are any counts to transmit. |
| 82 | * It is a component of PgStat_TableStatus (within-backend state) and |
| 83 | * PgStat_TableEntry (the transmitted message format). |
| 84 | * |
| 85 | * Note: for a table, tuples_returned is the number of tuples successfully |
| 86 | * fetched by heap_getnext, while tuples_fetched is the number of tuples |
| 87 | * successfully fetched by heap_fetch under the control of bitmap indexscans. |
| 88 | * For an index, tuples_returned is the number of index entries returned by |
| 89 | * the index AM, while tuples_fetched is the number of tuples successfully |
| 90 | * fetched by heap_fetch under the control of simple indexscans for this index. |
| 91 | * |
| 92 | * tuples_inserted/updated/deleted/hot_updated count attempted actions, |
| 93 | * regardless of whether the transaction committed. delta_live_tuples, |
| 94 | * delta_dead_tuples, and changed_tuples are set depending on commit or abort. |
| 95 | * Note that delta_live_tuples and delta_dead_tuples can be negative! |
| 96 | * ---------- |
| 97 | */ |
| 98 | typedef struct PgStat_TableCounts |
| 99 | { |
| 100 | PgStat_Counter t_numscans; |
| 101 | |
| 102 | PgStat_Counter t_tuples_returned; |
| 103 | PgStat_Counter t_tuples_fetched; |
| 104 | |
| 105 | PgStat_Counter t_tuples_inserted; |
| 106 | PgStat_Counter t_tuples_updated; |
| 107 | PgStat_Counter t_tuples_deleted; |
| 108 | PgStat_Counter t_tuples_hot_updated; |
| 109 | bool t_truncated; |
| 110 | |
| 111 | PgStat_Counter t_delta_live_tuples; |
| 112 | PgStat_Counter t_delta_dead_tuples; |
| 113 | PgStat_Counter t_changed_tuples; |
| 114 | |
| 115 | PgStat_Counter t_blocks_fetched; |
| 116 | PgStat_Counter t_blocks_hit; |
| 117 | } PgStat_TableCounts; |
| 118 | |
| 119 | /* Possible targets for resetting cluster-wide shared values */ |
| 120 | typedef enum PgStat_Shared_Reset_Target |
| 121 | { |
| 122 | RESET_ARCHIVER, |
| 123 | RESET_BGWRITER |
| 124 | } PgStat_Shared_Reset_Target; |
| 125 | |
| 126 | /* Possible object types for resetting single counters */ |
| 127 | typedef enum PgStat_Single_Reset_Type |
| 128 | { |
| 129 | RESET_TABLE, |
| 130 | RESET_FUNCTION |
| 131 | } PgStat_Single_Reset_Type; |
| 132 | |
| 133 | /* ------------------------------------------------------------ |
| 134 | * Structures kept in backend local memory while accumulating counts |
| 135 | * ------------------------------------------------------------ |
| 136 | */ |
| 137 | |
| 138 | |
| 139 | /* ---------- |
| 140 | * PgStat_TableStatus Per-table status within a backend |
| 141 | * |
| 142 | * Many of the event counters are nontransactional, ie, we count events |
| 143 | * in committed and aborted transactions alike. For these, we just count |
| 144 | * directly in the PgStat_TableStatus. However, delta_live_tuples, |
| 145 | * delta_dead_tuples, and changed_tuples must be derived from event counts |
| 146 | * with awareness of whether the transaction or subtransaction committed or |
| 147 | * aborted. Hence, we also keep a stack of per-(sub)transaction status |
| 148 | * records for every table modified in the current transaction. At commit |
| 149 | * or abort, we propagate tuples_inserted/updated/deleted up to the |
| 150 | * parent subtransaction level, or out to the parent PgStat_TableStatus, |
| 151 | * as appropriate. |
| 152 | * ---------- |
| 153 | */ |
| 154 | typedef struct PgStat_TableStatus |
| 155 | { |
| 156 | Oid t_id; /* table's OID */ |
| 157 | bool t_shared; /* is it a shared catalog? */ |
| 158 | struct PgStat_TableXactStatus *trans; /* lowest subxact's counts */ |
| 159 | PgStat_TableCounts t_counts; /* event counts to be sent */ |
| 160 | } PgStat_TableStatus; |
| 161 | |
| 162 | /* ---------- |
| 163 | * PgStat_TableXactStatus Per-table, per-subtransaction status |
| 164 | * ---------- |
| 165 | */ |
| 166 | typedef struct PgStat_TableXactStatus |
| 167 | { |
| 168 | PgStat_Counter tuples_inserted; /* tuples inserted in (sub)xact */ |
| 169 | PgStat_Counter tuples_updated; /* tuples updated in (sub)xact */ |
| 170 | PgStat_Counter tuples_deleted; /* tuples deleted in (sub)xact */ |
| 171 | bool truncated; /* relation truncated in this (sub)xact */ |
| 172 | PgStat_Counter inserted_pre_trunc; /* tuples inserted prior to truncate */ |
| 173 | PgStat_Counter updated_pre_trunc; /* tuples updated prior to truncate */ |
| 174 | PgStat_Counter deleted_pre_trunc; /* tuples deleted prior to truncate */ |
| 175 | int nest_level; /* subtransaction nest level */ |
| 176 | /* links to other structs for same relation: */ |
| 177 | struct PgStat_TableXactStatus *upper; /* next higher subxact if any */ |
| 178 | PgStat_TableStatus *parent; /* per-table status */ |
| 179 | /* structs of same subxact level are linked here: */ |
| 180 | struct PgStat_TableXactStatus *next; /* next of same subxact */ |
| 181 | } PgStat_TableXactStatus; |
| 182 | |
| 183 | |
| 184 | /* ------------------------------------------------------------ |
| 185 | * Message formats follow |
| 186 | * ------------------------------------------------------------ |
| 187 | */ |
| 188 | |
| 189 | |
| 190 | /* ---------- |
| 191 | * PgStat_MsgHdr The common message header |
| 192 | * ---------- |
| 193 | */ |
| 194 | typedef struct PgStat_MsgHdr |
| 195 | { |
| 196 | StatMsgType m_type; |
| 197 | int m_size; |
| 198 | } PgStat_MsgHdr; |
| 199 | |
| 200 | /* ---------- |
| 201 | * Space available in a message. This will keep the UDP packets below 1K, |
| 202 | * which should fit unfragmented into the MTU of the loopback interface. |
| 203 | * (Larger values of PGSTAT_MAX_MSG_SIZE would work for that on most |
| 204 | * platforms, but we're being conservative here.) |
| 205 | * ---------- |
| 206 | */ |
| 207 | #define PGSTAT_MAX_MSG_SIZE 1000 |
| 208 | #define PGSTAT_MSG_PAYLOAD (PGSTAT_MAX_MSG_SIZE - sizeof(PgStat_MsgHdr)) |
| 209 | |
| 210 | |
| 211 | /* ---------- |
| 212 | * PgStat_MsgDummy A dummy message, ignored by the collector |
| 213 | * ---------- |
| 214 | */ |
| 215 | typedef struct PgStat_MsgDummy |
| 216 | { |
| 217 | PgStat_MsgHdr m_hdr; |
| 218 | } PgStat_MsgDummy; |
| 219 | |
| 220 | |
| 221 | /* ---------- |
| 222 | * PgStat_MsgInquiry Sent by a backend to ask the collector |
| 223 | * to write the stats file(s). |
| 224 | * |
| 225 | * Ordinarily, an inquiry message prompts writing of the global stats file, |
| 226 | * the stats file for shared catalogs, and the stats file for the specified |
| 227 | * database. If databaseid is InvalidOid, only the first two are written. |
| 228 | * |
