1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
|
/*-
* Public Domain 2014-2015 MongoDB, Inc.
* Public Domain 2008-2014 WiredTiger, Inc.
*
* This is free and unencumbered software released into the public domain.
*
* Anyone is free to copy, modify, publish, use, compile, sell, or
* distribute this software, either in source code form or as a compiled
* binary, for any purpose, commercial or non-commercial, and by any
* means.
*
* In jurisdictions that recognize copyright laws, the author or authors
* of this software dedicate any and all copyright interest in the
* software to the public domain. We make this dedication for the benefit
* of the public at large and to the detriment of our heirs and
* successors. We intend this dedication to be an overt act of
* relinquishment in perpetuity of all present and future rights to this
* software under copyright law.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
* IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*/
#include "wtperf.h"
/* Default values. */
static const CONFIG default_cfg = {
"WT_TEST", /* home */
"WT_TEST", /* monitor dir */
NULL, /* partial logging */
NULL, /* base_uri */
NULL, /* uris */
NULL, /* helium_mount */
NULL, /* conn */
NULL, /* logf */
NULL, /* async */
NULL, NULL, /* compressor ext, blk */
NULL, NULL, /* populate, checkpoint threads */
NULL, /* worker threads */
0, /* worker thread count */
NULL, /* workloads */
0, /* workload count */
0, /* use_asyncops */
0, /* checkpoint operations */
0, /* insert operations */
0, /* read operations */
0, /* truncate operations */
0, /* update operations */
0, /* insert key */
0, /* checkpoint in progress */
0, /* thread error */
0, /* notify threads to stop */
0, /* in warmup phase */
0, /* total seconds running */
0, /* has truncate */
{NULL, NULL}, /* the truncate queue */
#define OPT_DEFINE_DEFAULT
#include "wtperf_opt.i"
#undef OPT_DEFINE_DEFAULT
};
static const char * const debug_cconfig = "";
static const char * const debug_tconfig = "";
static void *checkpoint_worker(void *);
static int create_tables(CONFIG *);
static int create_uris(CONFIG *);
static int drop_all_tables(CONFIG *);
static int execute_populate(CONFIG *);
static int execute_workload(CONFIG *);
static int find_table_count(CONFIG *);
static void *monitor(void *);
static void *populate_thread(void *);
static void randomize_value(CONFIG_THREAD *, char *);
static int start_all_runs(CONFIG *);
static int start_run(CONFIG *);
static int start_threads(CONFIG *,
WORKLOAD *, CONFIG_THREAD *, u_int, void *(*)(void *));
static int stop_threads(CONFIG *, u_int, CONFIG_THREAD *);
static void *thread_run_wtperf(void *);
static void *worker(void *);
static void worker_throttle(int64_t, int64_t *, struct timespec *);
static uint64_t wtperf_rand(CONFIG_THREAD *);
static uint64_t wtperf_value_range(CONFIG *);
#define HELIUM_NAME "dev1"
#define HELIUM_PATH \
"../../ext/test/helium/.libs/libwiredtiger_helium.so"
#define HELIUM_CONFIG ",type=helium"
#define INDEX_COL_NAMES ",columns=(key,val)"
/* Retrieve an ID for the next insert operation. */
static inline uint64_t
get_next_incr(CONFIG *cfg)
{
return (WT_ATOMIC_ADD8(cfg->insert_key, 1));
}
static inline void
generate_key(CONFIG *cfg, char *key_buf, uint64_t keyno)
{
/*
* Don't change to snprintf, sprintf is faster in some tests.
*/
sprintf(key_buf, "%0*" PRIu64, cfg->key_sz - 1, keyno);
}
static inline uint64_t
decode_key(CONFIG *cfg, char *key_buf)
{
char *endPtr;
uint32_t i;
i= 0;
for (;i < cfg->key_sz; i++)
if (key_buf[i] != 0)
break;
return (strtoull (key_buf + i, &endPtr, 10));
}
static void
randomize_value(CONFIG_THREAD *thread, char *value_buf)
{
uint8_t *vb;
uint32_t i;
/*
* Each time we're called overwrite value_buf[0] and one other
* randomly chosen byte (other than the trailing NUL).
* Make sure we don't write a NUL: keep the value the same length.
*/
i = __wt_random(&thread->rnd) % (thread->cfg->value_sz - 1);
while (value_buf[i] == '\0' && i > 0)
--i;
if (i > 0) {
vb = (uint8_t *)value_buf;
vb[0] = (__wt_random(&thread->rnd) % 255) + 1;
vb[i] = (__wt_random(&thread->rnd) % 255) + 1;
}
}
static int
cb_asyncop(WT_ASYNC_CALLBACK *cb, WT_ASYNC_OP *op, int ret, uint32_t flags)
{
CONFIG *cfg;
CONFIG_THREAD *thread;
TRACK *trk;
WT_ASYNC_OPTYPE type;
char *value;
uint32_t *tables;
int t_ret;
(void)cb;
(void)flags;
cfg = NULL; /* -Wconditional-uninitialized */
thread = NULL; /* -Wconditional-uninitialized */
type = op->get_type(op);
if (type != WT_AOP_COMPACT) {
thread = (CONFIG_THREAD *)op->app_private;
cfg = thread->cfg;
}
trk = NULL;
switch (type) {
case WT_AOP_COMPACT:
tables = (uint32_t *)op->app_private;
WT_ATOMIC_ADD4(*tables, (uint32_t)-1);
break;
case WT_AOP_INSERT:
trk = &thread->insert;
break;
case WT_AOP_SEARCH:
trk = &thread->read;
if (ret == 0 &&
(t_ret = op->get_value(op, &value)) != 0) {
ret = t_ret;
lprintf(cfg, ret, 0, "get_value in read.");
goto err;
}
break;
case WT_AOP_UPDATE:
trk = &thread->update;
break;
case WT_AOP_NONE:
case WT_AOP_REMOVE:
/* We never expect this type. */
lprintf(cfg, ret, 0, "No type in op %" PRIu64, op->get_id(op));
goto err;
}
/*
* Either we have success and we track it, or failure and panic.
*
* Reads and updates can fail with WT_NOTFOUND: we may be searching
* in a random range, or an insert op might have updated the
* last record in the table but not yet finished the actual insert.
*/
if (type == WT_AOP_COMPACT)
return (0);
if (ret == 0 || (ret == WT_NOTFOUND && type != WT_AOP_INSERT)) {
if (!cfg->in_warmup)
(void)WT_ATOMIC_ADD8(trk->ops, 1);
return (0);
}
err:
/* Panic if error */
lprintf(cfg, ret, 0, "Error in op %" PRIu64,
op->get_id(op));
cfg->error = cfg->stop = 1;
return (1);
}
static WT_ASYNC_CALLBACK cb = { cb_asyncop };
/*
* track_operation --
* Update an operation's tracking structure with new latency information.
*/
static inline void
track_operation(TRACK *trk, uint64_t usecs)
{
uint64_t v;
/* average microseconds per call */
v = (uint64_t)usecs;
trk->latency += usecs; /* track total latency */
if (v > trk->max_latency) /* track max/min latency */
trk->max_latency = (uint32_t)v;
if (v < trk->min_latency)
trk->min_latency = (uint32_t)v;
/*
* Update a latency bucket.
* First buckets: usecs from 100us to 1000us at 100us each.
*/
if (v < 1000)
++trk->us[v];
/*
* Second buckets: milliseconds from 1ms to 1000ms, at 1ms each.
*/
else if (v < ms_to_us(1000))
++trk->ms[us_to_ms(v)];
/*
* Third buckets are seconds from 1s to 100s, at 1s each.
*/
else if (v < sec_to_us(100))
++trk->sec[us_to_sec(v)];
/* >100 seconds, accumulate in the biggest bucket. */
else
++trk->sec[ELEMENTS(trk->sec) - 1];
}
static const char *
op_name(uint8_t *op)
{
switch (*op) {
case WORKER_INSERT:
return ("insert");
case WORKER_INSERT_RMW:
return ("insert_rmw");
case WORKER_READ:
return ("read");
case WORKER_TRUNCATE:
return ("truncate");
case WORKER_UPDATE:
return ("update");
default:
return ("unknown");
}
/* NOTREACHED */
}
static void *
worker_async(void *arg)
{
CONFIG *cfg;
CONFIG_THREAD *thread;
WT_ASYNC_OP *asyncop;
WT_CONNECTION *conn;
uint64_t next_val;
uint8_t *op, *op_end;
int ret;
char *key_buf, *value_buf;
thread = (CONFIG_THREAD *)arg;
cfg = thread->cfg;
conn = cfg->conn;
key_buf = thread->key_buf;
value_buf = thread->value_buf;
op = thread->workload->ops;
op_end = op + sizeof(thread->workload->ops);
while (!cfg->stop) {
/*
* Generate the next key and setup operation specific
* statistics tracking objects.
*/
switch (*op) {
case WORKER_INSERT:
case WORKER_INSERT_RMW:
if (cfg->random_range)
next_val = wtperf_rand(thread);
else
next_val = cfg->icount + get_next_incr(cfg);
break;
case WORKER_READ:
case WORKER_UPDATE:
next_val = wtperf_rand(thread);
/*
* If the workload is started without a populate phase
* we rely on at least one insert to get a valid item
* id.
*/
if (wtperf_value_range(cfg) < next_val)
continue;
break;
default:
goto err; /* can't happen */
}
generate_key(cfg, key_buf, next_val);
/*
* Spread the data out around the multiple databases.
* Sleep to allow workers a chance to run and process async ops.
* Then retry to get an async op.
