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
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
|
/**
* Copyright (C) 2014 MongoDB Inc.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License, version 3,
* as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Affero General Public License for more details.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*
* As a special exception, the copyright holders give permission to link the
* code of portions of this program with the OpenSSL library under certain
* conditions as described in each individual source file and distribute
* linked combinations including the program with the OpenSSL library. You
* must comply with the GNU Affero General Public License in all respects for
* all of the code used other than as permitted herein. If you modify file(s)
* with this exception, you may extend this exception to your version of the
* file(s), but you are not obligated to do so. If you do not wish to do so,
* delete this exception statement from your version. If you delete this
* exception statement from all source files in the program, then also delete
* it in the license file.
*/
#define MONGO_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kReplication
#include "mongo/platform/basic.h"
#include "mongo/db/repl/replication_coordinator_impl.h"
#include <algorithm>
#include <limits>
#include "mongo/base/status.h"
#include "mongo/client/fetcher.h"
#include "mongo/db/client.h"
#include "mongo/db/commands.h"
#include "mongo/db/concurrency/d_concurrency.h"
#include "mongo/db/index/index_descriptor.h"
#include "mongo/db/logical_clock.h"
#include "mongo/db/logical_time.h"
#include "mongo/db/operation_context_noop.h"
#include "mongo/db/repl/check_quorum_for_config_change.h"
#include "mongo/db/repl/data_replicator_external_state_initial_sync.h"
#include "mongo/db/repl/elect_cmd_runner.h"
#include "mongo/db/repl/freshness_checker.h"
#include "mongo/db/repl/freshness_scanner.h"
#include "mongo/db/repl/handshake_args.h"
#include "mongo/db/repl/is_master_response.h"
#include "mongo/db/repl/last_vote.h"
#include "mongo/db/repl/old_update_position_args.h"
#include "mongo/db/repl/read_concern_args.h"
#include "mongo/db/repl/repl_client_info.h"
#include "mongo/db/repl/repl_set_config_checks.h"
#include "mongo/db/repl/repl_set_heartbeat_args.h"
#include "mongo/db/repl/repl_set_heartbeat_args_v1.h"
#include "mongo/db/repl/repl_set_heartbeat_response.h"
#include "mongo/db/repl/repl_set_html_summary.h"
#include "mongo/db/repl/repl_set_request_votes_args.h"
#include "mongo/db/repl/repl_settings.h"
#include "mongo/db/repl/replication_executor.h"
#include "mongo/db/repl/rslog.h"
#include "mongo/db/repl/storage_interface.h"
#include "mongo/db/repl/topology_coordinator.h"
#include "mongo/db/repl/update_position_args.h"
#include "mongo/db/repl/vote_requester.h"
#include "mongo/db/server_options.h"
#include "mongo/db/server_parameters.h"
#include "mongo/db/storage/mmap_v1/dur.h"
#include "mongo/db/write_concern.h"
#include "mongo/db/write_concern_options.h"
#include "mongo/executor/connection_pool_stats.h"
#include "mongo/executor/network_interface.h"
#include "mongo/rpc/metadata/oplog_query_metadata.h"
#include "mongo/rpc/metadata/repl_set_metadata.h"
#include "mongo/rpc/metadata/server_selection_metadata.h"
#include "mongo/rpc/request_interface.h"
#include "mongo/stdx/functional.h"
#include "mongo/stdx/mutex.h"
#include "mongo/util/assert_util.h"
#include "mongo/util/fail_point_service.h"
#include "mongo/util/log.h"
#include "mongo/util/scopeguard.h"
#include "mongo/util/stacktrace.h"
#include "mongo/util/time_support.h"
#include "mongo/util/timer.h"
namespace mongo {
namespace repl {
MONGO_FP_DECLARE(transitionToPrimaryHangBeforeTakingGlobalExclusiveLock);
using CallbackArgs = executor::TaskExecutor::CallbackArgs;
using CallbackFn = executor::TaskExecutor::CallbackFn;
using CallbackHandle = executor::TaskExecutor::CallbackHandle;
using CBHandle = ReplicationExecutor::CallbackHandle;
using CBHStatus = StatusWith<CBHandle>;
using EventHandle = executor::TaskExecutor::EventHandle;
using executor::NetworkInterface;
using LockGuard = stdx::lock_guard<stdx::mutex>;
using NextAction = Fetcher::NextAction;
namespace {
const char kLocalDB[] = "local";
MONGO_EXPORT_SERVER_PARAMETER(numInitialSyncAttempts, int, 10);
void lockAndCall(stdx::unique_lock<stdx::mutex>* lk, const stdx::function<void()>& fn) {
if (!lk->owns_lock()) {
lk->lock();
}
fn();
}
/**
* Implements the force-reconfig behavior of incrementing config version by a large random
* number.
*/
BSONObj incrementConfigVersionByRandom(BSONObj config) {
BSONObjBuilder builder;
for (BSONObjIterator iter(config); iter.more(); iter.next()) {
BSONElement elem = *iter;
if (elem.fieldNameStringData() == ReplSetConfig::kVersionFieldName && elem.isNumber()) {
std::unique_ptr<SecureRandom> generator(SecureRandom::create());
const int random = std::abs(static_cast<int>(generator->nextInt64()) % 100000);
builder.appendIntOrLL(ReplSetConfig::kVersionFieldName,
elem.numberLong() + 10000 + random);
} else {
builder.append(elem);
}
}
return builder.obj();
}
const Seconds kNoopWriterPeriod(10);
} // namespace
BSONObj ReplicationCoordinatorImpl::SlaveInfo::toBSON() const {
BSONObjBuilder bo;
bo.append("id", memberId);
bo.append("rid", rid);
bo.append("host", hostAndPort.toString());
bo.append("lastDurableOpTime", lastDurableOpTime.toBSON());
bo.append("lastAppliedOpTime", lastAppliedOpTime.toBSON());
if (self)
bo.append("self", true);
if (down)
bo.append("down", true);
bo.append("lastUpdated", lastUpdate);
return bo.obj();
}
std::string ReplicationCoordinatorImpl::SlaveInfo::toString() const {
return toBSON().toString();
}
struct ReplicationCoordinatorImpl::WaiterInfo {
using FinishFunc = stdx::function<void()>;
WaiterInfo(unsigned int _opID,
const OpTime _opTime,
const WriteConcernOptions* _writeConcern,
stdx::condition_variable* _condVar)
: opID(_opID), opTime(_opTime), writeConcern(_writeConcern), condVar(_condVar) {}
// When waiter is signaled, finishCallback will be called while holding replCoord _mutex
// since WaiterLists are protected by _mutex.
WaiterInfo(const OpTime _opTime, FinishFunc _finishCallback)
: opTime(_opTime), finishCallback(_finishCallback) {}
BSONObj toBSON() const {
BSONObjBuilder bob;
bob.append("opId", opID);
bob.append("opTime", opTime.toBSON());
if (writeConcern) {
bob.append("writeConcern", writeConcern->toBSON());
}
return bob.obj();
};
std::string toString() const {
return toBSON().toString();
};
// It is invalid to call notify() unless holding ReplicationCoordinatorImpl::_mutex.
void notify() {
if (condVar) {
condVar->notify_all();
}
if (finishCallback) {
finishCallback();
}
}
const unsigned int opID = 0;
const OpTime opTime;
const WriteConcernOptions* writeConcern = nullptr;
stdx::condition_variable* condVar = nullptr;
// The callback that will be called when this waiter is notified.
FinishFunc finishCallback = nullptr;
};
struct ReplicationCoordinatorImpl::WaiterInfoGuard {
/**
* Constructor takes the list of waiters and enqueues itself on the list, removing itself
* in the destructor.
*
* Usually waiters will be signaled and removed when their criteria are satisfied, but
* wait_until() with timeout may signal waiters earlier and this guard will remove the waiter
* properly.
*
* _list is guarded by ReplicationCoordinatorImpl::_mutex, thus it is illegal to construct one
* of these without holding _mutex
*/
WaiterInfoGuard(WaiterList* list,
unsigned int opID,
const OpTime opTime,
const WriteConcernOptions* writeConcern,
stdx::condition_variable* condVar)
: waiter(opID, opTime, writeConcern, condVar), _list(list) {
list->add_inlock(&waiter);
}
~WaiterInfoGuard() {
_list->remove_inlock(&waiter);
}
WaiterInfo waiter;
private:
WaiterList* _list;
};
void ReplicationCoordinatorImpl::WaiterList::add_inlock(WaiterType waiter) {
_list.push_back(waiter);
}
void ReplicationCoordinatorImpl::WaiterList::signalAndRemoveIf_inlock(
stdx::function<bool(WaiterType)> func) {
std::vector<WaiterType>::iterator it = _list.end();
while (true) {
it = std::find_if(_list.begin(), _list.end(), func);
if (it == _list.end()) {
break;
}
(*it)->notify();
std::swap(*it, _list.back());
_list.pop_back();
}
}
void ReplicationCoordinatorImpl::WaiterList::signalAndRemoveAll_inlock() {
for (auto& waiter : _list) {
waiter->notify();
}
_list.clear();
}
bool ReplicationCoordinatorImpl::WaiterList::remove_inlock(WaiterType waiter) {
auto it = std::find(_list.begin(), _list.end(), waiter);
if (it != _list.end()) {
std::swap(*it, _list.back());
_list.pop_back();
return true;
}
return false;
}
namespace {
ReplicationCoordinator::Mode getReplicationModeFromSettings(const ReplSettings& settings) {
if (settings.usingReplSets()) {
return ReplicationCoordinator::modeReplSet;
}
if (settings.isMaster() || settings.isSlave()) {
return ReplicationCoordinator::modeMasterSlave;
}
return ReplicationCoordinator::modeNone;
}
DataReplicatorOptions createDataReplicatorOptions(
ReplicationCoordinator* replCoord, ReplicationCoordinatorExternalState* externalState) {
DataReplicatorOptions options;
options.getMyLastOptime = [replCoord]() { return replCoord->getMyLastAppliedOpTime(); };
options.setMyLastOptime = [replCoord, externalState](const OpTime& opTime) {
replCoord->setMyLastAppliedOpTime(opTime);
externalState->setGlobalTimestamp(replCoord->getServiceContext(), opTime.getTimestamp());
};
options.getSlaveDelay = [replCoord]() { return replCoord->getSlaveDelaySecs(); };
options.syncSourceSelector = replCoord;
options.replBatchLimitBytes = dur::UncommittedBytesLimit;
options.oplogFetcherMaxFetcherRestarts = externalState->getOplogFetcherMaxFetcherRestarts();
return options;
}
} // namespace
std::string ReplicationCoordinatorImpl::SnapshotInfo::toString() const {
BSONObjBuilder bob;
bob.append("optime", opTime.toBSON());
bob.append("name-id", name.toString());
return bob.obj().toString();
}
ReplicationCoordinatorImpl::ReplicationCoordinatorImpl(
ServiceContext* service,
const ReplSettings& settings,
std::unique_ptr<ReplicationCoordinatorExternalState> externalState,
std::unique_ptr<NetworkInterface> network,
std::unique_ptr<TopologyCoordinator> topCoord,
StorageInterface* storage,
int64_t prngSeed)
: _service(service),
_settings(settings),
_replMode(getReplicationModeFromSettings(settings)),
_topCoord(std::move(topCoord)),
_replExecutor(std::move(network), prngSeed),
_externalState(std::move(externalState)),
_inShutdown(false),
_memberState(MemberState::RS_STARTUP),
_rsConfigState(kConfigPreStart),
_selfIndex(-1),
_sleptLastElection(false),
_canAcceptNonLocalWrites(!(settings.usingReplSets() || settings.isSlave())),
_canServeNonLocalReads(0U),
_storage(storage) {
invariant(_service);
if (!isReplEnabled()) {
return;
}
std::unique_ptr<SecureRandom> rbidGenerator(SecureRandom::create());
_rbid = static_cast<int>(rbidGenerator->nextInt64());
if (_rbid < 0) {
// Ensure _rbid is always positive
_rbid = -_rbid;
}
// Make sure there is always an entry in _slaveInfo for ourself.