| 229 | * New file(s) will be written only if the existing file has a timestamp |
| 230 | * older than the specified cutoff_time; this prevents duplicated effort |
| 231 | * when multiple requests arrive at nearly the same time, assuming that |
| 232 | * backends send requests with cutoff_times a little bit in the past. |
| 233 | * |
| 234 | * clock_time should be the requestor's current local time; the collector |
| 235 | * uses this to check for the system clock going backward, but it has no |
| 236 | * effect unless that occurs. We assume clock_time >= cutoff_time, though. |
| 237 | * ---------- |
| 238 | */ |
| 239 | |
| 240 | typedef struct PgStat_MsgInquiry |
| 241 | { |
| 242 | PgStat_MsgHdr m_hdr; |
| 243 | TimestampTz clock_time; /* observed local clock time */ |
| 244 | TimestampTz cutoff_time; /* minimum acceptable file timestamp */ |
| 245 | Oid databaseid; /* requested DB (InvalidOid => shared only) */ |
| 246 | } PgStat_MsgInquiry; |
| 247 | |
| 248 | |
| 249 | /* ---------- |
| 250 | * PgStat_TableEntry Per-table info in a MsgTabstat |
| 251 | * ---------- |
| 252 | */ |
| 253 | typedef struct PgStat_TableEntry |
| 254 | { |
| 255 | Oid t_id; |
| 256 | PgStat_TableCounts t_counts; |
| 257 | } PgStat_TableEntry; |
| 258 | |
| 259 | /* ---------- |
| 260 | * PgStat_MsgTabstat Sent by the backend to report table |
| 261 | * and buffer access statistics. |
| 262 | * ---------- |
| 263 | */ |
| 264 | #define PGSTAT_NUM_TABENTRIES \ |
| 265 | ((PGSTAT_MSG_PAYLOAD - sizeof(Oid) - 3 * sizeof(int) - 2 * sizeof(PgStat_Counter)) \ |
| 266 | / sizeof(PgStat_TableEntry)) |
| 267 | |
| 268 | typedef struct PgStat_MsgTabstat |
| 269 | { |
| 270 | PgStat_MsgHdr m_hdr; |
| 271 | Oid m_databaseid; |
| 272 | int m_nentries; |
| 273 | int m_xact_commit; |
| 274 | int m_xact_rollback; |
| 275 | PgStat_Counter m_block_read_time; /* times in microseconds */ |
| 276 | PgStat_Counter m_block_write_time; |
| 277 | PgStat_TableEntry m_entry[PGSTAT_NUM_TABENTRIES]; |
| 278 | } PgStat_MsgTabstat; |
| 279 | |
| 280 | |
| 281 | /* ---------- |
| 282 | * PgStat_MsgTabpurge Sent by the backend to tell the collector |
| 283 | * about dead tables. |
| 284 | * ---------- |
| 285 | */ |
| 286 | #define PGSTAT_NUM_TABPURGE \ |
| 287 | ((PGSTAT_MSG_PAYLOAD - sizeof(Oid) - sizeof(int)) \ |
| 288 | / sizeof(Oid)) |
| 289 | |
| 290 | typedef struct PgStat_MsgTabpurge |
| 291 | { |
| 292 | PgStat_MsgHdr m_hdr; |
| 293 | Oid m_databaseid; |
| 294 | int m_nentries; |
| 295 | Oid m_tableid[PGSTAT_NUM_TABPURGE]; |
| 296 | } PgStat_MsgTabpurge; |
| 297 | |
| 298 | |
| 299 | /* ---------- |
| 300 | * PgStat_MsgDropdb Sent by the backend to tell the collector |
| 301 | * about a dropped database |
| 302 | * ---------- |
| 303 | */ |
| 304 | typedef struct PgStat_MsgDropdb |
| 305 | { |
| 306 | PgStat_MsgHdr m_hdr; |
| 307 | Oid m_databaseid; |
| 308 | } PgStat_MsgDropdb; |
| 309 | |
| 310 | |
| 311 | /* ---------- |
| 312 | * PgStat_MsgResetcounter Sent by the backend to tell the collector |
| 313 | * to reset counters |
| 314 | * ---------- |
| 315 | */ |
| 316 | typedef struct PgStat_MsgResetcounter |
| 317 | { |
| 318 | PgStat_MsgHdr m_hdr; |
| 319 | Oid m_databaseid; |
| 320 | } PgStat_MsgResetcounter; |
| 321 | |
| 322 | /* ---------- |
| 323 | * PgStat_MsgResetsharedcounter Sent by the backend to tell the collector |
| 324 | * to reset a shared counter |
| 325 | * ---------- |
| 326 | */ |
| 327 | typedef struct PgStat_MsgResetsharedcounter |
| 328 | { |
| 329 | PgStat_MsgHdr m_hdr; |
| 330 | PgStat_Shared_Reset_Target m_resettarget; |
| 331 | } PgStat_MsgResetsharedcounter; |
| 332 | |
| 333 | /* ---------- |
| 334 | * PgStat_MsgResetsinglecounter Sent by the backend to tell the collector |
| 335 | * to reset a single counter |
| 336 | * ---------- |
| 337 | */ |
| 338 | typedef struct PgStat_MsgResetsinglecounter |
| 339 | { |
| 340 | PgStat_MsgHdr m_hdr; |
| 341 | Oid m_databaseid; |
| 342 | PgStat_Single_Reset_Type m_resettype; |
| 343 | Oid m_objectid; |
| 344 | } PgStat_MsgResetsinglecounter; |
| 345 | |
| 346 | /* ---------- |
| 347 | * PgStat_MsgAutovacStart Sent by the autovacuum daemon to signal |
| 348 | * that a database is going to be processed |
| 349 | * ---------- |
| 350 | */ |
| 351 | typedef struct PgStat_MsgAutovacStart |
| 352 | { |
| 353 | PgStat_MsgHdr m_hdr; |
| 354 | Oid m_databaseid; |
| 355 | TimestampTz m_start_time; |
| 356 | } PgStat_MsgAutovacStart; |
| 357 | |
| 358 | |
| 359 | /* ---------- |
| 360 | * PgStat_MsgVacuum Sent by the backend or autovacuum daemon |
| 361 | * after VACUUM |
| 362 | * ---------- |
| 363 | */ |
| 364 | typedef struct PgStat_MsgVacuum |
| 365 | { |
| 366 | PgStat_MsgHdr m_hdr; |
| 367 | Oid m_databaseid; |
| 368 | Oid m_tableoid; |
| 369 | bool m_autovacuum; |
| 370 | TimestampTz m_vacuumtime; |
| 371 | PgStat_Counter m_live_tuples; |
| 372 | PgStat_Counter m_dead_tuples; |
| 373 | } PgStat_MsgVacuum; |
| 374 | |
| 375 | |
| 376 | /* ---------- |
| 377 | * PgStat_MsgAnalyze Sent by the backend or autovacuum daemon |
| 378 | * after ANALYZE |
| 379 | * ---------- |
| 380 | */ |
| 381 | typedef struct PgStat_MsgAnalyze |
| 382 | { |
| 383 | PgStat_MsgHdr m_hdr; |
| 384 | Oid m_databaseid; |
| 385 | Oid m_tableoid; |
| 386 | bool m_autovacuum; |
| 387 | bool m_resetcounter; |
| 388 | TimestampTz m_analyzetime; |
| 389 | PgStat_Counter m_live_tuples; |
| 390 | PgStat_Counter m_dead_tuples; |
| 391 | } PgStat_MsgAnalyze; |
| 392 | |
| 393 | |
| 394 | /* ---------- |
| 395 | * PgStat_MsgArchiver Sent by the archiver to update statistics. |
| 396 | * ---------- |
| 397 | */ |
| 398 | typedef struct PgStat_MsgArchiver |
| 399 | { |
| 400 | PgStat_MsgHdr m_hdr; |
| 401 | bool m_failed; /* Failed attempt */ |
| 402 | char m_xlog[MAX_XFN_CHARS + 1]; |
| 403 | TimestampTz m_timestamp; |
| 404 | } PgStat_MsgArchiver; |
| 405 | |
| 406 | /* ---------- |
| 407 | * PgStat_MsgBgWriter Sent by the bgwriter to update statistics. |
| 408 | * ---------- |
| 409 | */ |
| 410 | typedef struct PgStat_MsgBgWriter |
| 411 | { |
| 412 | PgStat_MsgHdr m_hdr; |
| 413 | |
| 414 | PgStat_Counter m_timed_checkpoints; |
| 415 | PgStat_Counter m_requested_checkpoints; |
| 416 | PgStat_Counter m_buf_written_checkpoints; |
| 417 | PgStat_Counter m_buf_written_clean; |
| 418 | PgStat_Counter m_maxwritten_clean; |
| 419 | PgStat_Counter m_buf_written_backend; |
| 420 | PgStat_Counter m_buf_fsync_backend; |
| 421 | PgStat_Counter m_buf_alloc; |
| 422 | PgStat_Counter m_checkpoint_write_time; /* times in milliseconds */ |