*/
while ((ret = conn->async_new_op(
conn, cfg->uris[next_val % cfg->table_count],
NULL, &cb, &asyncop)) == EBUSY)
(void)usleep(10000);
if (ret != 0)
goto err;
asyncop->app_private = thread;
asyncop->set_key(asyncop, key_buf);
switch (*op) {
case WORKER_READ:
ret = asyncop->search(asyncop);
if (ret == 0)
break;
goto op_err;
case WORKER_INSERT:
if (cfg->random_value)
randomize_value(thread, value_buf);
asyncop->set_value(asyncop, value_buf);
if ((ret = asyncop->insert(asyncop)) == 0)
break;
goto op_err;
case WORKER_UPDATE:
if (cfg->random_value)
randomize_value(thread, value_buf);
asyncop->set_value(asyncop, value_buf);
if ((ret = asyncop->update(asyncop)) == 0)
break;
goto op_err;
default:
op_err: lprintf(cfg, ret, 0,
"%s failed for: %s, range: %"PRIu64,
op_name(op), key_buf, wtperf_value_range(cfg));
goto err; /* can't happen */
}
/* Schedule the next operation */
if (++op == op_end)
op = thread->workload->ops;
}
if (conn->async_flush(conn) != 0)
goto err;
/* Notify our caller we failed and shut the system down. */
if (0) {
err: cfg->error = cfg->stop = 1;
}
return (NULL);
}
static void *
worker(void *arg)
{
struct timespec start, stop, interval;
TRUNCATE_QUEUE_ENTRY *truncate_item;
CONFIG *cfg;
CONFIG_THREAD *thread;
TRACK *trk;
WT_CONNECTION *conn;
WT_CURSOR **cursors, *cursor, *tmp_cursor;
WT_SESSION *session;
int64_t ops, ops_per_txn, throttle_ops;
size_t i;
uint64_t final_milestone_gap, needed_milestones, next_val;
uint64_t num_milestones, last_key, start_point_val, end_point_val;
uint64_t expected_total, last_total_inserts, total_gross_inserts;
uint64_t truncate_milestone_gap, usecs;
uint8_t *op, *op_end;
int measure_latency, ret;
char *value_buf, *key, *key_buf, *truncate_key, *value;
char buf[512];
double truncation_percentage;
thread = (CONFIG_THREAD *)arg;
cfg = thread->cfg;
conn = cfg->conn;
cursors = NULL;
needed_milestones = num_milestones = 0;
ops = 0;
ops_per_txn = thread->workload->ops_per_txn;
last_key = 0;
session = NULL;
trk = NULL;
throttle_ops = 0;
truncate_milestone_gap = 0;
total_gross_inserts = last_total_inserts = expected_total = 0;
if ((ret = conn->open_session(
conn, NULL, cfg->sess_config, &session)) != 0) {
lprintf(cfg, ret, 0, "worker: WT_CONNECTION.open_session");
goto err;
}
cursors = calloc(cfg->table_count, sizeof(WT_CURSOR *));
if (cursors == NULL) {
lprintf(cfg, ENOMEM, 0,
"worker: couldn't allocate cursor array");
goto err;
}
for (i = 0; i < cfg->table_count_idle; i++) {
snprintf(buf, 512, "%s_idle%05d", cfg->uris[0], (int)i);
if ((ret = session->open_cursor(
session, buf, NULL, NULL, &tmp_cursor)) != 0) {
lprintf(cfg, ret, 0,
"Error opening idle table %s", buf);
goto err;
}
if ((ret = tmp_cursor->close(tmp_cursor)) != 0) {
lprintf(cfg, ret, 0,
"Error closing idle table %s", buf);
goto err;
}
}
for (i = 0; i < cfg->table_count; i++) {
if ((ret = session->open_cursor(session,
cfg->uris[i], NULL, NULL, &cursors[i])) != 0) {
lprintf(cfg, ret, 0,
"worker: WT_SESSION.open_cursor: %s",
cfg->uris[i]);
goto err;
}
}
/* Setup the timer for throttling. */
if (thread->workload->throttle != 0 &&
(ret = __wt_epoch(NULL, &interval)) != 0) {
lprintf(cfg, ret, 0, "Get time call failed");
goto err;
}
key_buf = thread->key_buf;
value_buf = thread->value_buf;
op = thread->workload->ops;
op_end = op + sizeof(thread->workload->ops);
if (ops_per_txn != 0 &&
(ret = session->begin_transaction(session, NULL)) != 0) {
lprintf(cfg, ret, 0, "First transaction begin failed");
goto err;
}
/* Setup for truncate */
if (thread->workload->truncate != 0) {
cursor = cursors[cfg->icount % cfg->table_count];
/* Truncation percentage value. eg 10% is 0.1 */
truncation_percentage =
(double)thread->workload->truncate_pct / 100;
/* How many entries between each milestone */
truncate_milestone_gap =
thread->workload->truncate_count*truncation_percentage;
/* How many milestones we need */
needed_milestones = thread->workload->truncate_count /
truncate_milestone_gap;
final_milestone_gap = truncate_milestone_gap;
/* Reset this value for use again */
truncate_milestone_gap = 0;
cursor->next(cursor);
ret = cursor->get_key(cursor, &key);
/* We have data */
if (ret == 0) {
start_point_val = decode_key(cfg, key);
cursor->reset(cursor);
if ((ret = cursor->prev(cursor)) != 0) {
lprintf(cfg, ret, 0,
"truncate setup end: prev failed");
goto err;
}
if ((ret = cursor->get_key(cursor,
&key)) != 0) {
lprintf(cfg, ret, 0,
"truncate setup end: get_key failed");
goto err;
}
end_point_val = decode_key(cfg, key);
/* Not enough documents? */
if (start_point_val + needed_milestones
> end_point_val)
truncate_milestone_gap = 0;
else
truncate_milestone_gap =
(end_point_val - start_point_val)
/ needed_milestones;
} else if ( ret != WT_NOTFOUND ) {
lprintf(cfg, ret, 0,
"truncate setup start: get_key failed");
goto err;
}
/* If we have enough data alloc some milestones */
if (truncate_milestone_gap != 0) {
expected_total = (end_point_val - start_point_val);
for (i = 0; i < needed_milestones; i++) {
truncate_key = calloc(cfg->key_sz, 1);
truncate_item =
calloc(sizeof(TRUNCATE_QUEUE_ENTRY), 1);
if (truncate_item == NULL) {
(void)enomem(cfg);
goto err;
}
generate_key(cfg, truncate_key,
truncate_milestone_gap * (i+1));
truncate_item->key = truncate_key;
truncate_item->diff =
(truncate_milestone_gap * (i+1)) - last_key;
STAILQ_INSERT_TAIL( &cfg->truncate_stone_head,
truncate_item, q);
last_key = truncate_milestone_gap * (i+1);
num_milestones++;
}
}
truncate_milestone_gap = final_milestone_gap;
}
while (!cfg->stop) {
/*
* Generate the next key and setup operation specific
* statistics tracking objects.
*/
switch (*op) {
case WORKER_INSERT:
case WORKER_INSERT_RMW:
trk = &thread->insert;
if (cfg->random_range)
next_val = wtperf_rand(thread);
else
next_val = cfg->icount + get_next_incr(cfg);
break;
case WORKER_READ:
trk = &thread->read;
/* FALLTHROUGH */
case WORKER_UPDATE:
if (*op == WORKER_UPDATE)
trk = &thread->update;
next_val = wtperf_rand(thread);
/*
* If the workload is started without a populate phase
* we rely on at least one insert to get a valid item
* id.
*/
if (wtperf_value_range(cfg) < next_val)
continue;
break;
case WORKER_TRUNCATE:
total_gross_inserts = 0;
next_val = wtperf_rand(thread);
for (i=0; i < thread->cfg->workers_cnt; i++) {
total_gross_inserts +=
thread->cfg->workers[i].total_inserts;
}
break;
default:
goto err; /* can't happen */
}
generate_key(cfg, key_buf, next_val);
/*
* Spread the data out around the multiple databases.
*/
cursor = cursors[next_val % cfg->table_count];
/*
* Skip the first time we do an operation, when trk->ops
* is 0, to avoid first time latency spikes.
*/
measure_latency =
cfg->sample_interval != 0 && trk->ops != 0 && (
trk->ops % cfg->sample_rate == 0);
if (measure_latency &&
(ret = __wt_epoch(NULL, &start)) != 0) {
lprintf(cfg, ret, 0, "Get time call failed");
goto err;
}
cursor->set_key(cursor, key_buf);
switch (*op) {
case WORKER_READ:
/*
* Reads can fail with WT_NOTFOUND: we may be searching
* in a random range, or an insert thread might have
* updated the last record in the table but not yet
* finished the actual insert. Count failed search in
* a random range as a "read".
*/
ret = cursor->search(cursor);
if (ret == 0) {
if ((ret = cursor->get_value(
cursor, &value)) != 0) {
lprintf(cfg, ret, 0,
"get_value in read.");
goto err;
}
}
if (ret == 0 || ret == WT_NOTFOUND)
break;
goto op_err;
case WORKER_INSERT_RMW:
if ((ret = cursor->search(cursor)) != WT_NOTFOUND)
goto op_err;
/* The error return reset the cursor's key. */
cursor->set_key(cursor, key_buf);
/* FALLTHROUGH */
case WORKER_INSERT:
if (cfg->random_value)
randomize_value(thread, value_buf);
cursor->set_value(cursor, value_buf);
if ((ret = cursor->insert(cursor)) == 0) {
thread->total_inserts++;
break;
}
goto op_err;
case WORKER_TRUNCATE:
trk = &thread->truncate;
expected_total +=
(total_gross_inserts - last_total_inserts);
last_total_inserts = total_gross_inserts;
/* We have enough data */
if (expected_total > needed_milestones) {
while (num_milestones < needed_milestones) {
truncate_key = calloc(cfg->key_sz, 1);
truncate_item =
calloc(
sizeof(TRUNCATE_QUEUE_ENTRY), 1);
if (truncate_item == NULL) {
(void)enomem(cfg);
goto op_err;
}
generate_key(cfg,
truncate_key, last_key);
truncate_item->key = truncate_key;
truncate_item->diff =
truncate_milestone_gap;
STAILQ_INSERT_TAIL(
&cfg->truncate_stone_head,
truncate_item, q);
num_milestones++;
last_key += truncate_milestone_gap;
}
/* We have too much data, we need to truncate */
if (expected_total >
thread->workload->truncate_count &&
num_milestones > 0) {
truncate_item =
STAILQ_FIRST(
&cfg->truncate_stone_head);
num_milestones--;
STAILQ_REMOVE_HEAD(
&cfg->truncate_stone_head, q);
cursor->set_key(cursor,
truncate_item->key);
cursor->search(cursor);
ret = session->truncate(session,
NULL, NULL, cursor, NULL);
expected_total -= truncate_item->diff;
if ( ret != 0) {
lprintf(cfg, ret, 0,
"Truncate failed");
goto op_err;
}
free(truncate_item->key);
free(truncate_item);
truncate_item = NULL;
} else {
trk = &thread->truncate_sleep;
(void)usleep(1000);
}
} else {
trk = &thread->truncate_sleep;
(void)usleep(1000);
}
break;
case WORKER_UPDATE:
if ((ret = cursor->search(cursor)) == 0) {
if ((ret = cursor->get_value(
cursor, &value)) != 0) {
lprintf(cfg, ret, 0,
"get_value in update.");
goto err;
}
/*
* Copy as much of the previous value as is
* safe, and be sure to NUL-terminate.
