Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * instrument.c
4 : * functions for instrumentation of plan execution
5 : *
6 : *
7 : * Copyright (c) 2001-2025, 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 86530 : 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 86530 : instr = palloc0(n * sizeof(Instrumentation));
37 86530 : if (instrument_options & (INSTRUMENT_BUFFERS | INSTRUMENT_TIMER | INSTRUMENT_WAL))
38 : {
39 81636 : bool need_buffers = (instrument_options & INSTRUMENT_BUFFERS) != 0;
40 81636 : bool need_wal = (instrument_options & INSTRUMENT_WAL) != 0;
41 81636 : bool need_timer = (instrument_options & INSTRUMENT_TIMER) != 0;
42 : int i;
43 :
44 163272 : for (i = 0; i < n; i++)
45 : {
46 81636 : instr[i].need_bufusage = need_buffers;
47 81636 : instr[i].need_walusage = need_wal;
48 81636 : instr[i].need_timer = need_timer;
49 81636 : instr[i].async_mode = async_mode;
50 : }
51 : }
52 :
53 86530 : 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 12970542 : InstrStartNode(Instrumentation *instr)
69 : {
70 23097736 : if (instr->need_timer &&
71 10127194 : !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 12970542 : if (instr->need_bufusage)
76 10048442 : instr->bufusage_start = pgBufferUsage;
77 :
78 12970542 : if (instr->need_walusage)
79 149544 : instr->walusage_start = pgWalUsage;
80 12970542 : }
81 :
82 : /* Exit from a plan node */
83 : void
84 12967500 : InstrStopNode(Instrumentation *instr, double nTuples)
85 : {
86 12967500 : double save_tuplecount = instr->tuplecount;
87 : instr_time endtime;
88 :
89 : /* count the returned tuples */
90 12967500 : instr->tuplecount += nTuples;
91 :
92 : /* let's update the time only if the timer was requested */
93 12967500 : if (instr->need_timer)
94 : {
95 10124164 : if (INSTR_TIME_IS_ZERO(instr->starttime))
96 0 : elog(ERROR, "InstrStopNode called without start");
97 :
98 10124164 : INSTR_TIME_SET_CURRENT(endtime);
99 10124164 : INSTR_TIME_ACCUM_DIFF(instr->counter, endtime, instr->starttime);
100 :
101 10124164 : INSTR_TIME_SET_ZERO(instr->starttime);
102 : }
103 :
104 : /* Add delta of buffer usage since entry to node's totals */
105 12967500 : if (instr->need_bufusage)
106 10045412 : BufferUsageAccumDiff(&instr->bufusage,
107 10045412 : &pgBufferUsage, &instr->bufusage_start);
108 :
109 12967500 : if (instr->need_walusage)
110 146514 : WalUsageAccumDiff(&instr->walusage,
111 146514 : &pgWalUsage, &instr->walusage_start);
112 :
113 : /* Is this the first tuple of this cycle? */
114 12967500 : if (!instr->running)
115 : {
116 123818 : instr->running = true;
117 123818 : instr->firsttuple = INSTR_TIME_GET_DOUBLE(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 12843682 : if (instr->async_mode && save_tuplecount < 1.0)
126 0 : instr->firsttuple = INSTR_TIME_GET_DOUBLE(instr->counter);
127 : }
128 12967500 : }
129 :
130 : /* Update tuple count */
131 : void
132 2 : InstrUpdateTupleCount(Instrumentation *instr, double nTuples)
133 : {
134 : /* count the returned tuples */
135 2 : instr->tuplecount += nTuples;
136 2 : }
137 :
138 : /* Finish a run cycle for a plan node */
139 : void
140 125738 : InstrEndLoop(Instrumentation *instr)
141 : {
142 : double totaltime;
143 :
144 : /* Skip if nothing has happened, or already shut down */