| 423 | PgStat_Counter m_checkpoint_sync_time; |
| 424 | } PgStat_MsgBgWriter; |
| 425 | |
| 426 | /* ---------- |
| 427 | * PgStat_MsgRecoveryConflict Sent by the backend upon recovery conflict |
| 428 | * ---------- |
| 429 | */ |
| 430 | typedef struct PgStat_MsgRecoveryConflict |
| 431 | { |
| 432 | PgStat_MsgHdr m_hdr; |
| 433 | |
| 434 | Oid m_databaseid; |
| 435 | int m_reason; |
| 436 | } PgStat_MsgRecoveryConflict; |
| 437 | |
| 438 | /* ---------- |
| 439 | * PgStat_MsgTempFile Sent by the backend upon creating a temp file |
| 440 | * ---------- |
| 441 | */ |
| 442 | typedef struct PgStat_MsgTempFile |
| 443 | { |
| 444 | PgStat_MsgHdr m_hdr; |
| 445 | |
| 446 | Oid m_databaseid; |
| 447 | size_t m_filesize; |
| 448 | } PgStat_MsgTempFile; |
| 449 | |
| 450 | /* ---------- |
| 451 | * PgStat_FunctionCounts The actual per-function counts kept by a backend |
| 452 | * |
| 453 | * This struct should contain only actual event counters, because we memcmp |
| 454 | * it against zeroes to detect whether there are any counts to transmit. |
| 455 | * |
| 456 | * Note that the time counters are in instr_time format here. We convert to |
| 457 | * microseconds in PgStat_Counter format when transmitting to the collector. |
| 458 | * ---------- |
| 459 | */ |
| 460 | typedef struct PgStat_FunctionCounts |
| 461 | { |
| 462 | PgStat_Counter f_numcalls; |
| 463 | instr_time f_total_time; |
| 464 | instr_time f_self_time; |
| 465 | } PgStat_FunctionCounts; |
| 466 | |
| 467 | /* ---------- |
| 468 | * PgStat_BackendFunctionEntry Entry in backend's per-function hash table |
| 469 | * ---------- |
| 470 | */ |
| 471 | typedef struct PgStat_BackendFunctionEntry |
| 472 | { |
| 473 | Oid f_id; |
| 474 | PgStat_FunctionCounts f_counts; |
| 475 | } PgStat_BackendFunctionEntry; |
| 476 | |
| 477 | /* ---------- |
| 478 | * PgStat_FunctionEntry Per-function info in a MsgFuncstat |
| 479 | * ---------- |
| 480 | */ |
| 481 | typedef struct PgStat_FunctionEntry |
| 482 | { |
| 483 | Oid f_id; |
| 484 | PgStat_Counter f_numcalls; |
| 485 | PgStat_Counter f_total_time; /* times in microseconds */ |
| 486 | PgStat_Counter f_self_time; |
| 487 | } PgStat_FunctionEntry; |
| 488 | |
| 489 | /* ---------- |
| 490 | * PgStat_MsgFuncstat Sent by the backend to report function |
| 491 | * usage statistics. |
| 492 | * ---------- |
| 493 | */ |
| 494 | #define PGSTAT_NUM_FUNCENTRIES \ |
| 495 | ((PGSTAT_MSG_PAYLOAD - sizeof(Oid) - sizeof(int)) \ |
| 496 | / sizeof(PgStat_FunctionEntry)) |
| 497 | |
| 498 | typedef struct PgStat_MsgFuncstat |
| 499 | { |
| 500 | PgStat_MsgHdr m_hdr; |
| 501 | Oid m_databaseid; |
| 502 | int m_nentries; |
| 503 | PgStat_FunctionEntry m_entry[PGSTAT_NUM_FUNCENTRIES]; |
| 504 | } PgStat_MsgFuncstat; |
| 505 | |
| 506 | /* ---------- |
| 507 | * PgStat_MsgFuncpurge Sent by the backend to tell the collector |
| 508 | * about dead functions. |
| 509 | * ---------- |
| 510 | */ |
| 511 | #define PGSTAT_NUM_FUNCPURGE \ |
| 512 | ((PGSTAT_MSG_PAYLOAD - sizeof(Oid) - sizeof(int)) \ |
| 513 | / sizeof(Oid)) |
| 514 | |
| 515 | typedef struct PgStat_MsgFuncpurge |
| 516 | { |
| 517 | PgStat_MsgHdr m_hdr; |
| 518 | Oid m_databaseid; |
| 519 | int m_nentries; |
| 520 | Oid m_functionid[PGSTAT_NUM_FUNCPURGE]; |
| 521 | } PgStat_MsgFuncpurge; |
| 522 | |
| 523 | /* ---------- |
| 524 | * PgStat_MsgDeadlock Sent by the backend to tell the collector |
| 525 | * about a deadlock that occurred. |
| 526 | * ---------- |
| 527 | */ |
| 528 | typedef struct PgStat_MsgDeadlock |
| 529 | { |
| 530 | PgStat_MsgHdr m_hdr; |
| 531 | Oid m_databaseid; |
| 532 | } PgStat_MsgDeadlock; |
| 533 | |
| 534 | /* ---------- |
| 535 | * PgStat_MsgChecksumFailure Sent by the backend to tell the collector |
| 536 | * about checksum failures noticed. |
| 537 | * ---------- |
| 538 | */ |
| 539 | typedef struct PgStat_MsgChecksumFailure |
| 540 | { |
| 541 | PgStat_MsgHdr m_hdr; |
| 542 | Oid m_databaseid; |
| 543 | int m_failurecount; |
| 544 | TimestampTz m_failure_time; |
| 545 | } PgStat_MsgChecksumFailure; |
| 546 | |
| 547 | |
| 548 | /* ---------- |
| 549 | * PgStat_Msg Union over all possible messages. |
| 550 | * ---------- |
| 551 | */ |
| 552 | typedef union PgStat_Msg |
| 553 | { |
| 554 | PgStat_MsgHdr msg_hdr; |
| 555 | PgStat_MsgDummy msg_dummy; |
| 556 | PgStat_MsgInquiry msg_inquiry; |
| 557 | PgStat_MsgTabstat msg_tabstat; |
| 558 | PgStat_MsgTabpurge msg_tabpurge; |
| 559 | PgStat_MsgDropdb msg_dropdb; |
| 560 | PgStat_MsgResetcounter msg_resetcounter; |
| 561 | PgStat_MsgResetsharedcounter msg_resetsharedcounter; |
| 562 | PgStat_MsgResetsinglecounter msg_resetsinglecounter; |
| 563 | PgStat_MsgAutovacStart msg_autovacuum_start; |
| 564 | PgStat_MsgVacuum msg_vacuum; |
| 565 | PgStat_MsgAnalyze msg_analyze; |
| 566 | PgStat_MsgArchiver msg_archiver; |
| 567 | PgStat_MsgBgWriter msg_bgwriter; |
| 568 | PgStat_MsgFuncstat msg_funcstat; |
| 569 | PgStat_MsgFuncpurge msg_funcpurge; |
| 570 | PgStat_MsgRecoveryConflict msg_recoveryconflict; |
| 571 | PgStat_MsgDeadlock msg_deadlock; |
| 572 | PgStat_MsgTempFile msg_tempfile; |
| 573 | PgStat_MsgChecksumFailure msg_checksumfailure; |
| 574 | } PgStat_Msg; |
| 575 | |
| 576 | |
| 577 | /* ------------------------------------------------------------ |
| 578 | * Statistic collector data structures follow |
| 579 | * |
| 580 | * PGSTAT_FILE_FORMAT_ID should be changed whenever any of these |
| 581 | * data structures change. |
| 582 | * ------------------------------------------------------------ |
| 583 | */ |
| 584 | |
| 585 | #define PGSTAT_FILE_FORMAT_ID 0x01A5BC9D |
| 586 | |
| 587 | /* ---------- |
| 588 | * PgStat_StatDBEntry The collector's data per database |
| 589 | * ---------- |
| 590 | */ |
| 591 | typedef struct PgStat_StatDBEntry |
| 592 | { |
| 593 | Oid databaseid; |
| 594 | PgStat_Counter n_xact_commit; |
| 595 | PgStat_Counter n_xact_rollback; |
| 596 | PgStat_Counter n_blocks_fetched; |
| 597 | PgStat_Counter n_blocks_hit; |
| 598 | PgStat_Counter n_tuples_returned; |
| 599 | PgStat_Counter n_tuples_fetched; |
| 600 | PgStat_Counter n_tuples_inserted; |
| 601 | PgStat_Counter n_tuples_updated; |
| 602 | PgStat_Counter n_tuples_deleted; |
| 603 | TimestampTz last_autovac_time; |
| 604 | PgStat_Counter n_conflict_tablespace; |
| 605 | PgStat_Counter n_conflict_lock; |
| 606 | PgStat_Counter n_conflict_snapshot; |
| 607 | PgStat_Counter n_conflict_bufferpin; |
| 608 | PgStat_Counter n_conflict_startup_deadlock; |
| 609 | PgStat_Counter n_temp_files; |