*/
strncpy(value_buf, value, cfg->value_sz);
value_buf[cfg->value_sz - 1] = '\0';
if (value_buf[0] == 'a')
value_buf[0] = 'b';
else
value_buf[0] = 'a';
if (cfg->random_value)
randomize_value(thread, value_buf);
cursor->set_value(cursor, value_buf);
if ((ret = cursor->update(cursor)) == 0)
break;
goto op_err;
}
/*
* Reads can fail with WT_NOTFOUND: we may be searching
* in a random range, or an insert thread might have
* updated the last record in the table but not yet
* finished the actual insert. Count failed search in
* a random range as a "read".
*/
if (ret == WT_NOTFOUND)
break;
op_err: lprintf(cfg, ret, 0,
"%s failed for: %s, range: %"PRIu64,
op_name(op), key_buf, wtperf_value_range(cfg));
goto err;
default:
goto err; /* can't happen */
}
/* Release the cursor, if we have multiple tables. */
if (cfg->table_count > 1 && ret == 0 &&
*op != WORKER_INSERT && *op != WORKER_INSERT_RMW) {
if ((ret = cursor->reset(cursor)) != 0) {
lprintf(cfg, ret, 0, "Cursor reset failed");
goto err;
}
}
/* Gather statistics */
if (!cfg->in_warmup) {
if (measure_latency) {
if ((ret = __wt_epoch(NULL, &stop)) != 0) {
lprintf(cfg, ret, 0,
"Get time call failed");
goto err;
}
++trk->latency_ops;
usecs = ns_to_us(WT_TIMEDIFF(stop, start));
track_operation(trk, usecs);
}
/* Increment operation count */
++trk->ops;
}
/* Commit our work if configured for explicit transactions */
if (ops_per_txn != 0 && ops++ % ops_per_txn == 0) {
if ((ret = session->commit_transaction(
session, NULL)) != 0) {
lprintf(cfg, ret, 0,
"Worker transaction commit failed");
goto err;
}
if ((ret = session->begin_transaction(
session, NULL)) != 0) {
lprintf(cfg, ret, 0,
"Worker transaction commit failed");
goto err;
}
}
/* Schedule the next operation */
if (++op == op_end)
op = thread->workload->ops;
/*
* Check throttling periodically to avoid taking too
* many time samples.
*/
if (thread->workload->throttle != 0 &&
throttle_ops++ % THROTTLE_OPS == 0)
worker_throttle(thread->workload->throttle,
&throttle_ops, &interval);
}
if ((ret = session->close(session, NULL)) != 0) {
lprintf(cfg, ret, 0, "Session close in worker failed");
goto err;
}
/* Notify our caller we failed and shut the system down. */
if (0) {
err: cfg->error = cfg->stop = 1;
}
/* Empty the truncate queue before close */
if (thread->workload->truncate != 0) {
while (!STAILQ_EMPTY(&cfg->truncate_stone_head)) {
truncate_item = STAILQ_FIRST(&cfg->truncate_stone_head);
STAILQ_REMOVE_HEAD(&cfg->truncate_stone_head, q);
free(truncate_item->key);
free(truncate_item);
}
}
free(cursors);
return (NULL);
}
/*
* run_mix_schedule_op --
* Replace read operations with another operation, in the configured
* percentage.
*/
static void
run_mix_schedule_op(WORKLOAD *workp, int op, int64_t op_cnt)
{
int jump, pass;
uint8_t *p, *end;
/* Jump around the array to roughly spread out the operations. */
jump = 100 / op_cnt;
/*
* Find a read operation and replace it with another operation. This
* is roughly n-squared, but it's an N of 100, leave it.
*/
p = workp->ops;
end = workp->ops + sizeof(workp->ops);
while (op_cnt-- > 0) {
for (pass = 0; *p != WORKER_READ; ++p)
if (p == end) {
/*
* Passed a percentage of total operations and
* should always be a read operation to replace,
* but don't allow infinite loops.
*/
if (++pass > 1)
return;
p = workp->ops;
}
*p = (uint8_t)op;
if (end - jump < p)
p = workp->ops;
else
p += jump;
}
}
/*
* run_mix_schedule --
* Schedule the mixed-run operations.
*/
static int
run_mix_schedule(CONFIG *cfg, WORKLOAD *workp)
{
int64_t pct;
/* Confirm reads, inserts and updates cannot all be zero. */
if (workp->insert == 0 && workp->read == 0 &&
workp->update == 0 && workp->truncate == 0) {
lprintf(cfg, EINVAL, 0, "no operations scheduled");
return (EINVAL);
}
/*
* Check for a simple case where the thread is only doing insert or
* update operations (because the default operation for a job-mix is
* read, the subsequent code works fine if only reads are specified).
*/
if (workp->insert != 0 && workp->read == 0 &&
workp->update == 0 && workp->truncate == 0) {
memset(workp->ops,
cfg->insert_rmw ? WORKER_INSERT_RMW : WORKER_INSERT,
sizeof(workp->ops));
return (0);
}
if (workp->insert == 0 && workp->read == 0 &&
workp->truncate == 0 && workp->update != 0) {
memset(workp->ops, WORKER_UPDATE, sizeof(workp->ops));
return (0);
}
if (workp->insert == 0 && workp->read == 0 &&
workp->update == 0 && workp->truncate != 0) {
memset(workp->ops, WORKER_TRUNCATE, sizeof(workp->ops));
return (0);
}
/*
* The worker thread configuration is done as ratios of operations. If
* the caller gives us something insane like "reads=77,updates=23" (do
* 77 reads for every 23 updates), we don't want to do 77 reads followed
* by 23 updates, we want to uniformly distribute the read and update
* operations across the space. Convert to percentages and then lay out
* the operations across an array.
*
* Percentage conversion is lossy, the application can do stupid stuff
* here, for example, imagine a configured ratio of "reads=1,inserts=2,
* updates=999999". First, if the percentages are skewed enough, some
* operations might never be done. Second, we set the base operation to
* read, which means any fractional results from percentage conversion
* will be reads, implying read operations in some cases where reads
* weren't configured. We should be fine if the application configures
* something approaching a rational set of ratios.
*/
memset(workp->ops, WORKER_READ, sizeof(workp->ops));
pct = (workp->insert * 100) /
(workp->insert + workp->read + workp->update + workp->truncate);
if (pct != 0)
run_mix_schedule_op(workp,
cfg->insert_rmw ? WORKER_INSERT_RMW : WORKER_INSERT, pct);
pct = (workp->update * 100) /
(workp->insert + workp->read + workp->update + workp->truncate);
if (pct != 0)
run_mix_schedule_op(workp, WORKER_UPDATE, pct);
pct = (workp->truncate * 100) /
(workp->insert + workp->read + workp->update + workp->truncate);
if (pct != 0)
run_mix_schedule_op(workp, WORKER_TRUNCATE, pct);
return (0);
}
static void *
populate_thread(void *arg)
{
struct timespec start, stop;
CONFIG *cfg;
CONFIG_THREAD *thread;
TRACK *trk;
WT_CONNECTION *conn;
WT_CURSOR **cursors, *cursor;
WT_SESSION *session;
size_t i;
uint64_t op, usecs;
uint32_t opcount;
int intxn, measure_latency, ret, stress_checkpoint_due;
char *value_buf, *key_buf;
const char *cursor_config;
thread = (CONFIG_THREAD *)arg;
cfg = thread->cfg;
conn = cfg->conn;
session = NULL;
cursors = NULL;
ret = stress_checkpoint_due = 0;
trk = &thread->insert;
key_buf = thread->key_buf;
value_buf = thread->value_buf;
if ((ret = conn->open_session(
conn, NULL, cfg->sess_config, &session)) != 0) {
lprintf(cfg, ret, 0, "populate: WT_CONNECTION.open_session");
goto err;
}
/* Do bulk loads if populate is single-threaded. */
cursor_config =
(cfg->populate_threads == 1 && !cfg->index) ? "bulk" : NULL;
/* Create the cursors. */
cursors = calloc(cfg->table_count, sizeof(WT_CURSOR *));
if (cursors == NULL) {
lprintf(cfg, ENOMEM, 0,
"worker: couldn't allocate cursor array");
goto err;
}
for (i = 0; i < cfg->table_count; i++) {
if ((ret = session->open_cursor(
session, cfg->uris[i], NULL,
cursor_config, &cursors[i])) != 0) {
lprintf(cfg, ret, 0,
"populate: WT_SESSION.open_cursor: %s",
cfg->uris[i]);
goto err;
}
}
/* Populate the databases. */
for (intxn = 0, opcount = 0;;) {
op = get_next_incr(cfg);
if (op > cfg->icount)
break;
if (cfg->populate_ops_per_txn != 0 && !intxn) {
if ((ret = session->begin_transaction(
session, cfg->transaction_config)) != 0) {
lprintf(cfg, ret, 0,
"Failed starting transaction.");
goto err;
}
intxn = 1;
}
/*
* Figure out which table this op belongs to.
