Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * instrument.c
4 : * functions for instrumentation of plan execution
5 : *
6 : *
7 : * Copyright (c) 2001-2026, PostgreSQL Global Development Group
8 : *
9 : * IDENTIFICATION
10 : * src/backend/executor/instrument.c
11 : *
12 : *-------------------------------------------------------------------------
13 : */
14 : #include "postgres.h"
15 :
16 : #include <unistd.h>
17 :
18 : #include "executor/instrument.h"
19 :
20 : BufferUsage pgBufferUsage;
21 : static BufferUsage save_pgBufferUsage;
22 : WalUsage pgWalUsage;
23 : static WalUsage save_pgWalUsage;
24 :
25 : static void BufferUsageAdd(BufferUsage *dst, const BufferUsage *add);
26 : static void WalUsageAdd(WalUsage *dst, WalUsage *add);
27 :
28 :
29 : /* Allocate new instrumentation structure(s) */
30 : Instrumentation *
31 89070 : InstrAlloc(int n, int instrument_options, bool async_mode)
32 : {
33 : Instrumentation *instr;
34 :
35 : /* initialize all fields to zeroes, then modify as needed */
36 89070 : instr = palloc0(n * sizeof(Instrumentation));
37 89070 : if (instrument_options & (INSTRUMENT_BUFFERS | INSTRUMENT_TIMER | INSTRUMENT_WAL))
38 : {
39 84022 : bool need_buffers = (instrument_options & INSTRUMENT_BUFFERS) != 0;
40 84022 : bool need_wal = (instrument_options & INSTRUMENT_WAL) != 0;
41 84022 : bool need_timer = (instrument_options & INSTRUMENT_TIMER) != 0;
42 : int i;
43 :
44 168044 : for (i = 0; i < n; i++)
45 : {
46 84022 : instr[i].need_bufusage = need_buffers;
47 84022 : instr[i].need_walusage = need_wal;
48 84022 : instr[i].need_timer = need_timer;
49 84022 : instr[i].async_mode = async_mode;
50 : }
51 : }
52 :
53 89070 : return instr;
54 : }
55 :
56 : /* Initialize a pre-allocated instrumentation structure. */
57 : void
58 1680 : InstrInit(Instrumentation *instr, int instrument_options)
59 : {
60 1680 : memset(instr, 0, sizeof(Instrumentation));
61 1680 : instr->need_bufusage = (instrument_options & INSTRUMENT_BUFFERS) != 0;
62 1680 : instr->need_walusage = (instrument_options & INSTRUMENT_WAL) != 0;
63 1680 : instr->need_timer = (instrument_options & INSTRUMENT_TIMER) != 0;
64 1680 : }
65 :
66 : /* Entry to a plan node */
67 : void
68 13562604 : InstrStartNode(Instrumentation *instr)
69 : {
70 24277024 : if (instr->need_timer &&
71 10714420 : !INSTR_TIME_SET_CURRENT_LAZY(instr->starttime))
72 0 : elog(ERROR, "InstrStartNode called twice in a row");
73 :
74 : /* save buffer usage totals at node entry, if needed */
75 13562604 : if (instr->need_bufusage)
76 10635652 : instr->bufusage_start = pgBufferUsage;
77 :
78 13562604 : if (instr->need_walusage)
79 154202 : instr->walusage_start = pgWalUsage;
80 13562604 : }
81 :
82 : /* Exit from a plan node */
83 : void
84 13559484 : InstrStopNode(Instrumentation *instr, double nTuples)
85 : {
86 13559484 : double save_tuplecount = instr->tuplecount;
87 : instr_time endtime;
88 :
89 : /* count the returned tuples */
90 13559484 : instr->tuplecount += nTuples;
91 :
92 : /* let's update the time only if the timer was requested */
93 13559484 : if (instr->need_timer)
94 : {
95 10711312 : if (INSTR_TIME_IS_ZERO(instr->starttime))
96 0 : elog(ERROR, "InstrStopNode called without start");
97 :
98 10711312 : INSTR_TIME_SET_CURRENT(endtime);
99 10711312 : INSTR_TIME_ACCUM_DIFF(instr->counter, endtime, instr->starttime);
100 :
101 10711312 : INSTR_TIME_SET_ZERO(instr->starttime);
102 : }
103 :
104 : /* Add delta of buffer usage since entry to node's totals */
105 13559484 : if (instr->need_bufusage)
106 10632544 : BufferUsageAccumDiff(&instr->bufusage,
107 10632544 : &pgBufferUsage, &instr->bufusage_start);
108 :
109 13559484 : if (instr->need_walusage)
110 151094 : WalUsageAccumDiff(&instr->walusage,
111 151094 : &pgWalUsage, &instr->walusage_start);
112 :