| 610 | PgStat_Counter n_temp_bytes; |
| 611 | PgStat_Counter n_deadlocks; |
| 612 | PgStat_Counter n_checksum_failures; |
| 613 | TimestampTz last_checksum_failure; |
| 614 | PgStat_Counter n_block_read_time; /* times in microseconds */ |
| 615 | PgStat_Counter n_block_write_time; |
| 616 | |
| 617 | TimestampTz stat_reset_timestamp; |
| 618 | TimestampTz stats_timestamp; /* time of db stats file update */ |
| 619 | |
| 620 | /* |
| 621 | * tables and functions must be last in the struct, because we don't write |
| 622 | * the pointers out to the stats file. |
| 623 | */ |
| 624 | HTAB *tables; |
| 625 | HTAB *functions; |
| 626 | } PgStat_StatDBEntry; |
| 627 | |
| 628 | |
| 629 | /* ---------- |
| 630 | * PgStat_StatTabEntry The collector's data per table (or index) |
| 631 | * ---------- |
| 632 | */ |
| 633 | typedef struct PgStat_StatTabEntry |
| 634 | { |
| 635 | Oid tableid; |
| 636 | |
| 637 | PgStat_Counter numscans; |
| 638 | |
| 639 | PgStat_Counter tuples_returned; |
| 640 | PgStat_Counter tuples_fetched; |
| 641 | |
| 642 | PgStat_Counter tuples_inserted; |
| 643 | PgStat_Counter tuples_updated; |
| 644 | PgStat_Counter tuples_deleted; |
| 645 | PgStat_Counter tuples_hot_updated; |
| 646 | |
| 647 | PgStat_Counter n_live_tuples; |
| 648 | PgStat_Counter n_dead_tuples; |
| 649 | PgStat_Counter changes_since_analyze; |
| 650 | |
| 651 | PgStat_Counter blocks_fetched; |
| 652 | PgStat_Counter blocks_hit; |
| 653 | |
| 654 | TimestampTz vacuum_timestamp; /* user initiated vacuum */ |
| 655 | PgStat_Counter vacuum_count; |
| 656 | TimestampTz autovac_vacuum_timestamp; /* autovacuum initiated */ |
| 657 | PgStat_Counter autovac_vacuum_count; |
| 658 | TimestampTz analyze_timestamp; /* user initiated */ |
| 659 | PgStat_Counter analyze_count; |
| 660 | TimestampTz autovac_analyze_timestamp; /* autovacuum initiated */ |
| 661 | PgStat_Counter autovac_analyze_count; |
| 662 | } PgStat_StatTabEntry; |
| 663 | |
| 664 | |
| 665 | /* ---------- |
| 666 | * PgStat_StatFuncEntry The collector's data per function |
| 667 | * ---------- |
| 668 | */ |
| 669 | typedef struct PgStat_StatFuncEntry |
| 670 | { |
| 671 | Oid functionid; |
| 672 | |
| 673 | PgStat_Counter f_numcalls; |
| 674 | |
| 675 | PgStat_Counter f_total_time; /* times in microseconds */ |
| 676 | PgStat_Counter f_self_time; |
| 677 | } PgStat_StatFuncEntry; |
| 678 | |
| 679 | |
| 680 | /* |
| 681 | * Archiver statistics kept in the stats collector |
| 682 | */ |
| 683 | typedef struct PgStat_ArchiverStats |
| 684 | { |
| 685 | PgStat_Counter archived_count; /* archival successes */ |
| 686 | char last_archived_wal[MAX_XFN_CHARS + 1]; /* last WAL file |
| 687 | * archived */ |
| 688 | TimestampTz last_archived_timestamp; /* last archival success time */ |
| 689 | PgStat_Counter failed_count; /* failed archival attempts */ |
| 690 | char last_failed_wal[MAX_XFN_CHARS + 1]; /* WAL file involved in |
| 691 | * last failure */ |
| 692 | TimestampTz last_failed_timestamp; /* last archival failure time */ |
| 693 | TimestampTz stat_reset_timestamp; |
| 694 | } PgStat_ArchiverStats; |
| 695 | |
| 696 | /* |
| 697 | * Global statistics kept in the stats collector |
| 698 | */ |
| 699 | typedef struct PgStat_GlobalStats |
| 700 | { |
| 701 | TimestampTz stats_timestamp; /* time of stats file update */ |
| 702 | PgStat_Counter timed_checkpoints; |
| 703 | PgStat_Counter requested_checkpoints; |
| 704 | PgStat_Counter checkpoint_write_time; /* times in milliseconds */ |
| 705 | PgStat_Counter checkpoint_sync_time; |
| 706 | PgStat_Counter buf_written_checkpoints; |
| 707 | PgStat_Counter buf_written_clean; |
| 708 | PgStat_Counter maxwritten_clean; |
| 709 | PgStat_Counter buf_written_backend; |
| 710 | PgStat_Counter buf_fsync_backend; |
| 711 | PgStat_Counter buf_alloc; |
| 712 | TimestampTz stat_reset_timestamp; |
| 713 | } PgStat_GlobalStats; |
| 714 | |
| 715 | |
| 716 | /* ---------- |
| 717 | * Backend types |
| 718 | * ---------- |
| 719 | */ |
| 720 | typedef enum BackendType |
| 721 | { |
| 722 | B_AUTOVAC_LAUNCHER, |
| 723 | B_AUTOVAC_WORKER, |
| 724 | B_BACKEND, |
| 725 | B_BG_WORKER, |
| 726 | B_BG_WRITER, |
| 727 | B_CHECKPOINTER, |
| 728 | B_STARTUP, |
| 729 | B_WAL_RECEIVER, |
| 730 | B_WAL_SENDER, |
| 731 | B_WAL_WRITER |
| 732 | } BackendType; |
| 733 | |
| 734 | |
| 735 | /* ---------- |
| 736 | * Backend states |
| 737 | * ---------- |
| 738 | */ |
| 739 | typedef enum BackendState |
| 740 | { |
| 741 | STATE_UNDEFINED, |
| 742 | STATE_IDLE, |
| 743 | STATE_RUNNING, |
| 744 | STATE_IDLEINTRANSACTION, |
| 745 | STATE_FASTPATH, |
| 746 | STATE_IDLEINTRANSACTION_ABORTED, |
| 747 | STATE_DISABLED |
| 748 | } BackendState; |
| 749 | |
| 750 | |
| 751 | /* ---------- |
| 752 | * Wait Classes |
| 753 | * ---------- |
| 754 | */ |
| 755 | #define PG_WAIT_LWLOCK 0x01000000U |
| 756 | #define PG_WAIT_LOCK 0x03000000U |
| 757 | #define PG_WAIT_BUFFER_PIN 0x04000000U |
| 758 | #define PG_WAIT_ACTIVITY 0x05000000U |
| 759 | #define PG_WAIT_CLIENT 0x06000000U |
| 760 | #define PG_WAIT_EXTENSION 0x07000000U |
| 761 | #define PG_WAIT_IPC 0x08000000U |
| 762 | #define PG_WAIT_TIMEOUT 0x09000000U |
| 763 | #define PG_WAIT_IO 0x0A000000U |
| 764 | |
| 765 | /* ---------- |
| 766 | * Wait Events - Activity |
| 767 | * |
| 768 | * Use this category when a process is waiting because it has no work to do, |
| 769 | * unless the "Client" or "Timeout" category describes the situation better. |
| 770 | * Typically, this should only be used for background processes. |
| 771 | * ---------- |
| 772 | */ |
| 773 | typedef enum |
| 774 | { |
| 775 | WAIT_EVENT_ARCHIVER_MAIN = PG_WAIT_ACTIVITY, |
| 776 | WAIT_EVENT_AUTOVACUUM_MAIN, |
| 777 | WAIT_EVENT_BGWRITER_HIBERNATE, |
| 778 | WAIT_EVENT_BGWRITER_MAIN, |
| 779 | WAIT_EVENT_CHECKPOINTER_MAIN, |
| 780 | WAIT_EVENT_LOGICAL_APPLY_MAIN, |
| 781 | WAIT_EVENT_LOGICAL_LAUNCHER_MAIN, |
| 782 | WAIT_EVENT_PGSTAT_MAIN, |
| 783 | WAIT_EVENT_RECOVERY_WAL_ALL, |
| 784 | WAIT_EVENT_RECOVERY_WAL_STREAM, |
| 785 | WAIT_EVENT_SYSLOGGER_MAIN, |
| 786 | WAIT_EVENT_WAL_RECEIVER_MAIN, |
| 787 | WAIT_EVENT_WAL_SENDER_MAIN, |
| 788 | WAIT_EVENT_WAL_WRITER_MAIN |
| 789 | } WaitEventActivity; |
| 790 | |
| 791 | /* ---------- |
| 792 | * Wait Events - Client |
| 793 | * |
| 794 | * Use this category when a process is waiting to send data to or receive data |
| 795 | * from the frontend process to which it is connected. This is never used for |
| 796 | * a background process, which has no client connection. |
| 797 | * ---------- |