*/
cursor = cursors[op % cfg->table_count];
generate_key(cfg, key_buf, op);
measure_latency =
cfg->sample_interval != 0 && trk->ops != 0 && (
trk->ops % cfg->sample_rate == 0);
if (measure_latency &&
(ret = __wt_epoch(NULL, &start)) != 0) {
lprintf(cfg, ret, 0, "Get time call failed");
goto err;
}
cursor->set_key(cursor, key_buf);
if (cfg->random_value)
randomize_value(thread, value_buf);
cursor->set_value(cursor, value_buf);
if ((ret = cursor->insert(cursor)) == WT_ROLLBACK) {
lprintf(cfg, ret, 0, "insert retrying");
if ((ret = session->rollback_transaction(
session, NULL)) != 0) {
lprintf(cfg, ret, 0,
"Failed rollback_transaction");
goto err;
}
intxn = 0;
continue;
} else if (ret != 0) {
lprintf(cfg, ret, 0, "Failed inserting");
goto err;
}
/*
* Gather statistics.
* We measure the latency of inserting a single key. If there
* are multiple tables, it is the time for insertion into all
* of them.
*/
if (measure_latency) {
if ((ret = __wt_epoch(NULL, &stop)) != 0) {
lprintf(cfg, ret, 0, "Get time call failed");
goto err;
}
++trk->latency_ops;
usecs = ns_to_us(WT_TIMEDIFF(stop, start));
track_operation(trk, usecs);
}
++thread->insert.ops; /* Same as trk->ops */
if (cfg->checkpoint_stress_rate != 0 &&
(op % cfg->checkpoint_stress_rate) == 0)
stress_checkpoint_due = 1;
if (cfg->populate_ops_per_txn != 0) {
if (++opcount < cfg->populate_ops_per_txn)
continue;
opcount = 0;
if ((ret = session->commit_transaction(
session, NULL)) != 0)
lprintf(cfg, ret, 0,
"Fail committing, transaction was aborted");
intxn = 0;
}
if (stress_checkpoint_due && intxn == 0) {
stress_checkpoint_due = 0;
if ((ret = session->checkpoint(session, NULL)) != 0) {
lprintf(cfg, ret, 0, "Checkpoint failed");
goto err;
}
}
}
if (intxn &&
(ret = session->commit_transaction(session, NULL)) != 0)
lprintf(cfg, ret, 0,
"Fail committing, transaction was aborted");
if ((ret = session->close(session, NULL)) != 0) {
lprintf(cfg, ret, 0, "Error closing session in populate");
goto err;
}
/* Notify our caller we failed and shut the system down. */
if (0) {
err: cfg->error = cfg->stop = 1;
}
free(cursors);
return (NULL);
}
static void *
populate_async(void *arg)
{
struct timespec start, stop;
CONFIG *cfg;
CONFIG_THREAD *thread;
TRACK *trk;
WT_ASYNC_OP *asyncop;
WT_CONNECTION *conn;
WT_SESSION *session;
uint64_t op, usecs;
int measure_latency, ret;
char *value_buf, *key_buf;
thread = (CONFIG_THREAD *)arg;
cfg = thread->cfg;
conn = cfg->conn;
session = NULL;
ret = 0;
trk = &thread->insert;
key_buf = thread->key_buf;
value_buf = thread->value_buf;
if ((ret = conn->open_session(
conn, NULL, cfg->sess_config, &session)) != 0) {
lprintf(cfg, ret, 0, "populate: WT_CONNECTION.open_session");
goto err;
}
/*
* Measuring latency of one async op is not meaningful. We
* will measure the time it takes to do all of them, including
* the time to process by workers.
*/
measure_latency =
cfg->sample_interval != 0 && trk->ops != 0 && (
trk->ops % cfg->sample_rate == 0);
if (measure_latency &&
(ret = __wt_epoch(NULL, &start)) != 0) {
lprintf(cfg, ret, 0, "Get time call failed");
goto err;
}
/* Populate the databases. */
for (;;) {
op = get_next_incr(cfg);
if (op > cfg->icount)
break;
/*
* Allocate an async op for whichever table.
*/
while ((ret = conn->async_new_op(
conn, cfg->uris[op % cfg->table_count],
NULL, &cb, &asyncop)) == EBUSY)
(void)usleep(10000);
if (ret != 0)
goto err;
asyncop->app_private = thread;
generate_key(cfg, key_buf, op);
asyncop->set_key(asyncop, key_buf);
if (cfg->random_value)
randomize_value(thread, value_buf);
asyncop->set_value(asyncop, value_buf);
if ((ret = asyncop->insert(asyncop)) != 0) {
lprintf(cfg, ret, 0, "Failed inserting");
goto err;
}
}
/*
* Gather statistics.
* We measure the latency of inserting a single key. If there
* are multiple tables, it is the time for insertion into all
* of them. Note that currently every populate thread will call
* async_flush and those calls will convoy. That is not the
* most efficient way, but we want to flush before measuring latency.
*/
if (conn->async_flush(conn) != 0)
goto err;
if (measure_latency) {
if ((ret = __wt_epoch(NULL, &stop)) != 0) {
lprintf(cfg, ret, 0,
"Get time call failed");
goto err;
}
++trk->latency_ops;
usecs = ns_to_us(WT_TIMEDIFF(stop, start));
track_operation(trk, usecs);
}
if ((ret = session->close(session, NULL)) != 0) {
lprintf(cfg, ret, 0, "Error closing session in populate");
goto err;
}
/* Notify our caller we failed and shut the system down. */
if (0) {
err: cfg->error = cfg->stop = 1;
}
return (NULL);
}
static void *
monitor(void *arg)
{
struct timespec t;
struct tm *tm, _tm;
CONFIG *cfg;
FILE *fp;
size_t len;
uint64_t min_thr, reads, inserts, updates;
uint64_t cur_reads, cur_inserts, cur_updates;
uint64_t last_reads, last_inserts, last_updates;
uint32_t read_avg, read_min, read_max;
uint32_t insert_avg, insert_min, insert_max;
uint32_t update_avg, update_min, update_max;
uint32_t latency_max;
u_int i;
int ret;
char buf[64], *path;
cfg = (CONFIG *)arg;
assert(cfg->sample_interval != 0);
fp = NULL;
path = NULL;
min_thr = (uint64_t)cfg->min_throughput;
latency_max = (uint32_t)ms_to_us(cfg->max_latency);
/* Open the logging file. */
len = strlen(cfg->monitor_dir) + 100;
if ((path = malloc(len)) == NULL) {
(void)enomem(cfg);
goto err;
}
snprintf(path, len, "%s/monitor", cfg->monitor_dir);
if ((fp = fopen(path, "w")) == NULL) {
lprintf(cfg, errno, 0, "%s", path);
goto err;
}
/* Set line buffering for monitor file. */
(void)setvbuf(fp, NULL, _IOLBF, 0);
fprintf(fp,
"#time,"
"totalsec,"
"read ops per second,"
"insert ops per second,"
"update ops per second,"
"checkpoints,"
"read average latency(uS),"
"read minimum latency(uS),"
"read maximum latency(uS),"
"insert average latency(uS),"
"insert min latency(uS),"
"insert maximum latency(uS),"
"update average latency(uS),"
"update min latency(uS),"
"update maximum latency(uS)"
"\n");
last_reads = last_inserts = last_updates = 0;
while (!cfg->stop) {
for (i = 0; i < cfg->sample_interval; i++) {
sleep(1);
if (cfg->stop)
break;
}
/* If the workers are done, don't bother with a final call. */
if (cfg->stop)
break;
if (cfg->in_warmup)
continue;
if ((ret = __wt_epoch(NULL, &t)) != 0) {
lprintf(cfg, ret, 0, "Get time call failed");
goto err;
}
tm = localtime_r(&t.tv_sec, &_tm);
(void)strftime(buf, sizeof(buf), "%b %d %H:%M:%S", tm);
reads = sum_read_ops(cfg);
inserts = sum_insert_ops(cfg);
updates = sum_update_ops(cfg);
latency_read(cfg, &read_avg, &read_min, &read_max);
latency_insert(cfg, &insert_avg, &insert_min, &insert_max);
latency_update(cfg, &update_avg, &update_min, &update_max);
cur_reads = (reads - last_reads) / cfg->sample_interval;
cur_updates = (updates - last_updates) / cfg->sample_interval;
/*
* For now the only item we need to worry about changing is
* inserts when we transition from the populate phase to
* workload phase.