SlaveInfo selfInfo;
selfInfo.self = true;
_slaveInfo.push_back(selfInfo);
_externalState->setupNoopWriter(kNoopWriterPeriod);
}
ReplicationCoordinatorImpl::~ReplicationCoordinatorImpl() {}
void ReplicationCoordinatorImpl::waitForStartUpComplete_forTest() {
_waitForStartUpComplete();
}
void ReplicationCoordinatorImpl::_waitForStartUpComplete() {
CallbackHandle handle;
{
stdx::unique_lock<stdx::mutex> lk(_mutex);
while (_rsConfigState == kConfigPreStart || _rsConfigState == kConfigStartingUp) {
_rsConfigStateChange.wait(lk);
}
handle = _finishLoadLocalConfigCbh;
}
if (handle.isValid()) {
_replExecutor.wait(handle);
}
}
ReplSetConfig ReplicationCoordinatorImpl::getReplicaSetConfig_forTest() {
stdx::lock_guard<stdx::mutex> lk(_mutex);
return _rsConfig;
}
Date_t ReplicationCoordinatorImpl::getElectionTimeout_forTest() const {
stdx::lock_guard<stdx::mutex> lk(_mutex);
if (!_handleElectionTimeoutCbh.isValid()) {
return Date_t();
}
return _handleElectionTimeoutWhen;
}
Milliseconds ReplicationCoordinatorImpl::getRandomizedElectionOffset_forTest() {
return _getRandomizedElectionOffset();
}
boost::optional<Date_t> ReplicationCoordinatorImpl::getPriorityTakeover_forTest() const {
stdx::lock_guard<stdx::mutex> lk(_mutex);
if (!_priorityTakeoverCbh.isValid()) {
return boost::none;
}
return _priorityTakeoverWhen;
}
OpTime ReplicationCoordinatorImpl::getCurrentCommittedSnapshotOpTime() const {
stdx::lock_guard<stdx::mutex> lk(_mutex);
return _getCurrentCommittedSnapshotOpTime_inlock();
}
OpTime ReplicationCoordinatorImpl::_getCurrentCommittedSnapshotOpTime_inlock() const {
if (_currentCommittedSnapshot) {
return _currentCommittedSnapshot->opTime;
}
return OpTime();
}
void ReplicationCoordinatorImpl::appendConnectionStats(executor::ConnectionPoolStats* stats) const {
_replExecutor.appendConnectionStats(stats);
}
bool ReplicationCoordinatorImpl::_startLoadLocalConfig(OperationContext* opCtx) {
StatusWith<LastVote> lastVote = _externalState->loadLocalLastVoteDocument(opCtx);
if (!lastVote.isOK()) {
if (lastVote.getStatus() == ErrorCodes::NoMatchingDocument) {
log() << "Did not find local voted for document at startup.";
} else {
severe() << "Error loading local voted for document at startup; "
<< lastVote.getStatus();
fassertFailedNoTrace(40367);
}
} else {
LockGuard topoLock(_topoMutex);
_topCoord->loadLastVote(lastVote.getValue());
}
StatusWith<BSONObj> cfg = _externalState->loadLocalConfigDocument(opCtx);
if (!cfg.isOK()) {
log() << "Did not find local replica set configuration document at startup; "
<< cfg.getStatus();
return true;
}
ReplSetConfig localConfig;
Status status = localConfig.initialize(cfg.getValue());
if (!status.isOK()) {
error() << "Locally stored replica set configuration does not parse; See "
"http://www.mongodb.org/dochub/core/recover-replica-set-from-invalid-config "
"for information on how to recover from this. Got \""
<< status << "\" while parsing " << cfg.getValue();
fassertFailedNoTrace(28545);
}
// Read the last op from the oplog after cleaning up any partially applied batches.
_externalState->cleanUpLastApplyBatch(opCtx);
auto lastOpTimeStatus = _externalState->loadLastOpTime(opCtx);
// Use a callback here, because _finishLoadLocalConfig calls isself() which requires
// that the server's networking layer be up and running and accepting connections, which
// doesn't happen until startReplication finishes.
auto handle = _scheduleDBWork(stdx::bind(&ReplicationCoordinatorImpl::_finishLoadLocalConfig,
this,
stdx::placeholders::_1,
localConfig,
lastOpTimeStatus,
lastVote));
{
stdx::lock_guard<stdx::mutex> lk(_mutex);
_finishLoadLocalConfigCbh = handle;
}
return false;
}
void ReplicationCoordinatorImpl::_finishLoadLocalConfig(
const ReplicationExecutor::CallbackArgs& cbData,
const ReplSetConfig& localConfig,
const StatusWith<OpTime>& lastOpTimeStatus,
const StatusWith<LastVote>& lastVoteStatus) {
if (!cbData.status.isOK()) {
LOG(1) << "Loading local replica set configuration failed due to " << cbData.status;
return;
}
LockGuard topoLock(_topoMutex);
StatusWith<int> myIndex =
validateConfigForStartUp(_externalState.get(), _rsConfig, localConfig, getServiceContext());
if (!myIndex.isOK()) {
if (myIndex.getStatus() == ErrorCodes::NodeNotFound ||
myIndex.getStatus() == ErrorCodes::DuplicateKey) {
warning() << "Locally stored replica set configuration does not have a valid entry "
"for the current node; waiting for reconfig or remote heartbeat; Got \""
<< myIndex.getStatus() << "\" while validating " << localConfig.toBSON();
myIndex = StatusWith<int>(-1);
} else {
error() << "Locally stored replica set configuration is invalid; See "
"http://www.mongodb.org/dochub/core/recover-replica-set-from-invalid-config"
" for information on how to recover from this. Got \""
<< myIndex.getStatus() << "\" while validating " << localConfig.toBSON();
fassertFailedNoTrace(28544);
}
}
if (localConfig.getReplSetName() != _settings.ourSetName()) {
warning() << "Local replica set configuration document reports set name of "
<< localConfig.getReplSetName() << ", but command line reports "
<< _settings.ourSetName() << "; waitng for reconfig or remote heartbeat";
myIndex = StatusWith<int>(-1);
}
// Do not check optime, if this node is an arbiter.
bool isArbiter =
myIndex.getValue() != -1 && localConfig.getMemberAt(myIndex.getValue()).isArbiter();
OpTime lastOpTime;
if (!isArbiter) {
if (!lastOpTimeStatus.isOK()) {
warning() << "Failed to load timestamp of most recently applied operation: "
<< lastOpTimeStatus.getStatus();
} else {
lastOpTime = lastOpTimeStatus.getValue();
}
}
long long term = OpTime::kUninitializedTerm;
if (localConfig.getProtocolVersion() == 1) {
// Restore the current term according to the terms of last oplog entry and last vote.
// The initial term of OpTime() is 0.
term = lastOpTime.getTerm();
if (lastVoteStatus.isOK()) {
long long lastVoteTerm = lastVoteStatus.getValue().getTerm();
if (term < lastVoteTerm) {
term = lastVoteTerm;
}
}
}
stdx::unique_lock<stdx::mutex> lock(_mutex);
invariant(_rsConfigState == kConfigStartingUp);
const PostMemberStateUpdateAction action =
_setCurrentRSConfig_inlock(localConfig, myIndex.getValue());
if (!lastOpTime.isNull()) {
_setMyLastAppliedOpTime_inlock(lastOpTime, false);
_setMyLastDurableOpTime_inlock(lastOpTime, false);
_reportUpstream_inlock(std::move(lock));
// Unlocked below.
} else {
lock.unlock();
}
_externalState->setGlobalTimestamp(getServiceContext(), lastOpTime.getTimestamp());
// Step down is impossible, so we don't need to wait for the returned event.
_updateTerm_incallback(term);
LOG(1) << "Current term is now " << term;
_performPostMemberStateUpdateAction(action);
if (!isArbiter) {
_externalState->startThreads(_settings);
invariant(cbData.opCtx);
_startDataReplication(cbData.opCtx);
}
}
void ReplicationCoordinatorImpl::_stopDataReplication(OperationContext* opCtx) {
std::shared_ptr<DataReplicator> drCopy;
{
LockGuard lk(_mutex);
_dr.swap(drCopy);
}
if (drCopy) {
LOG(1)
<< "ReplicationCoordinatorImpl::_stopDataReplication calling DataReplicator::shutdown.";
const auto status = drCopy->shutdown();
if (!status.isOK()) {
warning() << "DataReplicator shutdown failed: " << status;
}
drCopy.reset();
// Do not return here, fall through.
}
LOG(1) << "ReplicationCoordinatorImpl::_stopDataReplication calling "
"ReplCoordExtState::stopDataReplication.";
_externalState->stopDataReplication(opCtx);
}
void ReplicationCoordinatorImpl::_startDataReplication(OperationContext* opCtx,
stdx::function<void()> startCompleted) {
// Check to see if we need to do an initial sync.
const auto lastOpTime = getMyLastAppliedOpTime();
const auto needsInitialSync =
lastOpTime.isNull() || _externalState->isInitialSyncFlagSet(opCtx);
if (!needsInitialSync) {
stdx::lock_guard<stdx::mutex> lk(_mutex);
if (!_inShutdown) {
// Start steady replication, since we already have data.
_externalState->startSteadyStateReplication(opCtx, this);
}
return;
}
// Do initial sync.
if (_externalState->shouldUseDataReplicatorInitialSync()) {
if (!_externalState->getTaskExecutor()) {
log() << "not running initial sync during test.";
return;
}
auto onCompletion = [this, startCompleted](const StatusWith<OpTimeWithHash>& status) {
{
stdx::lock_guard<stdx::mutex> lock(_mutex);
if (status == ErrorCodes::CallbackCanceled) {
log() << "Initial Sync has been cancelled: " << status.getStatus();
return;
} else if (!status.isOK()) {
if (_inShutdown) {
log() << "Initial Sync failed during shutdown due to "
<< status.getStatus();
return;
} else {
error() << "Initial sync failed, shutting down now. Restart the server "
"to attempt a new initial sync.";
fassertFailedWithStatusNoTrace(40088, status.getStatus());
}
}
const auto lastApplied = status.getValue();
_setMyLastAppliedOpTime_inlock(lastApplied.opTime, false);
}
// Clear maint. mode.
while (getMaintenanceMode()) {
setMaintenanceMode(false);
}
if (startCompleted) {
startCompleted();
}
// Repair local db (to compact it).
auto opCtx = cc().makeOperationContext();
uassertStatusOK(_externalState->runRepairOnLocalDB(opCtx.get()));
_externalState->startSteadyStateReplication(opCtx.get(), this);
};
std::shared_ptr<DataReplicator> drCopy;
try {
{
// Must take the lock to set _dr, but not call it.
stdx::lock_guard<stdx::mutex> lock(_mutex);
drCopy = std::make_shared<DataReplicator>(
createDataReplicatorOptions(this, _externalState.get()),
stdx::make_unique<DataReplicatorExternalStateInitialSync>(this,
_externalState.get()),
_storage,
onCompletion);
_dr = drCopy;
}
// DataReplicator::startup() must be called outside lock because it uses features (eg.