113 : /* Is this the first tuple of this cycle? */
114 13559484 : if (!instr->running)
115 : {
116 127008 : instr->running = true;
117 127008 : instr->firsttuple = instr->counter;
118 : }
119 : else
120 : {
121 : /*
122 : * In async mode, if the plan node hadn't emitted any tuples before,
123 : * this might be the first tuple
124 : */
125 13432476 : if (instr->async_mode && save_tuplecount < 1.0)
126 0 : instr->firsttuple = instr->counter;
127 : }
128 13559484 : }
129 :
130 : /* Update tuple count */
131 : void
132 4 : InstrUpdateTupleCount(Instrumentation *instr, double nTuples)
133 : {
134 : /* count the returned tuples */
135 4 : instr->tuplecount += nTuples;
136 4 : }
137 :
138 : /* Finish a run cycle for a plan node */
139 : void
140 128862 : InstrEndLoop(Instrumentation *instr)
141 : {
142 : /* Skip if nothing has happened, or already shut down */
143 128862 : if (!instr->running)
144 2198 : return;
145 :
146 126664 : if (!INSTR_TIME_IS_ZERO(instr->starttime))
147 0 : elog(ERROR, "InstrEndLoop called on running node");
148 :
149 : /* Accumulate per-cycle statistics into totals */
150 126664 : INSTR_TIME_ADD(instr->startup, instr->firsttuple);
151 126664 : INSTR_TIME_ADD(instr->total, instr->counter);
152 126664 : instr->ntuples += instr->tuplecount;
153 126664 : instr->nloops += 1;
154 :
155 : /* Reset for next cycle (if any) */
156 126664 : instr->running = false;
157 126664 : INSTR_TIME_SET_ZERO(instr->starttime);
158 126664 : INSTR_TIME_SET_ZERO(instr->counter);
159 126664 : INSTR_TIME_SET_ZERO(instr->firsttuple);
160 126664 : instr->tuplecount = 0;
161 : }
162 :
163 : /* aggregate instrumentation information */
164 : void
165 4056 : InstrAggNode(Instrumentation *dst, Instrumentation *add)
166 : {
167 4056 : if (!dst->running && add->running)
168 : {
169 0 : dst->running = true;
170 0 : dst->firsttuple = add->firsttuple;
171 : }
172 4056 : else if (dst->running && add->running &&
173 0 : INSTR_TIME_LT(dst->firsttuple, add->firsttuple))
174 0 : dst->firsttuple = add->firsttuple;
175 :
176 4056 : INSTR_TIME_ADD(dst->counter, add->counter);
177 :
178 4056 : dst->tuplecount += add->tuplecount;
179 4056 : INSTR_TIME_ADD(dst->startup, add->startup);
180 4056 : INSTR_TIME_ADD(dst->total, add->total);
181 4056 : dst->ntuples += add->ntuples;
182 4056 : dst->ntuples2 += add->ntuples2;
183 4056 : dst->nloops += add->nloops;
184 4056 : dst->nfiltered1 += add->nfiltered1;
185 4056 : dst->nfiltered2 += add->nfiltered2;
186 :
187 : /* Add delta of buffer usage since entry to node's totals */
188 4056 : if (dst->need_bufusage)
189 1908 : BufferUsageAdd(&dst->bufusage, &add->bufusage);
190 :
191 4056 : if (dst->need_walusage)
192 0 : WalUsageAdd(&dst->walusage, &add->walusage);
193 4056 : }
194 :
195 : /* note current values during parallel executor startup */
196 : void
197 2874 : InstrStartParallelQuery(void)
198 : {
199 2874 : save_pgBufferUsage = pgBufferUsage;
200 2874 : save_pgWalUsage = pgWalUsage;
201 2874 : }
202 :
203 : /* report usage after parallel executor shutdown */
204 : void
205 2862 : InstrEndParallelQuery(BufferUsage *bufusage, WalUsage *walusage)
206 : {
207 2862 : memset(bufusage, 0, sizeof(BufferUsage));
208 2862 : BufferUsageAccumDiff(bufusage, &pgBufferUsage, &save_pgBufferUsage);
209 2862 : memset(walusage, 0, sizeof(WalUsage));
210 2862 : WalUsageAccumDiff(walusage, &pgWalUsage, &save_pgWalUsage);
211 2862 : }
212 :
213 : /* accumulate work done by workers in leader's stats */
214 : void
215 2862 : InstrAccumParallelQuery(BufferUsage *bufusage, WalUsage *walusage)
216 : {
217 2862 : BufferUsageAdd(&pgBufferUsage, bufusage);
218 2862 : WalUsageAdd(&pgWalUsage, walusage);
219 2862 : }
220 :
221 : /* dst += add */
222 : static void
223 4770 : BufferUsageAdd(BufferUsage *dst, const BufferUsage *add)
224 : {