*/
if (inserts < last_inserts)
cur_inserts = 0;
else
cur_inserts =
(inserts - last_inserts) / cfg->sample_interval;
(void)fprintf(fp,
"%s,%" PRIu32
",%" PRIu64 ",%" PRIu64 ",%" PRIu64
",%c"
",%" PRIu32 ",%" PRIu32 ",%" PRIu32
",%" PRIu32 ",%" PRIu32 ",%" PRIu32
",%" PRIu32 ",%" PRIu32 ",%" PRIu32
"\n",
buf, cfg->totalsec,
cur_reads, cur_inserts, cur_updates,
cfg->ckpt ? 'Y' : 'N',
read_avg, read_min, read_max,
insert_avg, insert_min, insert_max,
update_avg, update_min, update_max);
if (latency_max != 0 &&
(read_max > latency_max || insert_max > latency_max ||
update_max > latency_max))
lprintf(cfg, WT_PANIC, 0,
"max latency exceeded: threshold %" PRIu32
" read max %" PRIu32 " insert max %" PRIu32
" update max %" PRIu32, latency_max,
read_max, insert_max, update_max);
if (min_thr != 0 &&
((cur_reads != 0 && cur_reads < min_thr) ||
(cur_inserts != 0 && cur_inserts < min_thr) ||
(cur_updates != 0 && cur_updates < min_thr)))
lprintf(cfg, WT_PANIC, 0,
"minimum throughput not met: threshold %" PRIu64
" reads %" PRIu64 " inserts %" PRIu64
" updates %" PRIu64, min_thr, cur_reads,
cur_inserts, cur_updates);
last_reads = reads;
last_inserts = inserts;
last_updates = updates;
}
/* Notify our caller we failed and shut the system down. */
if (0) {
err: cfg->error = cfg->stop = 1;
}
if (fp != NULL)
(void)fclose(fp);
free(path);
return (NULL);
}
static void *
checkpoint_worker(void *arg)
{
CONFIG *cfg;
CONFIG_THREAD *thread;
WT_CONNECTION *conn;
WT_SESSION *session;
struct timespec e, s;
uint32_t i;
int ret;
thread = (CONFIG_THREAD *)arg;
cfg = thread->cfg;
conn = cfg->conn;
session = NULL;
if ((ret = conn->open_session(
conn, NULL, cfg->sess_config, &session)) != 0) {
lprintf(cfg, ret, 0,
"open_session failed in checkpoint thread.");
goto err;
}
while (!cfg->stop) {
/* Break the sleep up, so we notice interrupts faster. */
for (i = 0; i < cfg->checkpoint_interval; i++) {
sleep(1);
if (cfg->stop)
break;
}
/* If the workers are done, don't bother with a final call. */
if (cfg->stop)
break;
if ((ret = __wt_epoch(NULL, &s)) != 0) {
lprintf(cfg, ret, 0, "Get time failed in checkpoint.");
goto err;
}
cfg->ckpt = 1;
if ((ret = session->checkpoint(session, NULL)) != 0) {
lprintf(cfg, ret, 0, "Checkpoint failed.");
goto err;
}
cfg->ckpt = 0;
++thread->ckpt.ops;
if ((ret = __wt_epoch(NULL, &e)) != 0) {
lprintf(cfg, ret, 0, "Get time failed in checkpoint.");
goto err;
}
}
if (session != NULL &&
((ret = session->close(session, NULL)) != 0)) {
lprintf(cfg, ret, 0,
"Error closing session in checkpoint worker.");
goto err;
}
/* Notify our caller we failed and shut the system down. */
if (0) {
err: cfg->error = cfg->stop = 1;
}
return (NULL);
}
static int
execute_populate(CONFIG *cfg)
{
struct timespec start, stop;
CONFIG_THREAD *popth;
WT_ASYNC_OP *asyncop;
size_t i;
uint64_t last_ops, msecs;
uint32_t interval, tables;
int elapsed, ret;
void *(*pfunc)(void *);
lprintf(cfg, 0, 1,
"Starting %" PRIu32
" populate thread(s) for %" PRIu32 " items",
cfg->populate_threads, cfg->icount);
cfg->insert_key = 0;
if ((cfg->popthreads =
calloc(cfg->populate_threads, sizeof(CONFIG_THREAD))) == NULL)
return (enomem(cfg));
if (cfg->use_asyncops > 0) {
lprintf(cfg, 0, 1, "Starting %" PRIu32 " async thread(s)",
cfg->async_threads);
pfunc = populate_async;
} else
pfunc = populate_thread;
if ((ret = start_threads(cfg, NULL,
cfg->popthreads, cfg->populate_threads, pfunc)) != 0)
return (ret);
if ((ret = __wt_epoch(NULL, &start)) != 0) {
lprintf(cfg, ret, 0, "Get time failed in populate.");
return (ret);
}
for (elapsed = 0, interval = 0, last_ops = 0;
cfg->insert_key < cfg->icount && cfg->error == 0;) {
/*
* Sleep for 100th of a second, report_interval is in second
* granularity, each 100th increment of elapsed is a single
* increment of interval.
*/
(void)usleep(10000);
if (cfg->report_interval == 0 || ++elapsed < 100)
continue;
elapsed = 0;
if (++interval < cfg->report_interval)
continue;
interval = 0;
cfg->totalsec += cfg->report_interval;
cfg->insert_ops = sum_pop_ops(cfg);
lprintf(cfg, 0, 1,
"%" PRIu64 " populate inserts (%" PRIu64 " of %"
PRIu32 ") in %" PRIu32 " secs (%" PRIu32 " total secs)",
cfg->insert_ops - last_ops, cfg->insert_ops,
cfg->icount, cfg->report_interval, cfg->totalsec);
last_ops = cfg->insert_ops;
}
if ((ret = __wt_epoch(NULL, &stop)) != 0) {
lprintf(cfg, ret, 0, "Get time failed in populate.");
return (ret);
}
/*
* Move popthreads aside to narrow possible race with the monitor
* thread. The latency tracking code also requires that popthreads be
* NULL when the populate phase is finished, to know that the workload
* phase has started.
*/
popth = cfg->popthreads;
cfg->popthreads = NULL;
ret = stop_threads(cfg, cfg->populate_threads, popth);
free(popth);
if (ret != 0)
return (ret);
/* Report if any worker threads didn't finish. */
if (cfg->error != 0) {
lprintf(cfg, WT_ERROR, 0,
"Populate thread(s) exited without finishing.");
return (WT_ERROR);
}
lprintf(cfg, 0, 1, "Finished load of %" PRIu32 " items", cfg->icount);
msecs = ns_to_ms(WT_TIMEDIFF(stop, start));
lprintf(cfg, 0, 1,
"Load time: %.2f\n" "load ops/sec: %" PRIu64,
(double)msecs / (double)MSEC_PER_SEC,
(uint64_t)((cfg->icount / msecs) / MSEC_PER_SEC));
/*
* If configured, compact to allow LSM merging to complete. We
* set an unlimited timeout because if we close the connection
* then any in-progress compact/merge is aborted.
*/
if (cfg->compact) {
assert(cfg->async_threads > 0);
lprintf(cfg, 0, 1, "Compact after populate");
if ((ret = __wt_epoch(NULL, &start)) != 0) {
lprintf(cfg, ret, 0, "Get time failed in populate.");
return (ret);
}
tables = cfg->table_count;
for (i = 0; i < cfg->table_count; i++) {
/*
* If no ops are available, retry. Any other error,
* return.
*/
while ((ret = cfg->conn->async_new_op(cfg->conn,
cfg->uris[i], "timeout=0", &cb, &asyncop)) == EBUSY)
(void)usleep(10000);
if (ret != 0)
return (ret);
asyncop->app_private = &tables;
if ((ret = asyncop->compact(asyncop)) != 0) {
lprintf(cfg, ret, 0, "Async compact failed.");
return (ret);
}
}
if ((ret = cfg->conn->async_flush(cfg->conn)) != 0) {
lprintf(cfg, ret, 0, "Populate async flush failed.");
return (ret);
}
if ((ret = __wt_epoch(NULL, &stop)) != 0) {
lprintf(cfg, ret, 0, "Get time failed in populate.");
return (ret);
}
lprintf(cfg, 0, 1,
"Compact completed in %" PRIu64 " seconds",
(uint64_t)(ns_to_sec(WT_TIMEDIFF(stop, start))));
assert(tables == 0);
}
return (0);
}
static int
close_reopen(CONFIG *cfg)
{
int ret;
/*
* Reopen the connection. We do this so that the workload phase always
* starts with the on-disk files, and so that read-only workloads can
* be identified. This is particularly important for LSM, where the
* merge algorithm is more aggressive for read-only trees.
*/
/* cfg->conn is released no matter the return value from close(). */
ret = cfg->conn->close(cfg->conn, NULL);
cfg->conn = NULL;
if (ret != 0) {
lprintf(cfg, ret, 0, "Closing the connection failed");
return (ret);
}
if ((ret = wiredtiger_open(
cfg->home, NULL, cfg->conn_config, &cfg->conn)) != 0) {
lprintf(cfg, ret, 0, "Re-opening the connection failed");
return (ret);
}
/*
* If we started async threads only for the purposes of compact,
* then turn it off before starting the workload so that those extra
* threads looking for work that will never arrive don't affect
* performance.
*/
if (cfg->compact && cfg->use_asyncops == 0) {
if ((ret = cfg->conn->reconfigure(
cfg->conn, "async=(enabled=false)")) != 0) {
lprintf(cfg, ret, 0, "Reconfigure async off failed");
return (ret);
}
}
return (0);
}
static int
execute_workload(CONFIG *cfg)
{
CONFIG_THREAD *threads;
WORKLOAD *workp;
uint64_t last_ckpts, last_inserts, last_reads, last_truncates;
uint64_t last_updates;
uint32_t interval, run_ops, run_time;
u_int i;
int ret, t_ret;
void *(*pfunc)(void *);
cfg->insert_key = 0;
cfg->insert_ops = cfg->read_ops =
cfg->truncate_ops = cfg->update_ops = 0;
last_ckpts = last_inserts =
last_reads =last_truncates = last_updates = 0;
ret = 0;
if (cfg->warmup != 0)
cfg->in_warmup = 1;
/* Allocate memory for the worker threads. */
if ((cfg->workers =
calloc((size_t)cfg->workers_cnt, sizeof(CONFIG_THREAD))) == NULL) {
ret = enomem(cfg);
goto err;
}
if (cfg->use_asyncops > 0) {
lprintf(cfg, 0, 1, "Starting %" PRIu32 " async thread(s)",
cfg->async_threads);
pfunc = worker_async;
} else
pfunc = worker;
/* Start each workload. */
for (threads = cfg->workers, i = 0,
workp = cfg->workload; i < cfg->workload_cnt; ++i, ++workp) {
lprintf(cfg, 0, 1,
"Starting workload #%d: %" PRId64 " threads, inserts=%"
PRId64 ", reads=%" PRId64 ", updates=%" PRId64
", truncate=%" PRId64, i + 1, workp->threads, workp->insert,
workp->read, workp->update, workp->truncate);
/* Figure out the workload's schedule. */
if ((ret = run_mix_schedule(cfg, workp)) != 0)
goto err;
/* Allocate the truncate threads queue */
if (workp->truncate > 0)
STAILQ_INIT(&cfg->truncate_stone_head);
/* Start the workload's threads. */
if ((ret = start_threads(
cfg, workp, threads, (u_int)workp->threads, pfunc)) != 0)
goto err;
threads += workp->threads;
}
if (cfg->warmup != 0) {
lprintf(cfg, 0, 1,
"Waiting for warmup duration of %" PRIu32, cfg->warmup);
sleep(cfg->warmup);
cfg->in_warmup = 0;
}
for (interval = cfg->report_interval, run_time = cfg->run_time,
run_ops = cfg->run_ops; cfg->error == 0;) {
/*
* Sleep for one second at a time.