| 798 | */ |
| 799 | typedef enum |
| 800 | { |
| 801 | WAIT_EVENT_CLIENT_READ = PG_WAIT_CLIENT, |
| 802 | WAIT_EVENT_CLIENT_WRITE, |
| 803 | WAIT_EVENT_LIBPQWALRECEIVER_CONNECT, |
| 804 | WAIT_EVENT_LIBPQWALRECEIVER_RECEIVE, |
| 805 | WAIT_EVENT_SSL_OPEN_SERVER, |
| 806 | WAIT_EVENT_WAL_RECEIVER_WAIT_START, |
| 807 | WAIT_EVENT_WAL_SENDER_WAIT_WAL, |
| 808 | WAIT_EVENT_WAL_SENDER_WRITE_DATA, |
| 809 | WAIT_EVENT_GSS_OPEN_SERVER, |
| 810 | } WaitEventClient; |
| 811 | |
| 812 | /* ---------- |
| 813 | * Wait Events - IPC |
| 814 | * |
| 815 | * Use this category when a process cannot complete the work it is doing because |
| 816 | * it is waiting for a notification from another process. |
| 817 | * ---------- |
| 818 | */ |
| 819 | typedef enum |
| 820 | { |
| 821 | WAIT_EVENT_BGWORKER_SHUTDOWN = PG_WAIT_IPC, |
| 822 | WAIT_EVENT_BGWORKER_STARTUP, |
| 823 | WAIT_EVENT_BTREE_PAGE, |
| 824 | WAIT_EVENT_CLOG_GROUP_UPDATE, |
| 825 | WAIT_EVENT_CHECKPOINT_DONE, |
| 826 | WAIT_EVENT_CHECKPOINT_START, |
| 827 | WAIT_EVENT_EXECUTE_GATHER, |
| 828 | WAIT_EVENT_HASH_BATCH_ALLOCATING, |
| 829 | WAIT_EVENT_HASH_BATCH_ELECTING, |
| 830 | WAIT_EVENT_HASH_BATCH_LOADING, |
| 831 | WAIT_EVENT_HASH_BUILD_ALLOCATING, |
| 832 | WAIT_EVENT_HASH_BUILD_ELECTING, |
| 833 | WAIT_EVENT_HASH_BUILD_HASHING_INNER, |
| 834 | WAIT_EVENT_HASH_BUILD_HASHING_OUTER, |
| 835 | WAIT_EVENT_HASH_GROW_BATCHES_ALLOCATING, |
| 836 | WAIT_EVENT_HASH_GROW_BATCHES_DECIDING, |
| 837 | WAIT_EVENT_HASH_GROW_BATCHES_ELECTING, |
| 838 | WAIT_EVENT_HASH_GROW_BATCHES_FINISHING, |
| 839 | WAIT_EVENT_HASH_GROW_BATCHES_REPARTITIONING, |
| 840 | WAIT_EVENT_HASH_GROW_BUCKETS_ALLOCATING, |
| 841 | WAIT_EVENT_HASH_GROW_BUCKETS_ELECTING, |
| 842 | WAIT_EVENT_HASH_GROW_BUCKETS_REINSERTING, |
| 843 | WAIT_EVENT_LOGICAL_SYNC_DATA, |
| 844 | WAIT_EVENT_LOGICAL_SYNC_STATE_CHANGE, |
| 845 | WAIT_EVENT_MQ_INTERNAL, |
| 846 | WAIT_EVENT_MQ_PUT_MESSAGE, |
| 847 | WAIT_EVENT_MQ_RECEIVE, |
| 848 | WAIT_EVENT_MQ_SEND, |
| 849 | WAIT_EVENT_PARALLEL_BITMAP_SCAN, |
| 850 | WAIT_EVENT_PARALLEL_CREATE_INDEX_SCAN, |
| 851 | WAIT_EVENT_PARALLEL_FINISH, |
| 852 | WAIT_EVENT_PROCARRAY_GROUP_UPDATE, |
| 853 | WAIT_EVENT_PROMOTE, |
| 854 | WAIT_EVENT_REPLICATION_ORIGIN_DROP, |
| 855 | WAIT_EVENT_REPLICATION_SLOT_DROP, |
| 856 | WAIT_EVENT_SAFE_SNAPSHOT, |
| 857 | WAIT_EVENT_SYNC_REP |
| 858 | } WaitEventIPC; |
| 859 | |
| 860 | /* ---------- |
| 861 | * Wait Events - Timeout |
| 862 | * |
| 863 | * Use this category when a process is waiting for a timeout to expire. |
| 864 | * ---------- |
| 865 | */ |
| 866 | typedef enum |
| 867 | { |
| 868 | WAIT_EVENT_BASE_BACKUP_THROTTLE = PG_WAIT_TIMEOUT, |
| 869 | WAIT_EVENT_PG_SLEEP, |
| 870 | WAIT_EVENT_RECOVERY_APPLY_DELAY |
| 871 | } WaitEventTimeout; |
| 872 | |
| 873 | /* ---------- |
| 874 | * Wait Events - IO |
| 875 | * |
| 876 | * Use this category when a process is waiting for a IO. |
| 877 | * ---------- |
| 878 | */ |
| 879 | typedef enum |
| 880 | { |
| 881 | WAIT_EVENT_BUFFILE_READ = PG_WAIT_IO, |
| 882 | WAIT_EVENT_BUFFILE_WRITE, |
| 883 | WAIT_EVENT_CONTROL_FILE_READ, |
| 884 | WAIT_EVENT_CONTROL_FILE_SYNC, |
| 885 | WAIT_EVENT_CONTROL_FILE_SYNC_UPDATE, |
| 886 | WAIT_EVENT_CONTROL_FILE_WRITE, |
| 887 | WAIT_EVENT_CONTROL_FILE_WRITE_UPDATE, |
| 888 | WAIT_EVENT_COPY_FILE_READ, |
| 889 | WAIT_EVENT_COPY_FILE_WRITE, |
| 890 | WAIT_EVENT_DATA_FILE_EXTEND, |
| 891 | WAIT_EVENT_DATA_FILE_FLUSH, |
| 892 | WAIT_EVENT_DATA_FILE_IMMEDIATE_SYNC, |
| 893 | WAIT_EVENT_DATA_FILE_PREFETCH, |
| 894 | WAIT_EVENT_DATA_FILE_READ, |
| 895 | WAIT_EVENT_DATA_FILE_SYNC, |
| 896 | WAIT_EVENT_DATA_FILE_TRUNCATE, |
| 897 | WAIT_EVENT_DATA_FILE_WRITE, |
| 898 | WAIT_EVENT_DSM_FILL_ZERO_WRITE, |
| 899 | WAIT_EVENT_LOCK_FILE_ADDTODATADIR_READ, |
| 900 | WAIT_EVENT_LOCK_FILE_ADDTODATADIR_SYNC, |
| 901 | WAIT_EVENT_LOCK_FILE_ADDTODATADIR_WRITE, |
| 902 | WAIT_EVENT_LOCK_FILE_CREATE_READ, |
| 903 | WAIT_EVENT_LOCK_FILE_CREATE_SYNC, |
| 904 | WAIT_EVENT_LOCK_FILE_CREATE_WRITE, |
| 905 | WAIT_EVENT_LOCK_FILE_RECHECKDATADIR_READ, |
| 906 | WAIT_EVENT_LOGICAL_REWRITE_CHECKPOINT_SYNC, |
| 907 | WAIT_EVENT_LOGICAL_REWRITE_MAPPING_SYNC, |
| 908 | WAIT_EVENT_LOGICAL_REWRITE_MAPPING_WRITE, |
| 909 | WAIT_EVENT_LOGICAL_REWRITE_SYNC, |
| 910 | WAIT_EVENT_LOGICAL_REWRITE_TRUNCATE, |
| 911 | WAIT_EVENT_LOGICAL_REWRITE_WRITE, |
| 912 | WAIT_EVENT_RELATION_MAP_READ, |
| 913 | WAIT_EVENT_RELATION_MAP_SYNC, |
| 914 | WAIT_EVENT_RELATION_MAP_WRITE, |
| 915 | WAIT_EVENT_REORDER_BUFFER_READ, |
| 916 | WAIT_EVENT_REORDER_BUFFER_WRITE, |
| 917 | WAIT_EVENT_REORDER_LOGICAL_MAPPING_READ, |
| 918 | WAIT_EVENT_REPLICATION_SLOT_READ, |
| 919 | WAIT_EVENT_REPLICATION_SLOT_RESTORE_SYNC, |
| 920 | WAIT_EVENT_REPLICATION_SLOT_SYNC, |
| 921 | WAIT_EVENT_REPLICATION_SLOT_WRITE, |
| 922 | WAIT_EVENT_SLRU_FLUSH_SYNC, |
| 923 | WAIT_EVENT_SLRU_READ, |
| 924 | WAIT_EVENT_SLRU_SYNC, |
| 925 | WAIT_EVENT_SLRU_WRITE, |
| 926 | WAIT_EVENT_SNAPBUILD_READ, |
| 927 | WAIT_EVENT_SNAPBUILD_SYNC, |
| 928 | WAIT_EVENT_SNAPBUILD_WRITE, |
| 929 | WAIT_EVENT_TIMELINE_HISTORY_FILE_SYNC, |
| 930 | WAIT_EVENT_TIMELINE_HISTORY_FILE_WRITE, |
| 931 | WAIT_EVENT_TIMELINE_HISTORY_READ, |
| 932 | WAIT_EVENT_TIMELINE_HISTORY_SYNC, |
| 933 | WAIT_EVENT_TIMELINE_HISTORY_WRITE, |
| 934 | WAIT_EVENT_TWOPHASE_FILE_READ, |
| 935 | WAIT_EVENT_TWOPHASE_FILE_SYNC, |
| 936 | WAIT_EVENT_TWOPHASE_FILE_WRITE, |
| 937 | WAIT_EVENT_WALSENDER_TIMELINE_HISTORY_READ, |
| 938 | WAIT_EVENT_WAL_BOOTSTRAP_SYNC, |
| 939 | WAIT_EVENT_WAL_BOOTSTRAP_WRITE, |
| 940 | WAIT_EVENT_WAL_COPY_READ, |
| 941 | WAIT_EVENT_WAL_COPY_SYNC, |
| 942 | WAIT_EVENT_WAL_COPY_WRITE, |
| 943 | WAIT_EVENT_WAL_INIT_SYNC, |
| 944 | WAIT_EVENT_WAL_INIT_WRITE, |
| 945 | WAIT_EVENT_WAL_READ, |
| 946 | WAIT_EVENT_WAL_SYNC, |
| 947 | WAIT_EVENT_WAL_SYNC_METHOD_ASSIGN, |
| 948 | WAIT_EVENT_WAL_WRITE |
| 949 | } WaitEventIO; |
| 950 | |
| 951 | /* ---------- |
| 952 | * Command type for progress reporting purposes |
| 953 | * ---------- |
| 954 | */ |
| 955 | typedef enum ProgressCommandType |
| 956 | { |
| 957 | PROGRESS_COMMAND_INVALID, |
| 958 | PROGRESS_COMMAND_VACUUM, |
| 959 | PROGRESS_COMMAND_CLUSTER, |
| 960 | PROGRESS_COMMAND_CREATE_INDEX |
| 961 | } ProgressCommandType; |
| 962 | |
| 963 | #define PGSTAT_NUM_PROGRESS_PARAM 20 |
| 964 | |
| 965 | /* ---------- |
| 966 | * Shared-memory data structures |
| 967 | * ---------- |
| 968 | */ |
| 969 | |
| 970 | |
| 971 | /* |
| 972 | * PgBackendSSLStatus |
| 973 | * |
| 974 | * For each backend, we keep the SSL status in a separate struct, that |