225 4770 : dst->shared_blks_hit += add->shared_blks_hit;
226 4770 : dst->shared_blks_read += add->shared_blks_read;
227 4770 : dst->shared_blks_dirtied += add->shared_blks_dirtied;
228 4770 : dst->shared_blks_written += add->shared_blks_written;
229 4770 : dst->local_blks_hit += add->local_blks_hit;
230 4770 : dst->local_blks_read += add->local_blks_read;
231 4770 : dst->local_blks_dirtied += add->local_blks_dirtied;
232 4770 : dst->local_blks_written += add->local_blks_written;
233 4770 : dst->temp_blks_read += add->temp_blks_read;
234 4770 : dst->temp_blks_written += add->temp_blks_written;
235 4770 : INSTR_TIME_ADD(dst->shared_blk_read_time, add->shared_blk_read_time);
236 4770 : INSTR_TIME_ADD(dst->shared_blk_write_time, add->shared_blk_write_time);
237 4770 : INSTR_TIME_ADD(dst->local_blk_read_time, add->local_blk_read_time);
238 4770 : INSTR_TIME_ADD(dst->local_blk_write_time, add->local_blk_write_time);
239 4770 : INSTR_TIME_ADD(dst->temp_blk_read_time, add->temp_blk_read_time);
240 4770 : INSTR_TIME_ADD(dst->temp_blk_write_time, add->temp_blk_write_time);
241 4770 : }
242 :
243 : /* dst += add - sub */
244 : void
245 10792728 : BufferUsageAccumDiff(BufferUsage *dst,
246 : const BufferUsage *add,
247 : const BufferUsage *sub)
248 : {
249 10792728 : dst->shared_blks_hit += add->shared_blks_hit - sub->shared_blks_hit;
250 10792728 : dst->shared_blks_read += add->shared_blks_read - sub->shared_blks_read;
251 10792728 : dst->shared_blks_dirtied += add->shared_blks_dirtied - sub->shared_blks_dirtied;
252 10792728 : dst->shared_blks_written += add->shared_blks_written - sub->shared_blks_written;
253 10792728 : dst->local_blks_hit += add->local_blks_hit - sub->local_blks_hit;
254 10792728 : dst->local_blks_read += add->local_blks_read - sub->local_blks_read;
255 10792728 : dst->local_blks_dirtied += add->local_blks_dirtied - sub->local_blks_dirtied;
256 10792728 : dst->local_blks_written += add->local_blks_written - sub->local_blks_written;
257 10792728 : dst->temp_blks_read += add->temp_blks_read - sub->temp_blks_read;
258 10792728 : dst->temp_blks_written += add->temp_blks_written - sub->temp_blks_written;
259 10792728 : INSTR_TIME_ACCUM_DIFF(dst->shared_blk_read_time,
260 : add->shared_blk_read_time, sub->shared_blk_read_time);
261 10792728 : INSTR_TIME_ACCUM_DIFF(dst->shared_blk_write_time,
262 : add->shared_blk_write_time, sub->shared_blk_write_time);
263 10792728 : INSTR_TIME_ACCUM_DIFF(dst->local_blk_read_time,
264 : add->local_blk_read_time, sub->local_blk_read_time);
265 10792728 : INSTR_TIME_ACCUM_DIFF(dst->local_blk_write_time,
266 : add->local_blk_write_time, sub->local_blk_write_time);
267 10792728 : INSTR_TIME_ACCUM_DIFF(dst->temp_blk_read_time,
268 : add->temp_blk_read_time, sub->temp_blk_read_time);
269 10792728 : INSTR_TIME_ACCUM_DIFF(dst->temp_blk_write_time,
270 : add->temp_blk_write_time, sub->temp_blk_write_time);
271 10792728 : }
272 :
273 : /* helper functions for WAL usage accumulation */
274 : static void
275 2862 : WalUsageAdd(WalUsage *dst, WalUsage *add)
276 : {
277 2862 : dst->wal_bytes += add->wal_bytes;
278 2862 : dst->wal_records += add->wal_records;
279 2862 : dst->wal_fpi += add->wal_fpi;
280 2862 : dst->wal_fpi_bytes += add->wal_fpi_bytes;
281 2862 : dst->wal_buffers_full += add->wal_buffers_full;
282 2862 : }
283 :
284 : void
285 349624 : WalUsageAccumDiff(WalUsage *dst, const WalUsage *add, const WalUsage *sub)
286 : {
287 349624 : dst->wal_bytes += add->wal_bytes - sub->wal_bytes;
288 349624 : dst->wal_records += add->wal_records - sub->wal_records;
289 349624 : dst->wal_fpi += add->wal_fpi - sub->wal_fpi;
290 349624 : dst->wal_fpi_bytes += add->wal_fpi_bytes - sub->wal_fpi_bytes;
291 349624 : dst->wal_buffers_full += add->wal_buffers_full - sub->wal_buffers_full;
292 349624 : }
|