* If we are tracking run time, check to see if we're done, and
* if we're only tracking run time, go back to sleep.
*/
sleep(1);
if (run_time != 0) {
if (--run_time == 0)
break;
if (!interval && !run_ops)
continue;
}
/* Sum the operations we've done. */
cfg->ckpt_ops = sum_ckpt_ops(cfg);
cfg->insert_ops = sum_insert_ops(cfg);
cfg->read_ops = sum_read_ops(cfg);
cfg->update_ops = sum_update_ops(cfg);
cfg->truncate_ops = sum_truncate_ops(cfg);
/* If we're checking total operations, see if we're done. */
if (run_ops != 0 && run_ops <=
cfg->insert_ops + cfg->read_ops + cfg->update_ops)
break;
/* If writing out throughput information, see if it's time. */
if (interval == 0 || --interval > 0)
continue;
interval = cfg->report_interval;
cfg->totalsec += cfg->report_interval;
lprintf(cfg, 0, 1,
"%" PRIu64 " reads, %" PRIu64 " inserts, %" PRIu64
" updates, %" PRIu64 " truncates, %" PRIu64
" checkpoints in %" PRIu32 " secs (%" PRIu32 " total secs)",
cfg->read_ops - last_reads,
cfg->insert_ops - last_inserts,
cfg->update_ops - last_updates,
cfg->truncate_ops - last_truncates,
cfg->ckpt_ops - last_ckpts,
cfg->report_interval, cfg->totalsec);
last_reads = cfg->read_ops;
last_inserts = cfg->insert_ops;
last_updates = cfg->update_ops;
last_truncates = cfg->truncate_ops;
last_ckpts = cfg->ckpt_ops;
}
/* Notify the worker threads they are done. */
err: cfg->stop = 1;
if ((t_ret = stop_threads(
cfg, (u_int)cfg->workers_cnt, cfg->workers)) != 0 && ret == 0)
ret = t_ret;
/* Drop tables if configured to and this isn't an error path */
if (ret == 0 && cfg->drop_tables && (ret = drop_all_tables(cfg)) != 0)
lprintf(cfg, ret, 0, "Drop tables failed.");
/* Report if any worker threads didn't finish. */
if (cfg->error != 0) {
lprintf(cfg, WT_ERROR, 0,
"Worker thread(s) exited without finishing.");
if (ret == 0)
ret = WT_ERROR;
}
return (ret);
}
/*
* Ensure that icount matches the number of records in the
* existing table.
*/
static int
find_table_count(CONFIG *cfg)
{
WT_CONNECTION *conn;
WT_CURSOR *cursor;
WT_SESSION *session;
uint32_t i, max_icount, table_icount;
int ret, t_ret;
char *key;
conn = cfg->conn;
max_icount = 0;
if ((ret = conn->open_session(
conn, NULL, cfg->sess_config, &session)) != 0) {
lprintf(cfg, ret, 0,
"find_table_count: open_session failed");
goto out;
}
for (i = 0; i < cfg->table_count; i++) {
if ((ret = session->open_cursor(session, cfg->uris[i],
NULL, NULL, &cursor)) != 0) {
lprintf(cfg, ret, 0,
"find_table_count: open_cursor failed");
goto err;
}
if ((ret = cursor->prev(cursor)) != 0) {
lprintf(cfg, ret, 0,
"find_table_count: cursor prev failed");
goto err;
}
if ((ret = cursor->get_key(cursor, &key)) != 0) {
lprintf(cfg, ret, 0,
"find_table_count: cursor get_key failed");
goto err;
}
table_icount = (uint32_t)atoi(key);
if (table_icount > max_icount)
max_icount = table_icount;
if ((ret = cursor->close(cursor)) != 0) {
lprintf(cfg, ret, 0,
"find_table_count: cursor close failed");
goto err;
}
}
err: if ((t_ret = session->close(session, NULL)) != 0) {
if (ret == 0)
ret = t_ret;
lprintf(cfg, ret, 0,
"find_table_count: session close failed");
}
cfg->icount = max_icount;
out: return (ret);
}
/*
* Populate the uri array if more than one table is being used.
*/
static int
create_uris(CONFIG *cfg)
{
size_t base_uri_len;
uint32_t i;
int ret;
char *uri;
ret = 0;
base_uri_len = strlen(cfg->base_uri);
cfg->uris = calloc(cfg->table_count, sizeof(char *));
if (cfg->uris == NULL) {
ret = ENOMEM;
goto err;
}
for (i = 0; i < cfg->table_count; i++) {
uri = cfg->uris[i] = calloc(base_uri_len + 5, 1);
if (uri == NULL) {
ret = ENOMEM;
goto err;
}
/*
* If there is only one table, just use base name.
*/
if (cfg->table_count == 1)
memcpy(uri, cfg->base_uri, base_uri_len);
else
sprintf(uri, "%s%05d", cfg->base_uri, i);
}
err: if (ret != 0 && cfg->uris != NULL) {
for (i = 0; i < cfg->table_count; i++)
free(cfg->uris[i]);
free(cfg->uris);
cfg->uris = NULL;
}
return (ret);
}
static int
create_tables(CONFIG *cfg)
{
WT_SESSION *session;
size_t i;
int ret;
char buf[512];
if (cfg->create == 0)
return (0);
if ((ret = cfg->conn->open_session(
cfg->conn, NULL, cfg->sess_config, &session)) != 0) {
lprintf(cfg, ret, 0,
"Error opening a session on %s", cfg->home);
return (ret);
}
for (i = 0; i < cfg->table_count_idle; i++) {
snprintf(buf, 512, "%s_idle%05d", cfg->uris[0], (int)i);
if ((ret = session->create(
session, buf, cfg->table_config)) != 0) {
lprintf(cfg, ret, 0,
"Error creating idle table %s", buf);
return (ret);
}
}
for (i = 0; i < cfg->table_count; i++) {
if (cfg->log_partial && i > 0) {
if (((ret = session->create(session,
cfg->uris[i], cfg->partial_config)) != 0)) {
lprintf(cfg, ret, 0,
"Error creating table %s", cfg->uris[i]);
return (ret);
}
} else if ((ret = session->create(
session, cfg->uris[i], cfg->table_config)) != 0) {
lprintf(cfg, ret, 0,
"Error creating table %s", cfg->uris[i]);
return (ret);
}
if (cfg->index) {
snprintf(buf, 512, "index:%s:val_idx",
cfg->uris[i] + strlen("table:"));
if ((ret = session->create(
session, buf, "columns=(val)")) != 0) {
lprintf(cfg, ret, 0,
"Error creating index %s", buf);
return (ret);
}
}
}
if ((ret = session->close(session, NULL)) != 0) {
lprintf(cfg, ret, 0, "Error closing session");
return (ret);
}
return (0);
}
static int
start_all_runs(CONFIG *cfg)
{
CONFIG *next_cfg, **configs;
pthread_t *threads;
size_t cmd_len, home_len, i;
int ret, t_ret;
char *cmd_buf, *new_home;
ret = 0;
configs = NULL;
cmd_buf = NULL;
if (cfg->database_count == 1)
return (start_run(cfg));
/* Allocate an array to hold our config struct copies. */
configs = calloc(cfg->database_count, sizeof(CONFIG *));
if (configs == NULL)
return (ENOMEM);
/* Allocate an array to hold our thread IDs. */
threads = calloc(cfg->database_count, sizeof(pthread_t));
if (threads == NULL) {
ret = ENOMEM;
goto err;
}
home_len = strlen(cfg->home);
cmd_len = (home_len * 2) + 30; /* Add some slop. */
cmd_buf = calloc(cmd_len, 1);
if (cmd_buf == NULL) {
ret = ENOMEM;
goto err;
}
for (i = 0; i < cfg->database_count; i++) {
next_cfg = calloc(1, sizeof(CONFIG));
if (next_cfg == NULL) {
ret = ENOMEM;
goto err;
}
configs[i] = next_cfg;
if ((ret = config_assign(next_cfg, cfg)) != 0)
goto err;
/* Setup a unique home directory for each database. */
new_home = malloc(home_len + 5);
if (new_home == NULL) {
ret = ENOMEM;
goto err;
}
snprintf(new_home, home_len + 5, "%s/D%02d", cfg->home, (int)i);
next_cfg->home = new_home;
/* If the monitor dir is default, update it too. */
if (strcmp(cfg->monitor_dir, cfg->home) == 0)
next_cfg->monitor_dir = new_home;
/* Create clean home directories. */
snprintf(cmd_buf, cmd_len, "rm -rf %s && mkdir %s",
next_cfg->home, next_cfg->home);
if ((ret = system(cmd_buf)) != 0) {
fprintf(stderr, "%s: failed\n", cmd_buf);
goto err;
}
if ((ret = pthread_create(
&threads[i], NULL, thread_run_wtperf, next_cfg)) != 0) {
lprintf(cfg, ret, 0, "Error creating thread");
goto err;
}
}
/* Wait for threads to finish. */
for (i = 0; i < cfg->database_count; i++) {
if ((t_ret = pthread_join(threads[i], NULL)) != 0) {
lprintf(cfg, ret, 0, "Error joining thread");
if (ret == 0)
ret = t_ret;
}
}
err: for (i = 0; i < cfg->database_count && configs[i] != NULL; i++) {
free((char *)configs[i]->home);
config_free(configs[i]);
free(configs[i]);
}
free(configs);
free(threads);
free(cmd_buf);
return (ret);
}
/* Run an instance of wtperf for a given configuration. */
static void *
thread_run_wtperf(void *arg)
{
CONFIG *cfg;
int ret;
cfg = (CONFIG *)arg;
if ((ret = start_run(cfg)) != 0)
lprintf(cfg, ret, 0, "Run failed for: %s.", cfg->home);
return (NULL);
}
static int
start_run(CONFIG *cfg)
{
pthread_t monitor_thread;
uint64_t total_ops;
int monitor_created, ret, t_ret;
char helium_buf[256];
monitor_created = ret = 0;
/* [-Wconditional-uninitialized] */
memset(&monitor_thread, 0, sizeof(monitor_thread));
if ((ret = setup_log_file(cfg)) != 0)
goto err;
if ((ret = wiredtiger_open( /* Open the real connection. */
cfg->home, NULL, cfg->conn_config, &cfg->conn)) != 0) {
lprintf(cfg, ret, 0, "Error connecting to %s", cfg->home);
goto err;
}
/* Configure optional Helium volume. */
if (cfg->helium_mount != NULL) {
snprintf(helium_buf, sizeof(helium_buf),
"entry=wiredtiger_extension_init,config=["
"%s=[helium_devices=\"he://./%s\","
"helium_o_volume_truncate=1]]",
HELIUM_NAME, cfg->helium_mount);
if ((ret = cfg->conn->load_extension(
cfg->conn, HELIUM_PATH, helium_buf)) != 0)
lprintf(cfg,
ret, 0, "Error loading Helium: %s", helium_buf);
}
if ((ret = create_uris(cfg)) != 0)
goto err;
if ((ret = create_tables(cfg)) != 0)
goto err;
/* Start the monitor thread. */
if (cfg->sample_interval != 0) {
if ((ret = pthread_create(
&monitor_thread, NULL, monitor, cfg)) != 0) {
lprintf(
cfg, ret, 0, "Error creating monitor thread.");
goto err;
}
monitor_created = 1;
}
/* If creating, populate the table. */
if (cfg->create != 0 && execute_populate(cfg) != 0)
goto err;
/* Optional workload. */
if (cfg->workers_cnt != 0 &&
(cfg->run_time != 0 || cfg->run_ops != 0)) {
/*
* If we have a workload, close and reopen the connection so
* that LSM can detect read-only workloads.