// setting the initial sync flag) which depend on the ReplicationCoordinatorImpl.
uassertStatusOK(drCopy->startup(opCtx, numInitialSyncAttempts.load()));
} catch (...) {
auto status = exceptionToStatus();
log() << "Initial Sync failed to start: " << status;
if (ErrorCodes::CallbackCanceled == status ||
ErrorCodes::isShutdownError(status.code())) {
return;
}
fassertFailedWithStatusNoTrace(40354, status);
}
} else {
_externalState->startInitialSync([this, startCompleted](OperationContext* opCtx) {
stdx::lock_guard<stdx::mutex> lk(_mutex);
if (!_inShutdown) {
if (startCompleted) {
startCompleted();
}
_externalState->startSteadyStateReplication(opCtx, this);
}
});
}
}
void ReplicationCoordinatorImpl::startup(OperationContext* opCtx) {
if (!isReplEnabled()) {
stdx::lock_guard<stdx::mutex> lk(_mutex);
_setConfigState_inlock(kConfigReplicationDisabled);
return;
}
{
OID rid = _externalState->ensureMe(opCtx);
stdx::lock_guard<stdx::mutex> lk(_mutex);
fassert(18822, !_inShutdown);
_setConfigState_inlock(kConfigStartingUp);
_myRID = rid;
_slaveInfo[_getMyIndexInSlaveInfo_inlock()].rid = rid;
}
if (!_settings.usingReplSets()) {
// Must be Master/Slave
invariant(_settings.isMaster() || _settings.isSlave());
_externalState->startMasterSlave(opCtx);
return;
}
_replExecutor.startup();
_topCoord->setStorageEngineSupportsReadCommitted(
_externalState->isReadCommittedSupportedByStorageEngine(opCtx));
bool doneLoadingConfig = _startLoadLocalConfig(opCtx);
if (doneLoadingConfig) {
// If we're not done loading the config, then the config state will be set by
// _finishLoadLocalConfig.
stdx::lock_guard<stdx::mutex> lk(_mutex);
invariant(!_rsConfig.isInitialized());
_setConfigState_inlock(kConfigUninitialized);
}
}
void ReplicationCoordinatorImpl::shutdown(OperationContext* opCtx) {
// Shutdown must:
// * prevent new threads from blocking in awaitReplication
// * wake up all existing threads blocking in awaitReplication
// * tell the ReplicationExecutor to shut down
// * wait for the thread running the ReplicationExecutor to finish
if (!_settings.usingReplSets()) {
return;
}
log() << "shutting down replication subsystems";
// Used to shut down outside of the lock.
std::shared_ptr<DataReplicator> drCopy;
{
stdx::unique_lock<stdx::mutex> lk(_mutex);
fassert(28533, !_inShutdown);
_inShutdown = true;
if (_rsConfigState == kConfigPreStart) {
warning() << "ReplicationCoordinatorImpl::shutdown() called before "
"startup() finished. Shutting down without cleaning up the "
"replication system";
return;
}
if (_rsConfigState == kConfigStartingUp) {
// Wait until we are finished starting up, so that we can cleanly shut everything down.
lk.unlock();
_waitForStartUpComplete();
lk.lock();
fassert(18823, _rsConfigState != kConfigStartingUp);
}
_replicationWaiterList.signalAndRemoveAll_inlock();
_opTimeWaiterList.signalAndRemoveAll_inlock();
_currentCommittedSnapshotCond.notify_all();
_dr.swap(drCopy);
}
{
stdx::lock_guard<stdx::mutex> topoLock(_topoMutex);
_signalStepDownWaiter_inlock();
}
// joining the replication executor is blocking so it must be run outside of the mutex
if (drCopy) {
LOG(1) << "ReplicationCoordinatorImpl::shutdown calling DataReplicator::shutdown.";
const auto status = drCopy->shutdown();
if (!status.isOK()) {
warning() << "DataReplicator shutdown failed: " << status;
}
drCopy->join();
drCopy.reset();
}
_externalState->shutdown(opCtx);
_replExecutor.shutdown();
_replExecutor.join();
}
const ReplSettings& ReplicationCoordinatorImpl::getSettings() const {
return _settings;
}
ReplicationCoordinator::Mode ReplicationCoordinatorImpl::getReplicationMode() const {
return _replMode;
}
MemberState ReplicationCoordinatorImpl::getMemberState() const {
stdx::lock_guard<stdx::mutex> lk(_mutex);
return _getMemberState_inlock();
}
MemberState ReplicationCoordinatorImpl::_getMemberState_inlock() const {
return _memberState;
}
Status ReplicationCoordinatorImpl::waitForMemberState(MemberState expectedState,
Milliseconds timeout) {
if (timeout < Milliseconds(0)) {
return Status(ErrorCodes::BadValue, "Timeout duration cannot be negative");
}
stdx::unique_lock<stdx::mutex> lk(_mutex);
auto pred = [this, expectedState]() { return _memberState == expectedState; };
if (!_memberStateChange.wait_for(lk, timeout.toSystemDuration(), pred)) {
return Status(ErrorCodes::ExceededTimeLimit,
str::stream() << "Timed out waiting for state to become "
<< expectedState.toString()
<< ". Current state is "
<< _memberState.toString());
}
return Status::OK();
}
Seconds ReplicationCoordinatorImpl::getSlaveDelaySecs() const {
stdx::lock_guard<stdx::mutex> lk(_mutex);
invariant(_rsConfig.isInitialized());
if (_selfIndex == -1) {
// We aren't currently in the set. Return 0 seconds so we can clear out the applier's
// queue of work.
return Seconds(0);
}
return _rsConfig.getMemberAt(_selfIndex).getSlaveDelay();
}
void ReplicationCoordinatorImpl::clearSyncSourceBlacklist() {
LockGuard topoLock(_topoMutex);
_topCoord->clearSyncSourceBlacklist();
}
ReplicationExecutor::EventHandle ReplicationCoordinatorImpl::setFollowerMode_nonBlocking(
const MemberState& newState, bool* success) {
StatusWith<ReplicationExecutor::EventHandle> finishedSettingFollowerMode =
_replExecutor.makeEvent();
if (finishedSettingFollowerMode.getStatus() == ErrorCodes::ShutdownInProgress) {
return ReplicationExecutor::EventHandle();
}
fassert(18812, finishedSettingFollowerMode.getStatus());
_setFollowerModeFinish(newState, finishedSettingFollowerMode.getValue(), success);
return finishedSettingFollowerMode.getValue();
}
bool ReplicationCoordinatorImpl::setFollowerMode(const MemberState& newState) {
bool success = false;
if (auto eventHandle = setFollowerMode_nonBlocking(newState, &success)) {
_replExecutor.waitForEvent(eventHandle);
}
return success;
}
void ReplicationCoordinatorImpl::_setFollowerModeFinish(
const MemberState& newState,
const ReplicationExecutor::EventHandle& finishedSettingFollowerMode,
bool* success) {
LockGuard topoLock(_topoMutex);
if (newState == _topCoord->getMemberState()) {
*success = true;
_replExecutor.signalEvent(finishedSettingFollowerMode);
return;
}
if (_topCoord->getRole() == TopologyCoordinator::Role::leader) {
*success = false;
_replExecutor.signalEvent(finishedSettingFollowerMode);
return;
}
if (_topCoord->getRole() == TopologyCoordinator::Role::candidate) {
// We are a candidate, which means _topCoord believes us to be in state RS_SECONDARY, and
// we know that newState != RS_SECONDARY because we would have returned early, above if
// the old and new state were equal. So, cancel the running election and try again to
// finish setting the follower mode.
if (isV1ElectionProtocol()) {
invariant(_voteRequester);
_voteRequester->cancel();
} else {
invariant(_freshnessChecker);
_freshnessChecker->cancel();
if (_electCmdRunner) {
_electCmdRunner->cancel();
}
}
_replExecutor.onEvent(
_electionFinishedEvent,
_wrapAsCallbackFn(stdx::bind(&ReplicationCoordinatorImpl::_setFollowerModeFinish,
this,
newState,
finishedSettingFollowerMode,
success)));
return;
}
stdx::unique_lock<stdx::mutex> lk(_mutex);
_topCoord->setFollowerMode(newState.s);
const PostMemberStateUpdateAction action = _updateMemberStateFromTopologyCoordinator_inlock();
lk.unlock();
_performPostMemberStateUpdateAction(action);
*success = true;
_replExecutor.signalEvent(finishedSettingFollowerMode);
}
ReplicationCoordinator::ApplierState ReplicationCoordinatorImpl::getApplierState() {
stdx::lock_guard<stdx::mutex> lk(_mutex);
return _applierState;
}
void ReplicationCoordinatorImpl::signalDrainComplete(OperationContext* opCtx,
long long termWhenBufferIsEmpty) {
// This logic is a little complicated in order to avoid acquiring the global exclusive lock
// unnecessarily. This is important because the applier may call signalDrainComplete()
// whenever it wants, not only when the ReplicationCoordinator is expecting it.
//
// The steps are:
// 1.) Check to see if we're waiting for this signal. If not, return early.
// 2.) Otherwise, release the mutex while acquiring the global exclusive lock,
// since that might take a while (NB there's a deadlock cycle otherwise, too).
// 3.) Re-check to see if we've somehow left drain mode. If we have not, clear
// producer and applier's states, set the flag allowing non-local database writes and
// drop the mutex. At this point, no writes can occur from other threads, due to the
// global exclusive lock.
// 4.) Drop all temp collections, and log the drops to the oplog.
// 5.) Log transition to primary in the oplog and set that OpTime as the floor for what we will
// consider to be committed.
// 6.) Drop the global exclusive lock.
//
// Because replicatable writes are forbidden while in drain mode, and we don't exit drain
// mode until we have the global exclusive lock, which forbids all other threads from making
// writes, we know that from the time that _canAcceptNonLocalWrites is set until
// this method returns, no external writes will be processed. This is important so that a new
// temp collection isn't introduced on the new primary before we drop all the temp collections.
// When we go to drop all temp collections, we must replicate the drops.
invariant(opCtx->writesAreReplicated());
stdx::unique_lock<stdx::mutex> lk(_mutex);
if (_applierState != ApplierState::Draining) {
return;
}
lk.unlock();
_externalState->onDrainComplete(opCtx);
if (MONGO_FAIL_POINT(transitionToPrimaryHangBeforeTakingGlobalExclusiveLock)) {
log() << "transition to primary - "
"transitionToPrimaryHangBeforeTakingGlobalExclusiveLock fail point enabled. "
"Blocking until fail point is disabled.";
while (MONGO_FAIL_POINT(transitionToPrimaryHangBeforeTakingGlobalExclusiveLock)) {
mongo::sleepsecs(1);
{
stdx::lock_guard<stdx::mutex> lock(_mutex);
if (_inShutdown) {
break;
}
}
}
}
ScopedTransaction transaction(opCtx, MODE_X);
Lock::GlobalWrite globalWriteLock(opCtx->lockState());
lk.lock();
// Exit drain mode when the buffer is empty in the current term and we're in Draining mode.
if (_applierState != ApplierState::Draining || termWhenBufferIsEmpty != _cachedTerm) {
return;
}
_applierState = ApplierState::Stopped;
_drainFinishedCond.notify_all();
invariant(_getMemberState_inlock().primary());
invariant(!_canAcceptNonLocalWrites);
_canAcceptNonLocalWrites = true;
lk.unlock();
_setFirstOpTimeOfMyTerm(_externalState->onTransitionToPrimary(opCtx, isV1ElectionProtocol()));
lk.lock();
// Must calculate the commit level again because firstOpTimeOfMyTerm wasn't set when we logged
// our election in onTransitionToPrimary(), above.