| 975 | * is only filled in if SSL is enabled. |
| 976 | * |
| 977 | * All char arrays must be null-terminated. |
| 978 | */ |
| 979 | typedef struct PgBackendSSLStatus |
| 980 | { |
| 981 | /* Information about SSL connection */ |
| 982 | int ssl_bits; |
| 983 | bool ssl_compression; |
| 984 | char ssl_version[NAMEDATALEN]; |
| 985 | char ssl_cipher[NAMEDATALEN]; |
| 986 | char ssl_client_dn[NAMEDATALEN]; |
| 987 | |
| 988 | /* |
| 989 | * serial number is max "20 octets" per RFC 5280, so this size should be |
| 990 | * fine |
| 991 | */ |
| 992 | char ssl_client_serial[NAMEDATALEN]; |
| 993 | |
| 994 | char ssl_issuer_dn[NAMEDATALEN]; |
| 995 | } PgBackendSSLStatus; |
| 996 | |
| 997 | /* |
| 998 | * PgBackendGSSStatus |
| 999 | * |
| 1000 | * For each backend, we keep the GSS status in a separate struct, that |
| 1001 | * is only filled in if GSS is enabled. |
| 1002 | * |
| 1003 | * All char arrays must be null-terminated. |
| 1004 | */ |
| 1005 | typedef struct PgBackendGSSStatus |
| 1006 | { |
| 1007 | /* Information about GSSAPI connection */ |
| 1008 | char gss_princ[NAMEDATALEN]; /* GSSAPI Principal used to auth */ |
| 1009 | bool gss_auth; /* If GSSAPI authentication was used */ |
| 1010 | bool gss_enc; /* If encryption is being used */ |
| 1011 | |
| 1012 | } PgBackendGSSStatus; |
| 1013 | |
| 1014 | |
| 1015 | /* ---------- |
| 1016 | * PgBackendStatus |
| 1017 | * |
| 1018 | * Each live backend maintains a PgBackendStatus struct in shared memory |
| 1019 | * showing its current activity. (The structs are allocated according to |
| 1020 | * BackendId, but that is not critical.) Note that the collector process |
| 1021 | * has no involvement in, or even access to, these structs. |
| 1022 | * |
| 1023 | * Each auxiliary process also maintains a PgBackendStatus struct in shared |
| 1024 | * memory. |
| 1025 | * ---------- |
| 1026 | */ |
| 1027 | typedef struct PgBackendStatus |
| 1028 | { |
| 1029 | /* |
| 1030 | * To avoid locking overhead, we use the following protocol: a backend |
| 1031 | * increments st_changecount before modifying its entry, and again after |
| 1032 | * finishing a modification. A would-be reader should note the value of |
| 1033 | * st_changecount, copy the entry into private memory, then check |
| 1034 | * st_changecount again. If the value hasn't changed, and if it's even, |
| 1035 | * the copy is valid; otherwise start over. This makes updates cheap |
| 1036 | * while reads are potentially expensive, but that's the tradeoff we want. |
| 1037 | * |
| 1038 | * The above protocol needs memory barriers to ensure that the apparent |
| 1039 | * order of execution is as it desires. Otherwise, for example, the CPU |
| 1040 | * might rearrange the code so that st_changecount is incremented twice |
| 1041 | * before the modification on a machine with weak memory ordering. Hence, |
| 1042 | * use the macros defined below for manipulating st_changecount, rather |
| 1043 | * than touching it directly. |
| 1044 | */ |
| 1045 | int st_changecount; |
| 1046 | |
| 1047 | /* The entry is valid iff st_procpid > 0, unused if st_procpid == 0 */ |
| 1048 | int st_procpid; |
| 1049 | |
| 1050 | /* Type of backends */ |
| 1051 | BackendType st_backendType; |
| 1052 | |
| 1053 | /* Times when current backend, transaction, and activity started */ |
| 1054 | TimestampTz st_proc_start_timestamp; |
| 1055 | TimestampTz st_xact_start_timestamp; |
| 1056 | TimestampTz st_activity_start_timestamp; |
| 1057 | TimestampTz st_state_start_timestamp; |
| 1058 | |
| 1059 | /* Database OID, owning user's OID, connection client address */ |
| 1060 | Oid st_databaseid; |
| 1061 | Oid st_userid; |
| 1062 | SockAddr st_clientaddr; |
| 1063 | char *st_clienthostname; /* MUST be null-terminated */ |
| 1064 | |
| 1065 | /* Information about SSL connection */ |
| 1066 | bool st_ssl; |
| 1067 | PgBackendSSLStatus *st_sslstatus; |
| 1068 | |
| 1069 | /* Information about GSSAPI connection */ |
| 1070 | bool st_gss; |
| 1071 | PgBackendGSSStatus *st_gssstatus; |
| 1072 | |
| 1073 | /* current state */ |
| 1074 | BackendState st_state; |
| 1075 | |
| 1076 | /* application name; MUST be null-terminated */ |
| 1077 | char *st_appname; |
| 1078 | |
| 1079 | /* |
| 1080 | * Current command string; MUST be null-terminated. Note that this string |
| 1081 | * possibly is truncated in the middle of a multi-byte character. As |
| 1082 | * activity strings are stored more frequently than read, that allows to |
| 1083 | * move the cost of correct truncation to the display side. Use |
| 1084 | * pgstat_clip_activity() to truncate correctly. |
| 1085 | */ |
| 1086 | char *st_activity_raw; |
| 1087 | |
| 1088 | /* |
| 1089 | * Command progress reporting. Any command which wishes can advertise |
| 1090 | * that it is running by setting st_progress_command, |
| 1091 | * st_progress_command_target, and st_progress_param[]. |
| 1092 | * st_progress_command_target should be the OID of the relation which the |
| 1093 | * command targets (we assume there's just one, as this is meant for |
| 1094 | * utility commands), but the meaning of each element in the |
| 1095 | * st_progress_param array is command-specific. |
| 1096 | */ |
| 1097 | ProgressCommandType st_progress_command; |
| 1098 | Oid st_progress_command_target; |
| 1099 | int64 st_progress_param[PGSTAT_NUM_PROGRESS_PARAM]; |
| 1100 | } PgBackendStatus; |
| 1101 | |
| 1102 | /* |
| 1103 | * Macros to load and store st_changecount with appropriate memory barriers. |
| 1104 | * |
| 1105 | * Use PGSTAT_BEGIN_WRITE_ACTIVITY() before, and PGSTAT_END_WRITE_ACTIVITY() |
| 1106 | * after, modifying the current process's PgBackendStatus data. Note that, |
| 1107 | * since there is no mechanism for cleaning up st_changecount after an error, |
| 1108 | * THESE MACROS FORM A CRITICAL SECTION. Any error between them will be |
| 1109 | * promoted to PANIC, causing a database restart to clean up shared memory! |
| 1110 | * Hence, keep the critical section as short and straight-line as possible. |
| 1111 | * Aside from being safer, that minimizes the window in which readers will |
| 1112 | * have to loop. |
| 1113 | * |
| 1114 | * Reader logic should follow this sketch: |
| 1115 | * |
| 1116 | * for (;;) |
| 1117 | * { |
| 1118 | * int before_ct, after_ct; |
| 1119 | * |
| 1120 | * pgstat_begin_read_activity(beentry, before_ct); |
| 1121 | * ... copy beentry data to local memory ... |
| 1122 | * pgstat_end_read_activity(beentry, after_ct); |