145 125738 : if (!instr->running)
146 2242 : return;
147 :
148 123496 : if (!INSTR_TIME_IS_ZERO(instr->starttime))
149 0 : elog(ERROR, "InstrEndLoop called on running node");
150 :
151 : /* Accumulate per-cycle statistics into totals */
152 123496 : totaltime = INSTR_TIME_GET_DOUBLE(instr->counter);
153 :
154 123496 : instr->startup += instr->firsttuple;
155 123496 : instr->total += totaltime;
156 123496 : instr->ntuples += instr->tuplecount;
157 123496 : instr->nloops += 1;
158 :
159 : /* Reset for next cycle (if any) */
160 123496 : instr->running = false;
161 123496 : INSTR_TIME_SET_ZERO(instr->starttime);
162 123496 : INSTR_TIME_SET_ZERO(instr->counter);
163 123496 : instr->firsttuple = 0;
164 123496 : instr->tuplecount = 0;
165 : }
166 :
167 : /* aggregate instrumentation information */
168 : void
169 4056 : InstrAggNode(Instrumentation *dst, Instrumentation *add)
170 : {
171 4056 : if (!dst->running && add->running)
172 : {
173 0 : dst->running = true;
174 0 : dst->firsttuple = add->firsttuple;
175 : }
176 4056 : else if (dst->running && add->running && dst->firsttuple > add->firsttuple)
177 0 : dst->firsttuple = add->firsttuple;
178 :
179 4056 : INSTR_TIME_ADD(dst->counter, add->counter);
180 :
181 4056 : dst->tuplecount += add->tuplecount;
182 4056 : dst->startup += add->startup;
183 4056 : dst->total += add->total;
184 4056 : dst->ntuples += add->ntuples;
185 4056 : dst->ntuples2 += add->ntuples2;
186 4056 : dst->nloops += add->nloops;
187 4056 : dst->nfiltered1 += add->nfiltered1;
188 4056 : dst->nfiltered2 += add->nfiltered2;
189 :
190 : /* Add delta of buffer usage since entry to node's totals */
191 4056 : if (dst->need_bufusage)
192 1908 : BufferUsageAdd(&dst->bufusage, &add->bufusage);
193 :
194 4056 : if (dst->need_walusage)
195 0 : WalUsageAdd(&dst->walusage, &add->walusage);
196 4056 : }
197 :
198 : /* note current values during parallel executor startup */
199 : void
200 2740 : InstrStartParallelQuery(void)
201 : {
202 2740 : save_pgBufferUsage = pgBufferUsage;
203 2740 : save_pgWalUsage = pgWalUsage;
204 2740 : }
205 :
206 : /* report usage after parallel executor shutdown */
207 : void
208 2728 : InstrEndParallelQuery(BufferUsage *bufusage, WalUsage *walusage)
209 : {
210 2728 : memset(bufusage, 0, sizeof(BufferUsage));
211 2728 : BufferUsageAccumDiff(bufusage, &pgBufferUsage, &save_pgBufferUsage);
212 2728 : memset(walusage, 0, sizeof(WalUsage));
213 2728 : WalUsageAccumDiff(walusage, &pgWalUsage, &save_pgWalUsage);
214 2728 : }
215 :
216 : /* accumulate work done by workers in leader's stats */
217 : void
218 2728 : InstrAccumParallelQuery(BufferUsage *bufusage, WalUsage *walusage)
219 : {
220 2728 : BufferUsageAdd(&pgBufferUsage, bufusage);
221 2728 : WalUsageAdd(&pgWalUsage, walusage);
222 2728 : }
223 :
224 : /* dst += add */
225 : static void
226 4636 : BufferUsageAdd(BufferUsage *dst, const BufferUsage *add)
227 : {
228 4636 : dst->shared_blks_hit += add->shared_blks_hit;
229 4636 : dst->shared_blks_read += add->shared_blks_read;
230 4636 : dst->shared_blks_dirtied += add->shared_blks_dirtied;
231 4636 : dst->shared_blks_written += add->shared_blks_written;
232 4636 : dst->local_blks_hit += add->local_blks_hit;
233 4636 : dst->local_blks_read += add->local_blks_read;