_updateLastCommittedOpTime_inlock();
log() << "transition to primary complete; database writes are now permitted" << rsLog;
_externalState->startNoopWriter(_getMyLastAppliedOpTime_inlock());
}
Status ReplicationCoordinatorImpl::waitForDrainFinish(Milliseconds timeout) {
if (timeout < Milliseconds(0)) {
return Status(ErrorCodes::BadValue, "Timeout duration cannot be negative");
}
stdx::unique_lock<stdx::mutex> lk(_mutex);
auto pred = [this]() { return _applierState != ApplierState::Draining; };
if (!_drainFinishedCond.wait_for(lk, timeout.toSystemDuration(), pred)) {
return Status(ErrorCodes::ExceededTimeLimit,
"Timed out waiting to finish draining applier buffer");
}
return Status::OK();
}
void ReplicationCoordinatorImpl::signalUpstreamUpdater() {
_externalState->forwardSlaveProgress();
}
ReplicationCoordinatorImpl::SlaveInfo* ReplicationCoordinatorImpl::_findSlaveInfoByMemberID_inlock(
int memberId) {
for (SlaveInfoVector::iterator it = _slaveInfo.begin(); it != _slaveInfo.end(); ++it) {
if (it->memberId == memberId) {
return &(*it);
}
}
return NULL;
}
ReplicationCoordinatorImpl::SlaveInfo* ReplicationCoordinatorImpl::_findSlaveInfoByRID_inlock(
const OID& rid) {
for (SlaveInfoVector::iterator it = _slaveInfo.begin(); it != _slaveInfo.end(); ++it) {
if (it->rid == rid) {
return &(*it);
}
}
return NULL;
}
void ReplicationCoordinatorImpl::_addSlaveInfo_inlock(const SlaveInfo& slaveInfo) {
invariant(getReplicationMode() == modeMasterSlave);
_slaveInfo.push_back(slaveInfo);
_updateLastCommittedOpTime_inlock();
// Wake up any threads waiting for replication that now have their replication
// check satisfied
_wakeReadyWaiters_inlock();
}
void ReplicationCoordinatorImpl::_updateSlaveInfoAppliedOpTime_inlock(SlaveInfo* slaveInfo,
const OpTime& opTime) {
slaveInfo->lastAppliedOpTime = opTime;
slaveInfo->lastUpdate = _replExecutor.now();
slaveInfo->down = false;
_updateLastCommittedOpTime_inlock();
// Wake up any threads waiting for replication that now have their replication
// check satisfied
_wakeReadyWaiters_inlock();
}
void ReplicationCoordinatorImpl::_updateSlaveInfoDurableOpTime_inlock(SlaveInfo* slaveInfo,
const OpTime& opTime) {
// lastAppliedOpTime cannot be behind lastDurableOpTime.
if (slaveInfo->lastAppliedOpTime < opTime) {
log() << "Durable progress (" << opTime << ") is ahead of the applied progress ("
<< slaveInfo->lastAppliedOpTime << ". This is likely due to a "
"rollback. slaveInfo: "
<< slaveInfo->toString();
return;
}
slaveInfo->lastDurableOpTime = opTime;
slaveInfo->lastUpdate = _replExecutor.now();
slaveInfo->down = false;
_updateLastCommittedOpTime_inlock();
// Wake up any threads waiting for replication that now have their replication
// check satisfied
_wakeReadyWaiters_inlock();
}
void ReplicationCoordinatorImpl::_updateSlaveInfoFromConfig_inlock() {
invariant(_settings.usingReplSets());
SlaveInfoVector oldSlaveInfos;
_slaveInfo.swap(oldSlaveInfos);
if (_selfIndex == -1) {
// If we aren't in the config then the only data we care about is for ourself
for (SlaveInfoVector::const_iterator it = oldSlaveInfos.begin(); it != oldSlaveInfos.end();
++it) {
if (it->self) {
SlaveInfo slaveInfo = *it;
slaveInfo.memberId = -1;
_slaveInfo.push_back(slaveInfo);
return;
}
}
invariant(false); // There should always have been an entry for ourself
}
for (int i = 0; i < _rsConfig.getNumMembers(); ++i) {
const MemberConfig& memberConfig = _rsConfig.getMemberAt(i);
int memberId = memberConfig.getId();
const HostAndPort& memberHostAndPort = memberConfig.getHostAndPort();
SlaveInfo slaveInfo;
// Check if the node existed with the same member ID and hostname in the old data
for (SlaveInfoVector::const_iterator it = oldSlaveInfos.begin(); it != oldSlaveInfos.end();
++it) {
if ((it->memberId == memberId && it->hostAndPort == memberHostAndPort) ||
(i == _selfIndex && it->self)) {
slaveInfo = *it;
}
}
// Make sure you have the most up-to-date info for member ID and hostAndPort.
slaveInfo.memberId = memberId;
slaveInfo.hostAndPort = memberHostAndPort;
_slaveInfo.push_back(slaveInfo);
}
invariant(static_cast<int>(_slaveInfo.size()) == _rsConfig.getNumMembers());
}
size_t ReplicationCoordinatorImpl::_getMyIndexInSlaveInfo_inlock() const {
if (getReplicationMode() == modeMasterSlave) {
// Self data always lives in the first entry in _slaveInfo for master/slave
return 0;
} else {
invariant(_settings.usingReplSets());
if (_selfIndex == -1) {
invariant(_slaveInfo.size() == 1);
return 0;
} else {
return _selfIndex;
}
}
}
Status ReplicationCoordinatorImpl::setLastOptimeForSlave(const OID& rid, const Timestamp& ts) {
stdx::unique_lock<stdx::mutex> lock(_mutex);
massert(28576,
"Received an old style replication progress update, which is only used for Master/"
"Slave replication now, but this node is not using Master/Slave replication. "
"This is likely caused by an old (pre-2.6) member syncing from this node.",
getReplicationMode() == modeMasterSlave);
// term == -1 for master-slave
OpTime opTime(ts, OpTime::kUninitializedTerm);
SlaveInfo* slaveInfo = _findSlaveInfoByRID_inlock(rid);
if (slaveInfo) {
if (slaveInfo->lastAppliedOpTime < opTime) {
_updateSlaveInfoAppliedOpTime_inlock(slaveInfo, opTime);
}
} else {
SlaveInfo newSlaveInfo;
newSlaveInfo.rid = rid;
newSlaveInfo.lastAppliedOpTime = opTime;
_addSlaveInfo_inlock(newSlaveInfo);
}
return Status::OK();
}
void ReplicationCoordinatorImpl::setMyHeartbeatMessage(const std::string& msg) {
LockGuard topoLock(_topoMutex);
_topCoord->setMyHeartbeatMessage(_replExecutor.now(), msg);
}
void ReplicationCoordinatorImpl::setMyLastAppliedOpTimeForward(const OpTime& opTime) {
stdx::unique_lock<stdx::mutex> lock(_mutex);
if (opTime > _getMyLastAppliedOpTime_inlock()) {
const bool allowRollback = false;
_setMyLastAppliedOpTime_inlock(opTime, allowRollback);
_reportUpstream_inlock(std::move(lock));
}
}
void ReplicationCoordinatorImpl::setMyLastDurableOpTimeForward(const OpTime& opTime) {
stdx::unique_lock<stdx::mutex> lock(_mutex);
if (opTime > _getMyLastDurableOpTime_inlock()) {
_setMyLastDurableOpTime_inlock(opTime, false);
_reportUpstream_inlock(std::move(lock));
}
}
void ReplicationCoordinatorImpl::setMyLastAppliedOpTime(const OpTime& opTime) {
stdx::unique_lock<stdx::mutex> lock(_mutex);
_setMyLastAppliedOpTime_inlock(opTime, false);
_reportUpstream_inlock(std::move(lock));
}
void ReplicationCoordinatorImpl::setMyLastDurableOpTime(const OpTime& opTime) {
stdx::unique_lock<stdx::mutex> lock(_mutex);
_setMyLastDurableOpTime_inlock(opTime, false);
_reportUpstream_inlock(std::move(lock));
}
void ReplicationCoordinatorImpl::resetMyLastOpTimes() {
stdx::unique_lock<stdx::mutex> lock(_mutex);
_resetMyLastOpTimes_inlock();
_reportUpstream_inlock(std::move(lock));
}
void ReplicationCoordinatorImpl::_resetMyLastOpTimes_inlock() {
LOG(1) << "resetting durable/applied optimes.";
// Reset to uninitialized OpTime
_setMyLastAppliedOpTime_inlock(OpTime(), true);
_setMyLastDurableOpTime_inlock(OpTime(), true);
}
void ReplicationCoordinatorImpl::_reportUpstream_inlock(stdx::unique_lock<stdx::mutex> lock) {
invariant(lock.owns_lock());
if (getReplicationMode() != modeReplSet) {
return;
}
if (_getMemberState_inlock().primary()) {
return;
}
lock.unlock();
_externalState->forwardSlaveProgress(); // Must do this outside _mutex
}
void ReplicationCoordinatorImpl::_setMyLastAppliedOpTime_inlock(const OpTime& opTime,
bool isRollbackAllowed) {
SlaveInfo* mySlaveInfo = &_slaveInfo[_getMyIndexInSlaveInfo_inlock()];
invariant(isRollbackAllowed || mySlaveInfo->lastAppliedOpTime <= opTime);
_updateSlaveInfoAppliedOpTime_inlock(mySlaveInfo, opTime);
_opTimeWaiterList.signalAndRemoveIf_inlock(
[opTime](WaiterInfo* waiter) { return waiter->opTime <= opTime; });
}
void ReplicationCoordinatorImpl::_setMyLastDurableOpTime_inlock(const OpTime& opTime,
bool isRollbackAllowed) {
SlaveInfo* mySlaveInfo = &_slaveInfo[_getMyIndexInSlaveInfo_inlock()];
invariant(isRollbackAllowed || mySlaveInfo->lastDurableOpTime <= opTime);
// lastAppliedOpTime cannot be behind lastDurableOpTime.
if (mySlaveInfo->lastAppliedOpTime < opTime) {
log() << "My durable progress (" << opTime << ") is ahead of my applied progress ("
<< mySlaveInfo->lastAppliedOpTime << ". This is likely due to a "
"rollback. slaveInfo: "
<< mySlaveInfo->toString();
return;
}
_updateSlaveInfoDurableOpTime_inlock(mySlaveInfo, opTime);
}
OpTime ReplicationCoordinatorImpl::getMyLastAppliedOpTime() const {
stdx::lock_guard<stdx::mutex> lock(_mutex);
return _getMyLastAppliedOpTime_inlock();
}
OpTime ReplicationCoordinatorImpl::getMyLastDurableOpTime() const {
stdx::lock_guard<stdx::mutex> lock(_mutex);
return _getMyLastDurableOpTime_inlock();
}
Status ReplicationCoordinatorImpl::waitUntilOpTimeForRead(OperationContext* opCtx,
const ReadConcernArgs& settings) {
// We should never wait for replication if we are holding any locks, because this can
// potentially block for long time while doing network activity.
if (opCtx->lockState()->isLocked()) {
return {ErrorCodes::IllegalOperation,
"Waiting for replication not allowed while holding a lock"};
}
const bool isMajorityReadConcern =
settings.getLevel() == ReadConcernLevel::kMajorityReadConcern;
if (isMajorityReadConcern && !getSettings().isMajorityReadConcernEnabled()) {
// This is an opt-in feature. Fail if the user didn't opt-in.
return {ErrorCodes::ReadConcernMajorityNotEnabled,
"Majority read concern requested, but server was not started with "
"--enableMajorityReadConcern."};
}
const auto targetOpTime = settings.getOpTime();
if (targetOpTime.isNull()) {
return Status::OK();
}
if (getReplicationMode() != repl::ReplicationCoordinator::modeReplSet) {
// For master/slave and standalone nodes, readAfterOpTime is not supported, so we return an
// error. However, we consider all writes "committed" and can treat MajorityReadConcern as
// LocalReadConcern, which is immediately satisfied since there is no OpTime to wait for.
return {ErrorCodes::NotAReplicaSet,
"node needs to be a replica set member to use read concern"};
}
stdx::unique_lock<stdx::mutex> lock(_mutex);
if (isMajorityReadConcern && !_externalState->snapshotsEnabled()) {
return {ErrorCodes::CommandNotSupported,
"Current storage engine does not support majority readConcerns"};
}
auto getCurrentOpTime = [this, isMajorityReadConcern] {
auto committedOptime =
_currentCommittedSnapshot ? _currentCommittedSnapshot->opTime : OpTime();
return isMajorityReadConcern ? committedOptime : _getMyLastAppliedOpTime_inlock();
};
if (isMajorityReadConcern && targetOpTime > getCurrentOpTime()) {
LOG(1) << "waitUntilOpTime: waiting for optime:" << targetOpTime
<< " to be in a snapshot -- current snapshot: " << getCurrentOpTime();
}
while (targetOpTime > getCurrentOpTime()) {
if (_inShutdown) {
return {ErrorCodes::ShutdownInProgress, "Shutdown in progress"};
}
// If we are doing a majority read concern we only need to wait for a new snapshot.
if (isMajorityReadConcern) {
// Wait for a snapshot that meets our needs (< targetOpTime).