| 1123 | * if (pgstat_read_activity_complete(before_ct, after_ct)) |
| 1124 | * break; |
| 1125 | * CHECK_FOR_INTERRUPTS(); |
| 1126 | * } |
| 1127 | * |
| 1128 | * For extra safety, we generally use volatile beentry pointers, although |
| 1129 | * the memory barriers should theoretically be sufficient. |
| 1130 | */ |
| 1131 | #define PGSTAT_BEGIN_WRITE_ACTIVITY(beentry) \ |
| 1132 | do { \ |
| 1133 | START_CRIT_SECTION(); \ |
| 1134 | (beentry)->st_changecount++; \ |
| 1135 | pg_write_barrier(); \ |
| 1136 | } while (0) |
| 1137 | |
| 1138 | #define PGSTAT_END_WRITE_ACTIVITY(beentry) \ |
| 1139 | do { \ |
| 1140 | pg_write_barrier(); \ |
| 1141 | (beentry)->st_changecount++; \ |
| 1142 | Assert(((beentry)->st_changecount & 1) == 0); \ |
| 1143 | END_CRIT_SECTION(); \ |
| 1144 | } while (0) |
| 1145 | |
| 1146 | #define pgstat_begin_read_activity(beentry, before_changecount) \ |
| 1147 | do { \ |
| 1148 | (before_changecount) = (beentry)->st_changecount; \ |
| 1149 | pg_read_barrier(); \ |
| 1150 | } while (0) |
| 1151 | |
| 1152 | #define pgstat_end_read_activity(beentry, after_changecount) \ |
| 1153 | do { \ |
| 1154 | pg_read_barrier(); \ |
| 1155 | (after_changecount) = (beentry)->st_changecount; \ |
| 1156 | } while (0) |
| 1157 | |
| 1158 | #define pgstat_read_activity_complete(before_changecount, after_changecount) \ |
| 1159 | ((before_changecount) == (after_changecount) && \ |
| 1160 | ((before_changecount) & 1) == 0) |
| 1161 | |
| 1162 | |
| 1163 | /* ---------- |
| 1164 | * LocalPgBackendStatus |
| 1165 | * |
| 1166 | * When we build the backend status array, we use LocalPgBackendStatus to be |
| 1167 | * able to add new values to the struct when needed without adding new fields |
| 1168 | * to the shared memory. It contains the backend status as a first member. |
| 1169 | * ---------- |
| 1170 | */ |
| 1171 | typedef struct LocalPgBackendStatus |
| 1172 | { |
| 1173 | /* |
| 1174 | * Local version of the backend status entry. |
| 1175 | */ |
| 1176 | PgBackendStatus backendStatus; |
| 1177 | |
| 1178 | /* |
| 1179 | * The xid of the current transaction if available, InvalidTransactionId |
| 1180 | * if not. |
| 1181 | */ |
| 1182 | TransactionId backend_xid; |
| 1183 | |
| 1184 | /* |
| 1185 | * The xmin of the current session if available, InvalidTransactionId if |
| 1186 | * not. |
| 1187 | */ |
| 1188 | TransactionId backend_xmin; |
| 1189 | } LocalPgBackendStatus; |
| 1190 | |
| 1191 | /* |
| 1192 | * Working state needed to accumulate per-function-call timing statistics. |
| 1193 | */ |
| 1194 | typedef struct PgStat_FunctionCallUsage |
| 1195 | { |
| 1196 | /* Link to function's hashtable entry (must still be there at exit!) */ |
| 1197 | /* NULL means we are not tracking the current function call */ |
| 1198 | PgStat_FunctionCounts *fs; |
| 1199 | /* Total time previously charged to function, as of function start */ |
| 1200 | instr_time save_f_total_time; |
| 1201 | /* Backend-wide total time as of function start */ |
| 1202 | instr_time save_total; |
| 1203 | /* system clock as of function start */ |
| 1204 | instr_time f_start; |
| 1205 | } PgStat_FunctionCallUsage; |
| 1206 | |
| 1207 | |
| 1208 | /* ---------- |
| 1209 | * GUC parameters |
| 1210 | * ---------- |
| 1211 | */ |
| 1212 | extern bool pgstat_track_activities; |
| 1213 | extern bool pgstat_track_counts; |
| 1214 | extern int pgstat_track_functions; |
| 1215 | extern PGDLLIMPORT int pgstat_track_activity_query_size; |
| 1216 | extern char *pgstat_stat_directory; |
| 1217 | extern char *pgstat_stat_tmpname; |
| 1218 | extern char *pgstat_stat_filename; |
| 1219 | |
| 1220 | /* |
| 1221 | * BgWriter statistics counters are updated directly by bgwriter and bufmgr |
| 1222 | */ |
| 1223 | extern PgStat_MsgBgWriter BgWriterStats; |
| 1224 | |
| 1225 | /* |
| 1226 | * Updated by pgstat_count_buffer_*_time macros |
| 1227 | */ |
| 1228 | extern PgStat_Counter pgStatBlockReadTime; |
| 1229 | extern PgStat_Counter pgStatBlockWriteTime; |
| 1230 | |
| 1231 | /* ---------- |
| 1232 | * Functions called from postmaster |
| 1233 | * ---------- |
| 1234 | */ |
| 1235 | extern Size BackendStatusShmemSize(void); |
| 1236 | extern void CreateSharedBackendStatus(void); |
| 1237 | |
| 1238 | extern void pgstat_init(void); |
| 1239 | extern int pgstat_start(void); |
| 1240 | extern void pgstat_reset_all(void); |
| 1241 | extern void allow_immediate_pgstat_restart(void); |
| 1242 | |
| 1243 | #ifdef EXEC_BACKEND |
| 1244 | extern void PgstatCollectorMain(int argc, char *argv[]) pg_attribute_noreturn(); |
| 1245 | #endif |
| 1246 | |
| 1247 | |
| 1248 | /* ---------- |
| 1249 | * Functions called from backends |
| 1250 | * ---------- |
| 1251 | */ |
| 1252 | extern void pgstat_ping(void); |
| 1253 | |
| 1254 | extern void pgstat_report_stat(bool force); |
| 1255 | extern void pgstat_vacuum_stat(void); |
| 1256 | extern void pgstat_drop_database(Oid databaseid); |
| 1257 | |
| 1258 | extern void pgstat_clear_snapshot(void); |
| 1259 | extern void pgstat_reset_counters(void); |
| 1260 | extern void pgstat_reset_shared_counters(const char *); |
| 1261 | extern void pgstat_reset_single_counter(Oid objectid, PgStat_Single_Reset_Type type); |
| 1262 | |
| 1263 | extern void pgstat_report_autovac(Oid dboid); |
| 1264 | extern void pgstat_report_vacuum(Oid tableoid, bool shared, |
| 1265 | PgStat_Counter livetuples, PgStat_Counter deadtuples); |
| 1266 | extern void pgstat_report_analyze(Relation rel, |
| 1267 | PgStat_Counter livetuples, PgStat_Counter deadtuples, |
| 1268 | bool resetcounter); |
| 1269 | |
| 1270 | extern void pgstat_report_recovery_conflict(int reason); |
| 1271 | extern void pgstat_report_deadlock(void); |
| 1272 | extern void pgstat_report_checksum_failures_in_db(Oid dboid, int failurecount); |
| 1273 | extern void pgstat_report_checksum_failure(void); |
| 1274 | |
| 1275 | extern void pgstat_initialize(void); |
| 1276 | extern void pgstat_bestart(void); |
| 1277 | |
| 1278 | extern void pgstat_report_activity(BackendState state, const char *cmd_str); |
| 1279 | extern void pgstat_report_tempfile(size_t filesize); |
| 1280 | extern void pgstat_report_appname(const char *appname); |
| 1281 | extern void pgstat_report_xact_timestamp(TimestampTz tstamp); |
| 1282 | extern const char *pgstat_get_wait_event(uint32 wait_event_info); |
| 1283 | extern const char *pgstat_get_wait_event_type(uint32 wait_event_info); |
| 1284 | extern const char *pgstat_get_backend_current_activity(int pid, bool checkUser); |