*/
if (close_reopen(cfg) != 0)
goto err;
/* Didn't create, set insert count. */
if (cfg->create == 0 && find_table_count(cfg) != 0)
goto err;
/* Start the checkpoint thread. */
if (cfg->checkpoint_threads != 0) {
lprintf(cfg, 0, 1,
"Starting %" PRIu32 " checkpoint thread(s)",
cfg->checkpoint_threads);
if ((cfg->ckptthreads =
calloc(cfg->checkpoint_threads,
sizeof(CONFIG_THREAD))) == NULL) {
ret = enomem(cfg);
goto err;
}
if (start_threads(cfg, NULL, cfg->ckptthreads,
cfg->checkpoint_threads, checkpoint_worker) != 0)
goto err;
}
/* Execute the workload. */
if ((ret = execute_workload(cfg)) != 0)
goto err;
/* One final summation of the operations we've completed. */
cfg->read_ops = sum_read_ops(cfg);
cfg->insert_ops = sum_insert_ops(cfg);
cfg->update_ops = sum_update_ops(cfg);
cfg->ckpt_ops = sum_ckpt_ops(cfg);
total_ops = cfg->read_ops + cfg->insert_ops + cfg->update_ops;
lprintf(cfg, 0, 1,
"Executed %" PRIu64 " read operations (%" PRIu64
"%%) %" PRIu64 " ops/sec",
cfg->read_ops, (cfg->read_ops * 100) / total_ops,
cfg->read_ops / cfg->run_time);
lprintf(cfg, 0, 1,
"Executed %" PRIu64 " insert operations (%" PRIu64
"%%) %" PRIu64 " ops/sec",
cfg->insert_ops, (cfg->insert_ops * 100) / total_ops,
cfg->insert_ops / cfg->run_time);
lprintf(cfg, 0, 1,
"Executed %" PRIu64 " update operations (%" PRIu64
"%%) %" PRIu64 " ops/sec",
cfg->update_ops, (cfg->update_ops * 100) / total_ops,
cfg->update_ops / cfg->run_time);
lprintf(cfg, 0, 1,
"Executed %" PRIu64 " checkpoint operations",
cfg->ckpt_ops);
latency_print(cfg);
}
if (0) {
err: if (ret == 0)
ret = EXIT_FAILURE;
}
/* Notify the worker threads they are done. */
cfg->stop = 1;
if ((t_ret = stop_threads(cfg, 1, cfg->ckptthreads)) != 0)
if (ret == 0)
ret = t_ret;
if (monitor_created != 0 &&
(t_ret = pthread_join(monitor_thread, NULL)) != 0) {
lprintf(cfg, ret, 0, "Error joining monitor thread.");
if (ret == 0)
ret = t_ret;
}
if (cfg->conn != NULL &&
(t_ret = cfg->conn->close(cfg->conn, NULL)) != 0) {
lprintf(cfg, t_ret, 0,
"Error closing connection to %s", cfg->home);
if (ret == 0)
ret = t_ret;
}
if (ret == 0) {
if (cfg->run_time == 0 && cfg->run_ops == 0)
lprintf(cfg, 0, 1, "Run completed");
else
lprintf(cfg, 0, 1, "Run completed: %" PRIu32 " %s",
cfg->run_time == 0 ? cfg->run_ops : cfg->run_time,
cfg->run_time == 0 ? "operations" : "seconds");
}
if (cfg->logf != NULL) {
if ((t_ret = fflush(cfg->logf)) != 0 && ret == 0)
ret = t_ret;
if ((t_ret = fclose(cfg->logf)) != 0 && ret == 0)
ret = t_ret;
}
return (ret);
}
extern int __wt_optind, __wt_optreset;
extern char *__wt_optarg;
int
main(int argc, char *argv[])
{
CONFIG *cfg, _cfg;
size_t req_len;
int ch, monitor_set, ret;
const char *opts = "C:H:h:m:O:o:T:";
const char *config_opts;
char *cc_buf, *tc_buf, *user_cconfig, *user_tconfig;
monitor_set = ret = 0;
config_opts = NULL;
cc_buf = tc_buf = user_cconfig = user_tconfig = NULL;
/* Setup the default configuration values. */
cfg = &_cfg;
memset(cfg, 0, sizeof(*cfg));
if (config_assign(cfg, &default_cfg))
goto err;
/* Do a basic validation of options, and home is needed before open. */
while ((ch = __wt_getopt("wtperf", argc, argv, opts)) != EOF)
switch (ch) {
case 'C':
if (user_cconfig == NULL)
user_cconfig = strdup(__wt_optarg);
else {
user_cconfig = realloc(user_cconfig,
strlen(user_cconfig) +
strlen(__wt_optarg) + 2);
strcat(user_cconfig, ",");
strcat(user_cconfig, __wt_optarg);
}
break;
case 'H':
cfg->helium_mount = __wt_optarg;
break;
case 'O':
config_opts = __wt_optarg;
break;
case 'T':
if (user_tconfig == NULL)
user_tconfig = strdup(__wt_optarg);
else {
user_tconfig = realloc(user_tconfig,
strlen(user_tconfig) +
strlen(__wt_optarg) + 2);
strcat(user_tconfig, ",");
strcat(user_tconfig, __wt_optarg);
}
break;
case 'h':
cfg->home = __wt_optarg;
break;
case 'm':
cfg->monitor_dir = __wt_optarg;
monitor_set = 1;
break;
case '?':
fprintf(stderr, "Invalid option\n");
usage();
goto einval;
}
/*
* If the user did not specify a monitor directory then set the
* monitor directory to the home dir.
*/
if (!monitor_set)
cfg->monitor_dir = cfg->home;
/* Parse configuration settings from configuration file. */
if (config_opts != NULL && config_opt_file(cfg, config_opts) != 0)
goto einval;
/* Parse options that override values set via a configuration file. */
__wt_optreset = __wt_optind = 1;
while ((ch = __wt_getopt("wtperf", argc, argv, opts)) != EOF)
switch (ch) {
case 'o':
/* Allow -o key=value */
if (config_opt_line(cfg, __wt_optarg) != 0)
goto einval;
break;
}
cfg->async_config = NULL;
/*
* If the user specified async_threads we use async for all ops.
* If the user wants compaction, then we also enable async for
* the compact operation, but not for the workloads.
*/
if (cfg->async_threads > 0) {
if (cfg->has_truncate > 0)
goto err;
cfg->use_asyncops = 1;
}
if (cfg->compact && cfg->async_threads == 0)
cfg->async_threads = 2;
if (cfg->async_threads > 0) {
/*
* The maximum number of async threads is two digits, so just
* use that to compute the space we need. Assume the default
* of 1024 for the max ops. Although we could bump that up
* to 4096 if needed.