LOG(3) << "waitUntilOpTime: waiting for a new snapshot until " << opCtx->getDeadline();
auto waitStatus =
opCtx->waitForConditionOrInterruptNoAssert(_currentCommittedSnapshotCond, lock);
if (!waitStatus.isOK()) {
return waitStatus;
}
LOG(3) << "Got notified of new snapshot: " << _currentCommittedSnapshot->toString();
continue;
}
// We just need to wait for the opTime to catch up to what we need (not majority RC).
stdx::condition_variable condVar;
WaiterInfoGuard waitInfo(
&_opTimeWaiterList, opCtx->getOpID(), targetOpTime, nullptr, &condVar);
LOG(3) << "waituntilOpTime: waiting for OpTime " << waitInfo.waiter << " until "
<< opCtx->getDeadline();
auto waitStatus = opCtx->waitForConditionOrInterruptNoAssert(condVar, lock);
if (!waitStatus.isOK()) {
return waitStatus;
}
}
return Status::OK();
}
OpTime ReplicationCoordinatorImpl::_getMyLastAppliedOpTime_inlock() const {
return _slaveInfo[_getMyIndexInSlaveInfo_inlock()].lastAppliedOpTime;
}
OpTime ReplicationCoordinatorImpl::_getMyLastDurableOpTime_inlock() const {
return _slaveInfo[_getMyIndexInSlaveInfo_inlock()].lastDurableOpTime;
}
Status ReplicationCoordinatorImpl::setLastDurableOptime_forTest(long long cfgVer,
long long memberId,
const OpTime& opTime) {
stdx::lock_guard<stdx::mutex> lock(_mutex);
invariant(getReplicationMode() == modeReplSet);
const UpdatePositionArgs::UpdateInfo update(OpTime(), opTime, cfgVer, memberId);
long long configVersion;
const auto status = _setLastOptime_inlock(update, &configVersion);
_updateLastCommittedOpTime_inlock();
return status;
}
Status ReplicationCoordinatorImpl::setLastAppliedOptime_forTest(long long cfgVer,
long long memberId,
const OpTime& opTime) {
stdx::lock_guard<stdx::mutex> lock(_mutex);
invariant(getReplicationMode() == modeReplSet);
const UpdatePositionArgs::UpdateInfo update(opTime, OpTime(), cfgVer, memberId);
long long configVersion;
const auto status = _setLastOptime_inlock(update, &configVersion);
_updateLastCommittedOpTime_inlock();
return status;
}
Status ReplicationCoordinatorImpl::_setLastOptime_inlock(
const OldUpdatePositionArgs::UpdateInfo& args, long long* configVersion) {
if (_selfIndex == -1) {
// Ignore updates when we're in state REMOVED
return Status(ErrorCodes::NotMasterOrSecondary,
"Received replSetUpdatePosition command but we are in state REMOVED");
}
invariant(getReplicationMode() == modeReplSet);
if (args.memberId < 0) {
std::string errmsg = str::stream()
<< "Received replSetUpdatePosition for node with memberId " << args.memberId
<< " which is negative and therefore invalid";
LOG(1) << errmsg;
return Status(ErrorCodes::NodeNotFound, errmsg);
}
if (args.memberId == _rsConfig.getMemberAt(_selfIndex).getId()) {
// Do not let remote nodes tell us what our optime is.
return Status::OK();
}
LOG(2) << "received notification that node with memberID " << args.memberId
<< " in config with version " << args.cfgver
<< " has durably reached optime: " << args.ts;
SlaveInfo* slaveInfo = NULL;
if (args.cfgver != _rsConfig.getConfigVersion()) {
std::string errmsg = str::stream()
<< "Received replSetUpdatePosition for node with memberId " << args.memberId
<< " whose config version of " << args.cfgver << " doesn't match our config version of "
<< _rsConfig.getConfigVersion();
LOG(1) << errmsg;
*configVersion = _rsConfig.getConfigVersion();
return Status(ErrorCodes::InvalidReplicaSetConfig, errmsg);
}
slaveInfo = _findSlaveInfoByMemberID_inlock(args.memberId);
if (!slaveInfo) {
invariant(!_rsConfig.findMemberByID(args.memberId));
std::string errmsg = str::stream()
<< "Received replSetUpdatePosition for node with memberId " << args.memberId
<< " which doesn't exist in our config";
LOG(1) << errmsg;
return Status(ErrorCodes::NodeNotFound, errmsg);
}
invariant(args.memberId == slaveInfo->memberId);
LOG(3) << "Node with memberID " << args.memberId << " has durably applied operationss through "
<< slaveInfo->lastDurableOpTime << " and has applied operations through "
<< slaveInfo->lastAppliedOpTime << "; updating to new durable operation with timestamp "
<< args.ts;
// Only update remote optimes if they increase.
if (slaveInfo->lastAppliedOpTime < args.ts) {
_updateSlaveInfoAppliedOpTime_inlock(slaveInfo, args.ts);
}
if (slaveInfo->lastDurableOpTime < args.ts) {
_updateSlaveInfoDurableOpTime_inlock(slaveInfo, args.ts);
}
// Update liveness for this node.
slaveInfo->lastUpdate = _replExecutor.now();
slaveInfo->down = false;
_cancelAndRescheduleLivenessUpdate_inlock(args.memberId);
return Status::OK();
}
Status ReplicationCoordinatorImpl::_setLastOptime_inlock(const UpdatePositionArgs::UpdateInfo& args,
long long* configVersion) {
if (_selfIndex == -1) {
// Ignore updates when we're in state REMOVED.
return Status(ErrorCodes::NotMasterOrSecondary,
"Received replSetUpdatePosition command but we are in state REMOVED");
}
invariant(getReplicationMode() == modeReplSet);
if (args.memberId < 0) {
std::string errmsg = str::stream()
<< "Received replSetUpdatePosition for node with memberId " << args.memberId
<< " which is negative and therefore invalid";
LOG(1) << errmsg;
return Status(ErrorCodes::NodeNotFound, errmsg);
}
if (args.memberId == _rsConfig.getMemberAt(_selfIndex).getId()) {
// Do not let remote nodes tell us what our optime is.
return Status::OK();
}
LOG(2) << "received notification that node with memberID " << args.memberId
<< " in config with version " << args.cfgver
<< " has reached optime: " << args.appliedOpTime
<< " and is durable through: " << args.durableOpTime;
SlaveInfo* slaveInfo = NULL;
if (args.cfgver != _rsConfig.getConfigVersion()) {
std::string errmsg = str::stream()
<< "Received replSetUpdatePosition for node with memberId " << args.memberId
<< " whose config version of " << args.cfgver << " doesn't match our config version of "
<< _rsConfig.getConfigVersion();
LOG(1) << errmsg;
*configVersion = _rsConfig.getConfigVersion();
return Status(ErrorCodes::InvalidReplicaSetConfig, errmsg);
}
slaveInfo = _findSlaveInfoByMemberID_inlock(args.memberId);
if (!slaveInfo) {
invariant(!_rsConfig.findMemberByID(args.memberId));
std::string errmsg = str::stream()
<< "Received replSetUpdatePosition for node with memberId " << args.memberId
<< " which doesn't exist in our config";
LOG(1) << errmsg;
return Status(ErrorCodes::NodeNotFound, errmsg);
}
invariant(args.memberId == slaveInfo->memberId);
LOG(3) << "Node with memberID " << args.memberId << " currently has optime "
<< slaveInfo->lastAppliedOpTime << " durable through " << slaveInfo->lastDurableOpTime
<< "; updating to optime " << args.appliedOpTime << " and durable through "
<< args.durableOpTime;
// Only update remote optimes if they increase.
if (slaveInfo->lastAppliedOpTime < args.appliedOpTime) {
_updateSlaveInfoAppliedOpTime_inlock(slaveInfo, args.appliedOpTime);
}
if (slaveInfo->lastDurableOpTime < args.durableOpTime) {
_updateSlaveInfoDurableOpTime_inlock(slaveInfo, args.durableOpTime);
}
// Update liveness for this node.
slaveInfo->lastUpdate = _replExecutor.now();
slaveInfo->down = false;
_cancelAndRescheduleLivenessUpdate_inlock(args.memberId);
return Status::OK();
}
bool ReplicationCoordinatorImpl::_doneWaitingForReplication_inlock(
const OpTime& opTime, SnapshotName minSnapshot, const WriteConcernOptions& writeConcern) {
// The syncMode cannot be unset.
invariant(writeConcern.syncMode != WriteConcernOptions::SyncMode::UNSET);
Status status = _checkIfWriteConcernCanBeSatisfied_inlock(writeConcern);
if (!status.isOK()) {
return true;
}
const bool useDurableOpTime = writeConcern.syncMode == WriteConcernOptions::SyncMode::JOURNAL;
if (writeConcern.wMode.empty()) {
return _haveNumNodesReachedOpTime_inlock(opTime, writeConcern.wNumNodes, useDurableOpTime);
}
StringData patternName;
if (writeConcern.wMode == WriteConcernOptions::kMajority) {
if (_externalState->snapshotsEnabled()) {
// Make sure we have a valid "committed" snapshot up to the needed optime.
if (!_currentCommittedSnapshot) {
return false;
}
// Wait for the "current" snapshot to advance to/past the opTime.
const auto haveSnapshot = (_currentCommittedSnapshot->opTime >= opTime &&
_currentCommittedSnapshot->name >= minSnapshot);
if (!haveSnapshot) {
LOG(1) << "Required snapshot optime: " << opTime << " is not yet part of the "
<< "current 'committed' snapshot: " << *_currentCommittedSnapshot;
return false;
}
// Fallthrough to wait for "majority" write concern.
}
// Continue and wait for replication to the majority (of voters).
// *** Needed for J:True, writeConcernMajorityShouldJournal:False (appliedOpTime snapshot).
patternName = ReplSetConfig::kMajorityWriteConcernModeName;
} else {
patternName = writeConcern.wMode;
}
StatusWith<ReplSetTagPattern> tagPattern = _rsConfig.findCustomWriteMode(patternName);
if (!tagPattern.isOK()) {
return true;
}
return _haveTaggedNodesReachedOpTime_inlock(opTime, tagPattern.getValue(), useDurableOpTime);
}
bool ReplicationCoordinatorImpl::_haveNumNodesReachedOpTime_inlock(const OpTime& targetOpTime,
int numNodes,
bool durablyWritten) {
// Replication progress that is for some reason ahead of us should not allow us to
// satisfy a write concern if we aren't caught up ourselves.