| 1285 | extern const char *pgstat_get_crashed_backend_activity(int pid, char *buffer, |
| 1286 | int buflen); |
| 1287 | extern const char *pgstat_get_backend_desc(BackendType backendType); |
| 1288 | |
| 1289 | extern void pgstat_progress_start_command(ProgressCommandType cmdtype, |
| 1290 | Oid relid); |
| 1291 | extern void pgstat_progress_update_param(int index, int64 val); |
| 1292 | extern void pgstat_progress_update_multi_param(int nparam, const int *index, |
| 1293 | const int64 *val); |
| 1294 | extern void pgstat_progress_end_command(void); |
| 1295 | |
| 1296 | extern PgStat_TableStatus *find_tabstat_entry(Oid rel_id); |
| 1297 | extern PgStat_BackendFunctionEntry *find_funcstat_entry(Oid func_id); |
| 1298 | |
| 1299 | extern void pgstat_initstats(Relation rel); |
| 1300 | |
| 1301 | extern char *pgstat_clip_activity(const char *raw_activity); |
| 1302 | |
| 1303 | /* ---------- |
| 1304 | * pgstat_report_wait_start() - |
| 1305 | * |
| 1306 | * Called from places where server process needs to wait. This is called |
| 1307 | * to report wait event information. The wait information is stored |
| 1308 | * as 4-bytes where first byte represents the wait event class (type of |
| 1309 | * wait, for different types of wait, refer WaitClass) and the next |
| 1310 | * 3-bytes represent the actual wait event. Currently 2-bytes are used |
| 1311 | * for wait event which is sufficient for current usage, 1-byte is |
| 1312 | * reserved for future usage. |
| 1313 | * |
| 1314 | * NB: this *must* be able to survive being called before MyProc has been |
| 1315 | * initialized. |
| 1316 | * ---------- |
| 1317 | */ |
| 1318 | static inline void |
| 1319 | pgstat_report_wait_start(uint32 wait_event_info) |
| 1320 | { |
| 1321 | volatile PGPROC *proc = MyProc; |
| 1322 | |
| 1323 | if (!pgstat_track_activities || !proc) |
| 1324 | return; |
| 1325 | |
| 1326 | /* |
| 1327 | * Since this is a four-byte field which is always read and written as |
| 1328 | * four-bytes, updates are atomic. |
| 1329 | */ |
| 1330 | proc->wait_event_info = wait_event_info; |
| 1331 | } |
| 1332 | |
| 1333 | /* ---------- |
| 1334 | * pgstat_report_wait_end() - |
| 1335 | * |
| 1336 | * Called to report end of a wait. |
| 1337 | * |
| 1338 | * NB: this *must* be able to survive being called before MyProc has been |
| 1339 | * initialized. |
| 1340 | * ---------- |
| 1341 | */ |
| 1342 | static inline void |
| 1343 | pgstat_report_wait_end(void) |
| 1344 | { |
| 1345 | volatile PGPROC *proc = MyProc; |
| 1346 | |
| 1347 | if (!pgstat_track_activities || !proc) |
| 1348 | return; |
| 1349 | |
| 1350 | /* |
| 1351 | * Since this is a four-byte field which is always read and written as |
| 1352 | * four-bytes, updates are atomic. |
| 1353 | */ |
| 1354 | proc->wait_event_info = 0; |
| 1355 | } |
| 1356 | |
| 1357 | /* nontransactional event counts are simple enough to inline */ |
| 1358 | |
| 1359 | #define pgstat_count_heap_scan(rel) \ |
| 1360 | do { \ |
| 1361 | if ((rel)->pgstat_info != NULL) \ |
| 1362 | (rel)->pgstat_info->t_counts.t_numscans++; \ |
| 1363 | } while (0) |
| 1364 | #define pgstat_count_heap_getnext(rel) \ |
| 1365 | do { \ |
| 1366 | if ((rel)->pgstat_info != NULL) \ |
| 1367 | (rel)->pgstat_info->t_counts.t_tuples_returned++; \ |
| 1368 | } while (0) |
| 1369 | #define pgstat_count_heap_fetch(rel) \ |
| 1370 | do { \ |
| 1371 | if ((rel)->pgstat_info != NULL) \ |
| 1372 | (rel)->pgstat_info->t_counts.t_tuples_fetched++; \ |
| 1373 | } while (0) |
| 1374 | #define pgstat_count_index_scan(rel) \ |
| 1375 | do { \ |
| 1376 | if ((rel)->pgstat_info != NULL) \ |
| 1377 | (rel)->pgstat_info->t_counts.t_numscans++; \ |
| 1378 | } while (0) |
| 1379 | #define pgstat_count_index_tuples(rel, n) \ |
| 1380 | do { \ |
| 1381 | if ((rel)->pgstat_info != NULL) \ |
| 1382 | (rel)->pgstat_info->t_counts.t_tuples_returned += (n); \ |
| 1383 | } while (0) |
| 1384 | #define pgstat_count_buffer_read(rel) \ |
| 1385 | do { \ |
| 1386 | if ((rel)->pgstat_info != NULL) \ |
| 1387 | (rel)->pgstat_info->t_counts.t_blocks_fetched++; \ |
| 1388 | } while (0) |
| 1389 | #define pgstat_count_buffer_hit(rel) \ |
| 1390 | do { \ |
| 1391 | if ((rel)->pgstat_info != NULL) \ |
| 1392 | (rel)->pgstat_info->t_counts.t_blocks_hit++; \ |
| 1393 | } while (0) |
| 1394 | #define pgstat_count_buffer_read_time(n) \ |
| 1395 | (pgStatBlockReadTime += (n)) |
| 1396 | #define pgstat_count_buffer_write_time(n) \ |
| 1397 | (pgStatBlockWriteTime += (n)) |
| 1398 | |
| 1399 | extern void pgstat_count_heap_insert(Relation rel, PgStat_Counter n); |
| 1400 | extern void pgstat_count_heap_update(Relation rel, bool hot); |
| 1401 | extern void pgstat_count_heap_delete(Relation rel); |
| 1402 | extern void pgstat_count_truncate(Relation rel); |
| 1403 | extern void pgstat_update_heap_dead_tuples(Relation rel, int delta); |
| 1404 | |
| 1405 | extern void pgstat_init_function_usage(FunctionCallInfo fcinfo, |
| 1406 | PgStat_FunctionCallUsage *fcu); |
| 1407 | extern void pgstat_end_function_usage(PgStat_FunctionCallUsage *fcu, |
| 1408 | bool finalize); |
| 1409 | |
| 1410 | extern void AtEOXact_PgStat(bool isCommit, bool parallel); |
| 1411 | extern void AtEOSubXact_PgStat(bool isCommit, int nestDepth); |
| 1412 | |
| 1413 | extern void AtPrepare_PgStat(void); |
| 1414 | extern void PostPrepare_PgStat(void); |
| 1415 | |
| 1416 | extern void pgstat_twophase_postcommit(TransactionId xid, uint16 info, |
| 1417 | void *recdata, uint32 len); |
| 1418 | extern void pgstat_twophase_postabort(TransactionId xid, uint16 info, |
| 1419 | void *recdata, uint32 len); |
| 1420 | |
| 1421 | extern void pgstat_send_archiver(const char *xlog, bool failed); |
| 1422 | extern void pgstat_send_bgwriter(void); |
| 1423 | |
| 1424 | /* ---------- |
| 1425 | * Support functions for the SQL-callable functions to |
| 1426 | * generate the pgstat* views. |
| 1427 | * ---------- |
| 1428 | */ |
| 1429 | extern PgStat_StatDBEntry *pgstat_fetch_stat_dbentry(Oid dbid); |
| 1430 | extern PgStat_StatTabEntry *pgstat_fetch_stat_tabentry(Oid relid); |
| 1431 | extern PgBackendStatus *pgstat_fetch_stat_beentry(int beid); |
| 1432 | extern LocalPgBackendStatus *pgstat_fetch_stat_local_beentry(int beid); |
| 1433 | extern PgStat_StatFuncEntry *pgstat_fetch_stat_funcentry(Oid funcid); |
| 1434 | extern int pgstat_fetch_stat_numbackends(void); |
| 1435 | extern PgStat_ArchiverStats *pgstat_fetch_stat_archiver(void); |
| 1436 | extern PgStat_GlobalStats *pgstat_fetch_global(void); |
| 1437 | |
| 1438 | #endif /* PGSTAT_H */ |
| 1439 | |