*/
req_len = strlen(",async=(enabled=true,threads=)") + 4;
if ((cfg->async_config = calloc(req_len, 1)) == NULL) {
ret = enomem(cfg);
goto err;
}
snprintf(cfg->async_config, req_len,
",async=(enabled=true,threads=%d)",
cfg->async_threads);
}
if ((ret = config_compress(cfg)) != 0)
goto err;
/* You can't have truncate on a random collection */
if (cfg->has_truncate && cfg->random_range)
goto err;
/* Build the URI from the table name. */
req_len = strlen("table:") +
strlen(HELIUM_NAME) + strlen(cfg->table_name) + 2;
if ((cfg->base_uri = calloc(req_len, 1)) == NULL) {
ret = enomem(cfg);
goto err;
}
snprintf(cfg->base_uri, req_len, "table:%s%s%s",
cfg->helium_mount == NULL ? "" : HELIUM_NAME,
cfg->helium_mount == NULL ? "" : "/",
cfg->table_name);
/* Make stdout line buffered, so verbose output appears quickly. */
(void)setvbuf(stdout, NULL, _IOLBF, 32);
/* Concatenate non-default configuration strings. */
if (cfg->verbose > 1 || user_cconfig != NULL ||
cfg->compress_ext != NULL || cfg->async_config != NULL) {
req_len = strlen(cfg->conn_config) + strlen(debug_cconfig) + 3;
if (user_cconfig != NULL)
req_len += strlen(user_cconfig);
if (cfg->async_config != NULL)
req_len += strlen(cfg->async_config);
if (cfg->compress_ext != NULL)
req_len += strlen(cfg->compress_ext);
if ((cc_buf = calloc(req_len, 1)) == NULL) {
ret = enomem(cfg);
goto err;
}
/*
* This is getting hard to parse.
*/
snprintf(cc_buf, req_len, "%s%s%s%s%s%s%s",
cfg->conn_config,
cfg->async_config ? cfg->async_config : "",
cfg->compress_ext ? cfg->compress_ext : "",
cfg->verbose > 1 ? ",": "",
cfg->verbose > 1 ? debug_cconfig : "",
user_cconfig ? ",": "",
user_cconfig ? user_cconfig : "");
if ((ret = config_opt_str(cfg, "conn_config", cc_buf)) != 0)
goto err;
}
if (cfg->verbose > 1 || cfg->index || cfg->helium_mount != NULL ||
user_tconfig != NULL || cfg->compress_table != NULL) {
req_len = strlen(cfg->table_config) + strlen(HELIUM_CONFIG) +
strlen(debug_tconfig) + 3;
if (user_tconfig != NULL)
req_len += strlen(user_tconfig);
if (cfg->compress_table != NULL)
req_len += strlen(cfg->compress_table);
if (cfg->index)
req_len += strlen(INDEX_COL_NAMES);
if ((tc_buf = calloc(req_len, 1)) == NULL) {
ret = enomem(cfg);
goto err;
}
/*
* This is getting hard to parse.
*/
snprintf(tc_buf, req_len, "%s%s%s%s%s%s%s%s",
cfg->table_config,
cfg->index ? INDEX_COL_NAMES : "",
cfg->compress_table ? cfg->compress_table : "",
cfg->verbose > 1 ? ",": "",
cfg->verbose > 1 ? debug_tconfig : "",
user_tconfig ? ",": "",
user_tconfig ? user_tconfig : "",
cfg->helium_mount == NULL ? "" : HELIUM_CONFIG);
if ((ret = config_opt_str(cfg, "table_config", tc_buf)) != 0)
goto err;
}
if (cfg->log_partial && cfg->table_count > 1) {
req_len = strlen(cfg->table_config) +
strlen(LOG_PARTIAL_CONFIG) + 1;
if ((cfg->partial_config = calloc(req_len, 1)) == NULL) {
ret = enomem(cfg);
goto err;
}
snprintf((char *)cfg->partial_config, req_len, "%s%s",
(char *)cfg->table_config, LOG_PARTIAL_CONFIG);
}
/* Sanity-check the configuration. */
if ((ret = config_sanity(cfg)) != 0)
goto err;
/* Display the configuration. */
if (cfg->verbose > 1)
config_print(cfg);
if ((ret = start_all_runs(cfg)) != 0)
goto err;
if (0) {
einval: ret = EINVAL;
}
err: config_free(cfg);
free(cc_buf);
free(tc_buf);
free(user_cconfig);
free(user_tconfig);
return (ret == 0 ? EXIT_SUCCESS : EXIT_FAILURE);
}
static int
start_threads(CONFIG *cfg,
WORKLOAD *workp, CONFIG_THREAD *base, u_int num, void *(*func)(void *))
{
CONFIG_THREAD *thread;
u_int i, j;
int ret;
/* Initialize the threads. */
for (i = 0, thread = base; i < num; ++i, ++thread) {
thread->cfg = cfg;
thread->workload = workp;
/*
* We don't want the threads executing in lock-step, move each
* new RNG state further along in the sequence.
*/
if (i == 0)
__wt_random_init(&thread->rnd);
else
thread->rnd = (thread - 1)->rnd;
for (j = 0; j < 1000; ++j)
(void)__wt_random(&thread->rnd);
/*
* Every thread gets a key/data buffer because we don't bother
* to distinguish between threads needing them and threads that
* don't, it's not enough memory to bother. These buffers hold
* strings: trailing NUL is included in the size.
*/
if ((thread->key_buf = calloc(cfg->key_sz, 1)) == NULL)
return (enomem(cfg));
if ((thread->value_buf = calloc(cfg->value_sz, 1)) == NULL)
return (enomem(cfg));
/*
* Initialize and then toss in a bit of random values if needed.
*/
memset(thread->value_buf, 'a', cfg->value_sz - 1);
if (cfg->random_value)
randomize_value(thread, thread->value_buf);
/*
* Every thread gets tracking information and is initialized
* for latency measurements, for the same reason.
*/
thread->ckpt.min_latency =
thread->insert.min_latency = thread->read.min_latency =
thread->update.min_latency = UINT32_MAX;
thread->ckpt.max_latency = thread->insert.max_latency =
thread->read.max_latency = thread->update.max_latency = 0;
thread->total_inserts = 0;
}
/* Start the threads. */
for (i = 0, thread = base; i < num; ++i, ++thread)
if ((ret = pthread_create(
&thread->handle, NULL, func, thread)) != 0) {
lprintf(cfg, ret, 0, "Error creating thread");
return (ret);
}
return (0);
}
static int
stop_threads(CONFIG *cfg, u_int num, CONFIG_THREAD *threads)
{
u_int i;
int ret;
if (num == 0 || threads == NULL)
return (0);
for (i = 0; i < num; ++i, ++threads) {
if ((ret = pthread_join(threads->handle, NULL)) != 0) {
lprintf(cfg, ret, 0, "Error joining thread");
return (ret);
}
free(threads->key_buf);
threads->key_buf = NULL;
free(threads->value_buf);
threads->value_buf = NULL;
}
/*
* We don't free the thread structures or any memory referenced, or NULL
* the reference when we stop the threads; the thread structure is still
* being read by the monitor thread (among others). As a standalone
* program, leaking memory isn't a concern, and it's simpler that way.
*/
return (0);
}
/*
* TODO: Spread the stalls out, so we don't flood at the start of each
* second and then pause. Doing this every 10th of a second is probably enough
*/
static void
worker_throttle(int64_t throttle, int64_t *ops, struct timespec *interval)
{
struct timespec now;
uint64_t usecs_to_complete;
if (*ops < throttle)
return;
/* Ignore errors, we don't really care. */
if (__wt_epoch(NULL, &now) != 0)
return;
/*
* If we've completed enough operations, reset the counters.
* If we did enough operations in less than a second, sleep for
* the rest of the second.
*/
usecs_to_complete = ns_to_us(WT_TIMEDIFF(now, *interval));
if (usecs_to_complete < USEC_PER_SEC)
(void)usleep((useconds_t)(USEC_PER_SEC - usecs_to_complete));
*ops = 0;
*interval = now;
}
static int
drop_all_tables(CONFIG *cfg)
{
struct timespec start, stop;
WT_SESSION *session;
size_t i;
uint64_t msecs;
int ret, t_ret;
/* Drop any tables. */
if ((ret = cfg->conn->open_session(
cfg->conn, NULL, cfg->sess_config, &session)) != 0) {
lprintf(cfg, ret, 0,
"Error opening a session on %s", cfg->home);
return (ret);
}
(void)__wt_epoch(NULL, &start);
for (i = 0; i < cfg->table_count; i++) {
if ((ret = session->drop(
session, cfg->uris[i], NULL)) != 0) {
lprintf(cfg, ret, 0,
"Error dropping table %s", cfg->uris[i]);
goto err;
}
}
(void)__wt_epoch(NULL, &stop);
msecs = ns_to_ms(WT_TIMEDIFF(stop, start));
lprintf(cfg, 0, 1,
"Executed %" PRIu32 " drop operations average time %" PRIu64 "ms",
cfg->table_count, msecs / cfg->table_count);
err: if ((t_ret = session->close(session, NULL)) != 0 && ret == 0)
ret = t_ret;
return (ret);
}
static uint64_t
wtperf_value_range(CONFIG *cfg)
{
if (cfg->random_range)
return (cfg->icount + cfg->random_range);
return (cfg->icount + cfg->insert_key - (u_int)(cfg->workers_cnt + 1));
}
static uint64_t
wtperf_rand(CONFIG_THREAD *thread)
{
CONFIG *cfg;
double S1, S2, U;
uint64_t rval;
cfg = thread->cfg;
/*
* Use WiredTiger's random number routine: it's lock-free and fairly
* good.
*/
rval = __wt_random(&thread->rnd);
/* Use Pareto distribution to give 80/20 hot/cold values. */
if (cfg->pareto != 0) {
#define PARETO_SHAPE 1.5
S1 = (-1 / PARETO_SHAPE);
S2 = wtperf_value_range(cfg) *
(cfg->pareto / 100.0) * (PARETO_SHAPE - 1);
U = 1 - (double)rval / (double)UINT32_MAX;
rval = (pow(U, S1) - 1) * S2;
/*
* This Pareto calculation chooses out of range values about
* 2% of the time, from my testing. That will lead to the
* first item in the table being "hot".
*/
if (rval > wtperf_value_range(cfg))
rval = wtperf_value_range(cfg);
}
/*
* Wrap the key to within the expected range and avoid zero: we never
* insert that key.
*/
rval = (rval % wtperf_value_range(cfg)) + 1;
return (rval);
}
|