OpTime myOpTime =
durablyWritten ? _getMyLastDurableOpTime_inlock() : _getMyLastAppliedOpTime_inlock();
if (myOpTime < targetOpTime) {
return false;
}
for (SlaveInfoVector::iterator it = _slaveInfo.begin(); it != _slaveInfo.end(); ++it) {
const OpTime& slaveTime = durablyWritten ? it->lastDurableOpTime : it->lastAppliedOpTime;
if (slaveTime >= targetOpTime) {
--numNodes;
}
if (numNodes <= 0) {
return true;
}
}
return false;
}
bool ReplicationCoordinatorImpl::_haveTaggedNodesReachedOpTime_inlock(
const OpTime& opTime, const ReplSetTagPattern& tagPattern, bool durablyWritten) {
ReplSetTagMatch matcher(tagPattern);
for (SlaveInfoVector::iterator it = _slaveInfo.begin(); it != _slaveInfo.end(); ++it) {
const OpTime& slaveTime = durablyWritten ? it->lastDurableOpTime : it->lastAppliedOpTime;
if (slaveTime >= opTime) {
// This node has reached the desired optime, now we need to check if it is a part
// of the tagPattern.
const MemberConfig* memberConfig = _rsConfig.findMemberByID(it->memberId);
invariant(memberConfig);
for (MemberConfig::TagIterator it = memberConfig->tagsBegin();
it != memberConfig->tagsEnd();
++it) {
if (matcher.update(*it)) {
return true;
}
}
}
}
return false;
}
ReplicationCoordinator::StatusAndDuration ReplicationCoordinatorImpl::awaitReplication(
OperationContext* opCtx, const OpTime& opTime, const WriteConcernOptions& writeConcern) {
Timer timer;
WriteConcernOptions fixedWriteConcern = populateUnsetWriteConcernOptionsSyncMode(writeConcern);
stdx::unique_lock<stdx::mutex> lock(_mutex);
auto status =
_awaitReplication_inlock(&lock, opCtx, opTime, SnapshotName::min(), fixedWriteConcern);
return {std::move(status), duration_cast<Milliseconds>(timer.elapsed())};
}
ReplicationCoordinator::StatusAndDuration
ReplicationCoordinatorImpl::awaitReplicationOfLastOpForClient(
OperationContext* opCtx, const WriteConcernOptions& writeConcern) {
Timer timer;
WriteConcernOptions fixedWriteConcern = populateUnsetWriteConcernOptionsSyncMode(writeConcern);
stdx::unique_lock<stdx::mutex> lock(_mutex);
const auto& clientInfo = ReplClientInfo::forClient(opCtx->getClient());
auto status = _awaitReplication_inlock(
&lock, opCtx, clientInfo.getLastOp(), clientInfo.getLastSnapshot(), fixedWriteConcern);
return {std::move(status), duration_cast<Milliseconds>(timer.elapsed())};
}
Status ReplicationCoordinatorImpl::_awaitReplication_inlock(
stdx::unique_lock<stdx::mutex>* lock,
OperationContext* opCtx,
const OpTime& opTime,
SnapshotName minSnapshot,
const WriteConcernOptions& writeConcern) {
// We should never wait for replication if we are holding any locks, because this can
// potentially block for long time while doing network activity.
if (opCtx->lockState()->isLocked()) {
return {ErrorCodes::IllegalOperation,
"Waiting for replication not allowed while holding a lock"};
}
const Mode replMode = getReplicationMode();
if (replMode == modeNone) {
// no replication check needed (validated above)
return Status::OK();
}
if (replMode == modeMasterSlave && writeConcern.wMode == WriteConcernOptions::kMajority) {
// with master/slave, majority is equivalent to w=1
return Status::OK();
}
if (opTime.isNull() && minSnapshot == SnapshotName::min()) {
// If waiting for the empty optime, always say it's been replicated.
return Status::OK();
}
auto checkForStepDown = [&]() -> Status {
if (replMode == modeReplSet && !_memberState.primary()) {
return {ErrorCodes::PrimarySteppedDown,
"Primary stepped down while waiting for replication"};
}
if (opTime.getTerm() != _cachedTerm) {
return {
ErrorCodes::PrimarySteppedDown,
str::stream() << "Term changed from " << opTime.getTerm() << " to " << _cachedTerm
<< " while waiting for replication, indicating that this node must "
"have stepped down."};
}
if (_stepDownPending) {
return {ErrorCodes::PrimarySteppedDown,
"Received stepdown request while waiting for replication"};
}
return Status::OK();
};
Status stepdownStatus = checkForStepDown();
if (!stepdownStatus.isOK()) {
return stepdownStatus;
}
auto interruptStatus = opCtx->checkForInterruptNoAssert();
if (!interruptStatus.isOK()) {
return interruptStatus;
}
if (writeConcern.wMode.empty()) {
if (writeConcern.wNumNodes < 1) {
return Status::OK();
} else if (writeConcern.wNumNodes == 1 && _getMyLastAppliedOpTime_inlock() >= opTime) {
return Status::OK();
}
}
auto clockSource = opCtx->getServiceContext()->getFastClockSource();
const auto wTimeoutDate = [&]() -> const Date_t {
if (writeConcern.wDeadline != Date_t::max()) {
return writeConcern.wDeadline;
}
if (writeConcern.wTimeout == WriteConcernOptions::kNoTimeout) {
return Date_t::max();
}
return clockSource->now() + clockSource->getPrecision() +
Milliseconds{writeConcern.wTimeout};
}();
// Must hold _mutex before constructing waitInfo as it will modify _replicationWaiterList
stdx::condition_variable condVar;
WaiterInfoGuard waitInfo(
&_replicationWaiterList, opCtx->getOpID(), opTime, &writeConcern, &condVar);
while (!_doneWaitingForReplication_inlock(opTime, minSnapshot, writeConcern)) {
if (_inShutdown) {
return {ErrorCodes::ShutdownInProgress, "Replication is being shut down"};
}
auto status = opCtx->waitForConditionOrInterruptNoAssertUntil(condVar, *lock, wTimeoutDate);
if (!status.isOK()) {
return status.getStatus();
}
if (status.getValue() == stdx::cv_status::timeout) {
if (Command::testCommandsEnabled) {
// log state of replica set on timeout to help with diagnosis.
BSONObjBuilder progress;
_appendSlaveInfoData_inlock(&progress);
log() << "Replication for failed WC: " << writeConcern.toBSON()
<< ", waitInfo:" << waitInfo.waiter.toBSON()
<< ", progress: " << progress.done();
}
return {ErrorCodes::WriteConcernFailed, "waiting for replication timed out"};
}
stepdownStatus = checkForStepDown();
if (!stepdownStatus.isOK()) {
return stepdownStatus;
}
}
return _checkIfWriteConcernCanBeSatisfied_inlock(writeConcern);
}
Status ReplicationCoordinatorImpl::stepDown(OperationContext* opCtx,
const bool force,
const Milliseconds& waitTime,
const Milliseconds& stepdownTime) {
const Date_t startTime = _replExecutor.now();
const Date_t stepDownUntil = startTime + stepdownTime;
const Date_t waitUntil = startTime + waitTime;
if (!getMemberState().primary()) {
// Note this check is inherently racy - it's always possible for the node to
// stepdown from some other path before we acquire the global shared lock, but
// that's okay because we are resiliant to that happening in _stepDownContinue.
return {ErrorCodes::NotMaster, "not primary so can't step down"};
}
Lock::GlobalLock globalReadLock(opCtx->lockState(), MODE_S, Lock::GlobalLock::EnqueueOnly());
// We've requested the global shared lock which will stop new writes from coming in,
// but existing writes could take a long time to finish, so kill all user operations
// to help us get the global lock faster.
_externalState->killAllUserOperations(opCtx);
globalReadLock.waitForLock(durationCount<Milliseconds>(stepdownTime));
if (!globalReadLock.isLocked()) {
return {ErrorCodes::ExceededTimeLimit,
"Could not acquire the global shared lock within the amount of time "
"specified that we should step down for"};
}
try {
stdx::unique_lock<stdx::mutex> topoLock(_topoMutex);
bool restartHeartbeats = true;
opCtx->checkForInterrupt();
while (!_tryToStepDown(waitUntil, stepDownUntil, force)) {
if (restartHeartbeats) {
// We send out a fresh round of heartbeats because stepping down successfully
// without
// {force: true} is dependent on timely heartbeat data.
stdx::lock_guard<stdx::mutex> lk(_mutex);
_restartHeartbeats_inlock();
restartHeartbeats = false;
}
opCtx->waitForConditionOrInterruptUntil(
_stepDownWaiters, topoLock, std::min(stepDownUntil, waitUntil));
}
} catch (const DBException& ex) {
return ex.toStatus();
}
return Status::OK();
}
void ReplicationCoordinatorImpl::_signalStepDownWaiter_inlock() {
_stepDownWaiters.notify_all();
}
bool ReplicationCoordinatorImpl::_tryToStepDown(const Date_t waitUntil,
const Date_t stepDownUntil,
const bool force) {
if (_topCoord->getRole() != TopologyCoordinator::Role::leader) {
uasserted(ErrorCodes::NotMaster,
"Already stepped down from primary while processing step down request");
}
const Date_t now = _replExecutor.now();
if (now >= stepDownUntil) {
uasserted(ErrorCodes::ExceededTimeLimit,
"By the time we were ready to step down, we were already past the "
"time we were supposed to step down until");
}
const bool forceNow = now >= waitUntil ? force : false;
if (!_topCoord->stepDown(
stepDownUntil, forceNow, getMyLastAppliedOpTime(), getLastCommittedOpTime())) {
if (now >= waitUntil) {
uasserted(ErrorCodes::ExceededTimeLimit,
str::stream() << "No electable secondaries caught up as of "
<< dateToISOStringLocal(now)
<< ". Please use {force: true} to force node to step down.");
}
return false;
}
const auto action = [&] {
stdx::lock_guard<stdx::mutex> lk(_mutex);
return _updateMemberStateFromTopologyCoordinator_inlock();
}();
_performPostMemberStateUpdateAction(action);
// Schedule work to (potentially) step back up once the stepdown period has ended.
_scheduleWorkAt(
stepDownUntil,
stdx::bind(&ReplicationCoordinatorImpl::_handleTimePassing, this, stdx::placeholders::_1));
return true;
}
void ReplicationCoordinatorImpl::_handleTimePassing(
const ReplicationExecutor::CallbackArgs& cbData) {
if (!cbData.status.isOK()) {
return;
}
LockGuard topoLock(_topoMutex);
if (_topCoord->becomeCandidateIfStepdownPeriodOverAndSingleNodeSet(_replExecutor.now())) {
_performPostMemberStateUpdateAction(kActionWinElection);
}
}
bool ReplicationCoordinatorImpl::isMasterForReportingPurposes() {
if (_settings.usingReplSets()) {
stdx::lock_guard<stdx::mutex> lock(_mutex);
if (getReplicationMode() == modeReplSet && _getMemberState_inlock().primary()) {
return true;
}
return false;
}
if (!_settings.isSlave())
return true;
// TODO(dannenberg) replAllDead is bad and should be removed when master slave is removed
if (replAllDead) {
return false;
}
if (_settings.isMaster()) {
// if running with --master --slave, allow.
return true;
}
return false;
}
bool ReplicationCoordinatorImpl::canAcceptWritesForDatabase(OperationContext* opCtx,
StringData dbName) {
// The answer isn't meaningful unless we hold the global lock.
invariant(opCtx->lockState()->isLocked());
return canAcceptWritesForDatabase_UNSAFE(opCtx, dbName);
}
bool ReplicationCoordinatorImpl::canAcceptWritesForDatabase_UNSAFE(OperationContext* opCtx,
StringData dbName) {
// _canAcceptNonLocalWrites is always true for standalone nodes, always false for nodes
// started with --slave, and adjusted based on primary+drain state in replica sets.
//
// That is, stand-alone nodes, non-slave nodes and drained replica set primaries can always
// accept writes. Similarly, writes are always permitted to the "local" database. Finally,
// in the event that a node is started with --slave and --master, we allow writes unless the
// master/slave system has set the replAllDead flag.
if (_canAcceptNonLocalWrites) {
return true;
}
if (dbName == kLocalDB) {
return true;
}
return !replAllDead && _settings.isMaster();
}
bool ReplicationCoordinatorImpl::canAcceptWritesFor(OperationContext* opCtx,
const NamespaceString& ns) {
invariant(opCtx->lockState()->isLocked());
return canAcceptWritesFor_UNSAFE(opCtx, ns);
}
bool ReplicationCoordinatorImpl::canAcceptWritesFor_UNSAFE(OperationContext* opCtx,
const NamespaceString& ns) {
if (_memberState.rollback() && ns.isOplog()) {
return false;
}
StringData dbName = ns.db();
return canAcceptWritesForDatabase_UNSAFE(opCtx, dbName);
}
Status ReplicationCoordinatorImpl::checkCanServeReadsFor(OperationContext* opCtx,
const NamespaceString& ns,
bool slaveOk) {
invariant(opCtx->lockState()->isLocked());
return checkCanServeReadsFor_UNSAFE(opCtx, ns, slaveOk);
}
Status ReplicationCoordinatorImpl::checkCanServeReadsFor_UNSAFE(OperationContext* opCtx,
const NamespaceString& ns,
bool slaveOk) {
auto client = opCtx->getClient();
// Oplog reads are not allowed during STARTUP state, but we make an exception for internal
// reads and master-slave replication. Internel reads are required for cleaning up unfinished
// apply batches. Master-slave never sets the state so we make an exception for it as well.