234 4636 : dst->local_blks_dirtied += add->local_blks_dirtied;
235 4636 : dst->local_blks_written += add->local_blks_written;
236 4636 : dst->temp_blks_read += add->temp_blks_read;
237 4636 : dst->temp_blks_written += add->temp_blks_written;
238 4636 : INSTR_TIME_ADD(dst->shared_blk_read_time, add->shared_blk_read_time);
239 4636 : INSTR_TIME_ADD(dst->shared_blk_write_time, add->shared_blk_write_time);
240 4636 : INSTR_TIME_ADD(dst->local_blk_read_time, add->local_blk_read_time);
241 4636 : INSTR_TIME_ADD(dst->local_blk_write_time, add->local_blk_write_time);
242 4636 : INSTR_TIME_ADD(dst->temp_blk_read_time, add->temp_blk_read_time);
243 4636 : INSTR_TIME_ADD(dst->temp_blk_write_time, add->temp_blk_write_time);
244 4636 : }
245 :
246 : /* dst += add - sub */
247 : void
248 10192162 : BufferUsageAccumDiff(BufferUsage *dst,
249 : const BufferUsage *add,
250 : const BufferUsage *sub)
251 : {
252 10192162 : dst->shared_blks_hit += add->shared_blks_hit - sub->shared_blks_hit;
253 10192162 : dst->shared_blks_read += add->shared_blks_read - sub->shared_blks_read;
254 10192162 : dst->shared_blks_dirtied += add->shared_blks_dirtied - sub->shared_blks_dirtied;
255 10192162 : dst->shared_blks_written += add->shared_blks_written - sub->shared_blks_written;
256 10192162 : dst->local_blks_hit += add->local_blks_hit - sub->local_blks_hit;
257 10192162 : dst->local_blks_read += add->local_blks_read - sub->local_blks_read;
258 10192162 : dst->local_blks_dirtied += add->local_blks_dirtied - sub->local_blks_dirtied;
259 10192162 : dst->local_blks_written += add->local_blks_written - sub->local_blks_written;
260 10192162 : dst->temp_blks_read += add->temp_blks_read - sub->temp_blks_read;
261 10192162 : dst->temp_blks_written += add->temp_blks_written - sub->temp_blks_written;
262 10192162 : INSTR_TIME_ACCUM_DIFF(dst->shared_blk_read_time,
263 : add->shared_blk_read_time, sub->shared_blk_read_time);
264 10192162 : INSTR_TIME_ACCUM_DIFF(dst->shared_blk_write_time,
265 : add->shared_blk_write_time, sub->shared_blk_write_time);
266 10192162 : INSTR_TIME_ACCUM_DIFF(dst->local_blk_read_time,
267 : add->local_blk_read_time, sub->local_blk_read_time);
268 10192162 : INSTR_TIME_ACCUM_DIFF(dst->local_blk_write_time,
269 : add->local_blk_write_time, sub->local_blk_write_time);
270 10192162 : INSTR_TIME_ACCUM_DIFF(dst->temp_blk_read_time,
271 : add->temp_blk_read_time, sub->temp_blk_read_time);
272 10192162 : INSTR_TIME_ACCUM_DIFF(dst->temp_blk_write_time,
273 : add->temp_blk_write_time, sub->temp_blk_write_time);
274 10192162 : }
275 :
276 : /* helper functions for WAL usage accumulation */
277 : static void
278 2728 : WalUsageAdd(WalUsage *dst, WalUsage *add)
279 : {
280 2728 : dst->wal_bytes += add->wal_bytes;
281 2728 : dst->wal_records += add->wal_records;
282 2728 : dst->wal_fpi += add->wal_fpi;
283 2728 : dst->wal_buffers_full += add->wal_buffers_full;
284 2728 : }
285 :
286 : void
287 326942 : WalUsageAccumDiff(WalUsage *dst, const WalUsage *add, const WalUsage *sub)
288 : {
289 326942 : dst->wal_bytes += add->wal_bytes - sub->wal_bytes;
290 326942 : dst->wal_records += add->wal_records - sub->wal_records;
291 326942 : dst->wal_fpi += add->wal_fpi - sub->wal_fpi;
292 326942 : dst->wal_buffers_full += add->wal_buffers_full - sub->wal_buffers_full;
293 326942 : }
|