if (((_memberState.startup() && client->isFromUserConnection() &&
getReplicationMode() == modeReplSet) ||
_memberState.startup2() || _memberState.rollback()) &&
ns.isOplog()) {
return {ErrorCodes::NotMasterOrSecondary,
"Oplog collection reads are not allowed while in the rollback or startup state."};
}
if (client->isInDirectClient()) {
return Status::OK();
}
if (canAcceptWritesFor_UNSAFE(opCtx, ns)) {
return Status::OK();
}
if (_settings.isSlave() || _settings.isMaster()) {
return Status::OK();
}
if (slaveOk) {
if (_canServeNonLocalReads.loadRelaxed()) {
return Status::OK();
}
return Status(ErrorCodes::NotMasterOrSecondary,
"not master or secondary; cannot currently read from this replSet member");
}
return Status(ErrorCodes::NotMasterNoSlaveOk, "not master and slaveOk=false");
}
bool ReplicationCoordinatorImpl::isInPrimaryOrSecondaryState() const {
return _canServeNonLocalReads.loadRelaxed();
}
bool ReplicationCoordinatorImpl::shouldRelaxIndexConstraints(OperationContext* opCtx,
const NamespaceString& ns) {
return !canAcceptWritesFor(opCtx, ns);
}
OID ReplicationCoordinatorImpl::getElectionId() {
stdx::lock_guard<stdx::mutex> lock(_mutex);
return _electionId;
}
OID ReplicationCoordinatorImpl::getMyRID() const {
stdx::lock_guard<stdx::mutex> lock(_mutex);
return _getMyRID_inlock();
}
OID ReplicationCoordinatorImpl::_getMyRID_inlock() const {
return _myRID;
}
int ReplicationCoordinatorImpl::getMyId() const {
stdx::lock_guard<stdx::mutex> lock(_mutex);
return _getMyId_inlock();
}
int ReplicationCoordinatorImpl::_getMyId_inlock() const {
const MemberConfig& self = _rsConfig.getMemberAt(_selfIndex);
return self.getId();
}
Status ReplicationCoordinatorImpl::resyncData(OperationContext* opCtx, bool waitUntilCompleted) {
_stopDataReplication(opCtx);
auto finishedEvent = uassertStatusOK(_replExecutor.makeEvent());
stdx::function<void()> f;
if (waitUntilCompleted)
f = [&finishedEvent, this]() { _replExecutor.signalEvent(finishedEvent); };
stdx::unique_lock<stdx::mutex> lk(_mutex);
_resetMyLastOpTimes_inlock();
lk.unlock(); // unlock before calling into replCoordExtState.
_startDataReplication(opCtx, f);
if (waitUntilCompleted) {
_replExecutor.waitForEvent(finishedEvent);
}
return Status::OK();
}
StatusWith<BSONObj> ReplicationCoordinatorImpl::prepareReplSetUpdatePositionCommand(
ReplicationCoordinator::ReplSetUpdatePositionCommandStyle commandStyle) const {
BSONObjBuilder cmdBuilder;
{
stdx::lock_guard<stdx::mutex> lock(_mutex);
invariant(_rsConfig.isInitialized());
// Do not send updates if we have been removed from the config.
if (_selfIndex == -1) {
return Status(ErrorCodes::NodeNotFound,
"This node is not in the current replset configuration.");
}
cmdBuilder.append(UpdatePositionArgs::kCommandFieldName, 1);
// Create an array containing objects each live member connected to us and for ourself.
BSONArrayBuilder arrayBuilder(cmdBuilder.subarrayStart("optimes"));
for (const auto& slaveInfo : _slaveInfo) {
if (slaveInfo.lastAppliedOpTime.isNull()) {
// Don't include info on members we haven't heard from yet.
continue;
}
// Don't include members we think are down.
if (!slaveInfo.self && slaveInfo.down) {
continue;
}
BSONObjBuilder entry(arrayBuilder.subobjStart());
switch (commandStyle) {
case ReplSetUpdatePositionCommandStyle::kNewStyle:
slaveInfo.lastDurableOpTime.append(&entry,
UpdatePositionArgs::kDurableOpTimeFieldName);
slaveInfo.lastAppliedOpTime.append(&entry,
UpdatePositionArgs::kAppliedOpTimeFieldName);
break;
case ReplSetUpdatePositionCommandStyle::kOldStyle:
entry.append("_id", slaveInfo.rid);
if (isV1ElectionProtocol()) {
slaveInfo.lastDurableOpTime.append(&entry, "optime");
} else {
entry.append("optime", slaveInfo.lastDurableOpTime.getTimestamp());
}
break;
}
entry.append(UpdatePositionArgs::kMemberIdFieldName, slaveInfo.memberId);
entry.append(UpdatePositionArgs::kConfigVersionFieldName, _rsConfig.getConfigVersion());
}
arrayBuilder.done();
}
// Add metadata to command. Old style parsing logic will reject the metadata.
if (commandStyle == ReplSetUpdatePositionCommandStyle::kNewStyle) {
stdx::lock_guard<stdx::mutex> topoLock(_topoMutex);
stdx::lock_guard<stdx::mutex> lock(_mutex);
_prepareReplSetMetadata_inlock(OpTime(), &cmdBuilder);
}
return cmdBuilder.obj();
}
Status ReplicationCoordinatorImpl::processReplSetGetStatus(
BSONObjBuilder* response, ReplSetGetStatusResponseStyle responseStyle) {
BSONObj initialSyncProgress;
if (responseStyle == ReplSetGetStatusResponseStyle::kInitialSync) {
LockGuard lk(_mutex);
if (_dr) {
initialSyncProgress = _dr->getInitialSyncProgress();
}
}
LockGuard topoLock(_topoMutex);
Status result(ErrorCodes::InternalError, "didn't set status in prepareStatusResponse");
_topCoord->prepareStatusResponse(
TopologyCoordinator::ReplSetStatusArgs{
_replExecutor.now(),
static_cast<unsigned>(time(0) - serverGlobalParams.started),
getMyLastAppliedOpTime(),
getMyLastDurableOpTime(),
getLastCommittedOpTime(),
getCurrentCommittedSnapshotOpTime(),
initialSyncProgress},
response,
&result);
return result;
}
void ReplicationCoordinatorImpl::fillIsMasterForReplSet(IsMasterResponse* response) {
invariant(getSettings().usingReplSets());
{
LockGuard topoLock(_topoMutex);
_topCoord->fillIsMasterForReplSet(response);
}
LockGuard lock(_mutex);
OpTime lastOpTime = _getMyLastAppliedOpTime_inlock();
response->setLastWrite(lastOpTime, lastOpTime.getTimestamp().getSecs());
if (_currentCommittedSnapshot) {
OpTime majorityOpTime = _currentCommittedSnapshot->opTime;
response->setLastMajorityWrite(majorityOpTime, majorityOpTime.getTimestamp().getSecs());
}
// Report that we are secondary to ismaster callers until drain completes.
if (response->isMaster() && !_canAcceptNonLocalWrites) {
response->setIsMaster(false);
response->setIsSecondary(true);
}
}
void ReplicationCoordinatorImpl::appendSlaveInfoData(BSONObjBuilder* result) {
stdx::lock_guard<stdx::mutex> lock(_mutex);
_appendSlaveInfoData_inlock(result);
}
void ReplicationCoordinatorImpl::_appendSlaveInfoData_inlock(BSONObjBuilder* result) {
BSONArrayBuilder replicationProgress(result->subarrayStart("replicationProgress"));
{
for (SlaveInfoVector::const_iterator itr = _slaveInfo.begin(); itr != _slaveInfo.end();
++itr) {
BSONObjBuilder entry(replicationProgress.subobjStart());
entry.append("rid", itr->rid);
if (isV1ElectionProtocol()) {
BSONObjBuilder opTime(entry.subobjStart("optime"));
opTime.append("ts", itr->lastDurableOpTime.getTimestamp());
opTime.append("term", itr->lastDurableOpTime.getTerm());
opTime.done();
} else {
entry.append("optime", itr->lastDurableOpTime.getTimestamp());
}
entry.append("host", itr->hostAndPort.toString());
if (getReplicationMode() == modeReplSet) {
if (_selfIndex == -1) {
continue;
}
invariant(itr->memberId >= 0);
entry.append("memberId", itr->memberId);
}
}
}
}
ReplSetConfig ReplicationCoordinatorImpl::getConfig() const {
stdx::lock_guard<stdx::mutex> lock(_mutex);
return _rsConfig;
}
void ReplicationCoordinatorImpl::processReplSetGetConfig(BSONObjBuilder* result) {
stdx::lock_guard<stdx::mutex> lock(_mutex);
result->append("config", _rsConfig.toBSON());
}
void ReplicationCoordinatorImpl::processReplSetMetadata(const rpc::ReplSetMetadata& replMetadata) {
EventHandle evh;
{
LockGuard topoLock(_topoMutex);
evh = _processReplSetMetadata_incallback(replMetadata);
}
if (evh) {
_replExecutor.waitForEvent(evh);
}
}
void ReplicationCoordinatorImpl::cancelAndRescheduleElectionTimeout() {
stdx::lock_guard<stdx::mutex> lock(_mutex);
_cancelAndRescheduleElectionTimeout_inlock();
}
EventHandle ReplicationCoordinatorImpl::_processReplSetMetadata_incallback(
const rpc::ReplSetMetadata& replMetadata) {
if (replMetadata.getConfigVersion() != _rsConfig.getConfigVersion()) {
return EventHandle();
}
return _updateTerm_incallback(replMetadata.getTerm());
}
bool ReplicationCoordinatorImpl::getMaintenanceMode() {
LockGuard topoLock(_topoMutex);
return _topCoord->getMaintenanceCount() > 0;
}
Status ReplicationCoordinatorImpl::setMaintenanceMode(bool activate) {
if (getReplicationMode() != modeReplSet) {
return Status(ErrorCodes::NoReplicationEnabled,
"can only set maintenance mode on replica set members");
}
LockGuard topoLock(_topoMutex);
if (_topCoord->getRole() == TopologyCoordinator::Role::candidate) {
return Status(ErrorCodes::NotSecondary, "currently running for election");
}
stdx::unique_lock<stdx::mutex> lk(_mutex);
if (_getMemberState_inlock().primary()) {
return Status(ErrorCodes::NotSecondary, "primaries can't modify maintenance mode");
}
int curMaintenanceCalls = _topCoord->getMaintenanceCount();
if (activate) {
log() << "going into maintenance mode with " << curMaintenanceCalls
<< " other maintenance mode tasks in progress" << rsLog;
_topCoord->adjustMaintenanceCountBy(1);
} else if (curMaintenanceCalls > 0) {
invariant(_topCoord->getRole() == TopologyCoordinator::Role::follower);
_topCoord->adjustMaintenanceCountBy(-1);
log() << "leaving maintenance mode (" << curMaintenanceCalls - 1
<< " other maintenance mode tasks ongoing)" << rsLog;
} else {
warning() << "Attempted to leave maintenance mode but it is not currently active";
return Status(ErrorCodes::OperationFailed, "already out of maintenance mode");
}
const PostMemberStateUpdateAction action = _updateMemberStateFromTopologyCoordinator_inlock();
lk.unlock();
_performPostMemberStateUpdateAction(action);
return Status::OK();
}
Status ReplicationCoordinatorImpl::processReplSetSyncFrom(OperationContext* opCtx,
const HostAndPort& target,
BSONObjBuilder* resultObj) {
Status result(ErrorCodes::InternalError, "didn't set status in prepareSyncFromResponse");
auto doResync = false;
{
LockGuard topoLock(_topoMutex);
LockGuard lk(_mutex);
auto opTime = _getMyLastAppliedOpTime_inlock();
_topCoord->prepareSyncFromResponse(target, opTime, resultObj, &result);
// If we are in the middle of an initial sync, do a resync.
doResync = result.isOK() && _dr && _dr->isActive();
}
if (doResync) {
return resyncData(opCtx, false);
}
return result;
}
Status ReplicationCoordinatorImpl::processReplSetFreeze(int secs, BSONObjBuilder* resultObj) {
LockGuard topoLock(_topoMutex);
auto result = _topCoord->prepareFreezeResponse(_replExecutor.now(), secs, resultObj);
if (!result.isOK()) {
return result.getStatus();
}
if (TopologyCoordinator::PrepareFreezeResponseResult::kElectSelf == result.getValue()) {
// If we just unfroze and ended our stepdown period and we are a one node replica set,
// the topology coordinator will have gone into the candidate role to signal that we
// need to elect ourself.
_performPostMemberStateUpdateAction(kActionWinElection);
}
return Status::OK();
}
Status ReplicationCoordinatorImpl::processHeartbeat(const ReplSetHeartbeatArgs& args,
ReplSetHeartbeatResponse* response) {
{
stdx::lock_guard<stdx::mutex> lock(_mutex);
if (_rsConfigState == kConfigPreStart || _rsConfigState == kConfigStartingUp) {
return Status(ErrorCodes::NotYetInitialized,
"Received heartbeat while still initializing replication system");
}
}
auto senderHost(args.getSenderHost());
LockGuard topoLock(_topoMutex);
const Date_t now = _replExecutor.now();
Status result = _topCoord->prepareHeartbeatResponse(now,
args,
_settings.ourSetName(),
getMyLastAppliedOpTime(),
getMyLastDurableOpTime(),
response);
if ((result.isOK() || result == ErrorCodes::InvalidReplicaSetConfig) && _selfIndex < 0) {
// If this node does not belong to the configuration it knows about, send heartbeats
// back to any node that sends us a heartbeat, in case one of those remote nodes has
// a configuration that contains us. Chances are excellent that it will, since that
// is the only reason for a remote node to send this node a heartbeat request.
if (!senderHost.empty() && _seedList.insert(senderHost).second) {
_scheduleHeartbeatToTarget(senderHost, -1, now);
}
} else if (result.isOK() && response->getConfigVersion() < args.getConfigVersion()) {
// Schedule a heartbeat to the sender to fetch the new config.
// We cannot cancel the enqueued heartbeat, but either this one or the enqueued heartbeat
// will trigger reconfig, which cancels and reschedules all heartbeats.
if (args.hasSenderHost()) {
int senderIndex = _rsConfig.findMemberIndexByHostAndPort(senderHost);
_scheduleHeartbeatToTarget(senderHost, senderIndex, now);
}
}
return result;
}
Status ReplicationCoordinatorImpl::processReplSetReconfig(OperationContext* opCtx,
const ReplSetReconfigArgs& args,
BSONObjBuilder* resultObj) {
log() << "replSetReconfig admin command received from client";
stdx::unique_lock<stdx::mutex> lk(_mutex);
while (_rsConfigState == kConfigPreStart || _rsConfigState == kConfigStartingUp) {
_rsConfigStateChange.wait(lk);
}
switch (_rsConfigState) {
case kConfigSteady:
break;
case kConfigUninitialized:
return Status(ErrorCodes::NotYetInitialized,
"Node not yet initialized; use the replSetInitiate command");
case kConfigReplicationDisabled:
invariant(
false); // should be unreachable due to !_settings.usingReplSets() check above
case kConfigInitiating:
case kConfigReconfiguring:
case kConfigHBReconfiguring:
return Status(ErrorCodes::ConfigurationInProgress,
"Cannot run replSetReconfig because the node is currently updating "
"its configuration");
default:
severe() << "Unexpected _rsConfigState " << int(_rsConfigState);
fassertFailed(18914);
}
invariant(_rsConfig.isInitialized());
if (!args.force && !_getMemberState_inlock().primary()) {
return Status(ErrorCodes::NotMaster,
str::stream()
<< "replSetReconfig should only be run on PRIMARY, but my state is "
<< _getMemberState_inlock().toString()
<< "; use the \"force\" argument to override");
}
_setConfigState_inlock(kConfigReconfiguring);
ScopeGuard configStateGuard = MakeGuard(
lockAndCall,
&lk,
stdx::bind(&ReplicationCoordinatorImpl::_setConfigState_inlock, this, kConfigSteady));
ReplSetConfig oldConfig = _rsConfig;
lk.unlock();
ReplSetConfig newConfig;
BSONObj newConfigObj = args.newConfigObj;
if (args.force) {
newConfigObj = incrementConfigVersionByRandom(newConfigObj);
}
Status status = newConfig.initialize(
newConfigObj, oldConfig.getProtocolVersion() == 1, oldConfig.getReplicaSetId());
if (!status.isOK()) {
error() << "replSetReconfig got " << status << " while parsing " << newConfigObj;
return Status(ErrorCodes::InvalidReplicaSetConfig, status.reason());
;
}
if (newConfig.getReplSetName() != _settings.ourSetName()) {
str::stream errmsg;
errmsg << "Attempting to reconfigure a replica set with name " << newConfig.getReplSetName()
<< ", but command line reports " << _settings.ourSetName() << "; rejecting";
error() << std::string(errmsg);
return Status(ErrorCodes::InvalidReplicaSetConfig, errmsg);
}
StatusWith<int> myIndex = validateConfigForReconfig(
_externalState.get(), oldConfig, newConfig, opCtx->getServiceContext(), args.force);
if (!myIndex.isOK()) {
error() << "replSetReconfig got " << myIndex.getStatus() << " while validating "
<< newConfigObj;
return Status(ErrorCodes::NewReplicaSetConfigurationIncompatible,
myIndex.getStatus().reason());
}
log() << "replSetReconfig config object with " << newConfig.getNumMembers()
<< " members parses ok";
if (!args.force) {
status = checkQuorumForReconfig(&_replExecutor, newConfig, myIndex.getValue());
if (!status.isOK()) {
error() << "replSetReconfig failed; " << status;
return status;
}
}
status = _externalState->storeLocalConfigDocument(opCtx, newConfig.toBSON());
if (!status.isOK()) {
error() << "replSetReconfig failed to store config document; " << status;
return status;
}
const stdx::function<void(const ReplicationExecutor::CallbackArgs&)> reconfigFinishFn(
stdx::bind(&ReplicationCoordinatorImpl::_finishReplSetReconfig,
this,
stdx::placeholders::_1,
newConfig,
myIndex.getValue()));
// If it's a force reconfig, the primary node may not be electable after the configuration
// change. In case we are that primary node, finish the reconfig under the global lock,
// so that the step down occurs safely.
CBHStatus cbhStatus(ErrorCodes::InternalError, "reconfigFinishFn hasn't been scheduled");
if (args.force) {
cbhStatus = _replExecutor.scheduleWorkWithGlobalExclusiveLock(reconfigFinishFn);
} else {
cbhStatus = _replExecutor.scheduleWork(reconfigFinishFn);
}
if (cbhStatus.getStatus() == ErrorCodes::ShutdownInProgress) {
return cbhStatus.getStatus();
}
fassert(18824, cbhStatus.getStatus());
configStateGuard.Dismiss();
_replExecutor.wait(cbhStatus.getValue());
return Status::OK();
}
void ReplicationCoordinatorImpl::_finishReplSetReconfig(
const ReplicationExecutor::CallbackArgs& cbData, const ReplSetConfig& newConfig, int myIndex) {
LockGuard topoLock(_topoMutex);
stdx::unique_lock<stdx::mutex> lk(_mutex);
invariant(_rsConfigState == kConfigReconfiguring);
invariant(_rsConfig.isInitialized());
// Do not conduct an election during a reconfig, as the node may not be electable post-reconfig.
if (_topCoord->getRole() == TopologyCoordinator::Role::candidate) {
if (isV1ElectionProtocol()) {
invariant(_voteRequester);
_voteRequester->cancel();
} else {
invariant(_freshnessChecker);
_freshnessChecker->cancel();
if (_electCmdRunner) {
_electCmdRunner->cancel();
}
}
// Wait for the election to complete and the node's Role to be set to follower.
_replExecutor.onEvent(_electionFinishedEvent,
stdx::bind(&ReplicationCoordinatorImpl::_finishReplSetReconfig,
this,
stdx::placeholders::_1,
newConfig,
myIndex));
return;
}
const ReplSetConfig oldConfig = _rsConfig;
const PostMemberStateUpdateAction action = _setCurrentRSConfig_inlock(newConfig, myIndex);
// On a reconfig we drop all snapshots so we don't mistakenely read from the wrong one.
// For example, if we change the meaning of the "committed" snapshot from applied -> durable.
_dropAllSnapshots_inlock();
lk.unlock();
auto evh = _resetElectionInfoOnProtocolVersionUpgrade(oldConfig, newConfig);
if (evh) {
_replExecutor.onEvent(evh, [this, action](const CallbackArgs& cbArgs) {
LockGuard topoLock(_topoMutex);
_performPostMemberStateUpdateAction(action);
});
} else {
_performPostMemberStateUpdateAction(action);
}
}
Status ReplicationCoordinatorImpl::processReplSetInitiate(OperationContext* opCtx,
const BSONObj& configObj,
BSONObjBuilder* resultObj) {
log() << "replSetInitiate admin command received from client";
const auto replEnabled = _settings.usingReplSets();
stdx::unique_lock<stdx::mutex> lk(_mutex);
if (!replEnabled) {
return Status(ErrorCodes::NoReplicationEnabled, "server is not running with --replSet");
}
while (_rsConfigState == kConfigPreStart || _rsConfigState == kConfigStartingUp) {
_rsConfigStateChange.wait(lk);
}
if (_rsConfigState != kConfigUninitialized) {
resultObj->append("info", "try querying local.system.replset to see current configuration");
return Status(ErrorCodes::AlreadyInitialized, "already initialized");
}
invariant(!_rsConfig.isInitialized());
_setConfigState_inlock(kConfigInitiating);
ScopeGuard configStateGuard = MakeGuard(
lockAndCall,
&lk,
stdx::bind(
&ReplicationCoordinatorImpl::_setConfigState_inlock, this, kConfigUninitialized));
lk.unlock();
ReplSetConfig newConfig;
Status status = newConfig.initializeForInitiate(configObj, true);
if (!status.isOK()) {
error() << "replSet initiate got " << status << " while parsing " << configObj;
return Status(ErrorCodes::InvalidReplicaSetConfig, status.reason());
}
if (newConfig.getReplSetName() != _settings.ourSetName()) {
str::stream errmsg;
errmsg << "Attempting to initiate a replica set with name " << newConfig.getReplSetName()
<< ", but command line reports " << _settings.ourSetName() << "; rejecting";
error() << std::string(errmsg);
return Status(ErrorCodes::InvalidReplicaSetConfig, errmsg);
}
StatusWith<int> myIndex =
validateConfigForInitiate(_externalState.get(), newConfig, opCtx->getServiceContext());
if (!myIndex.isOK()) {
error() << "replSet initiate got " << myIndex.getStatus() << " while validating "
<< configObj;
return Status(ErrorCodes::InvalidReplicaSetConfig, myIndex.getStatus().reason());
}
log() << "replSetInitiate config object with " << newConfig.getNumMembers()
<< " members parses ok";
status = checkQuorumForInitiate(&_replExecutor, newConfig, myIndex.getValue());
if (!status.isOK()) {
error() << "replSetInitiate failed; " << status;
return status;
}
status = _externalState->initializeReplSetStorage(opCtx, newConfig.toBSON());
if (!status.isOK()) {
error() << "replSetInitiate failed to store config document or create the oplog; "
<< status;
return status;
}
// Since the JournalListener has not yet been set up, we must manually set our
// durableOpTime.
setMyLastDurableOpTime(getMyLastAppliedOpTime());
{
LockGuard topoLock(_topoMutex);
_finishReplSetInitiate(newConfig, myIndex.getValue());
}
// A configuration passed to replSetInitiate() with the current node as an arbiter
// will fail validation with a "replSet initiate got ... while validating" reason.
invariant(!newConfig.getMemberAt(myIndex.getValue()).isArbiter());
_externalState->startThreads(_settings);
_startDataReplication(opCtx);
configStateGuard.Dismiss();
return Status::OK();
}
void ReplicationCoordinatorImpl::_finishReplSetInitiate(const ReplSetConfig& newConfig,
int myIndex) {
stdx::unique_lock<stdx::mutex> lk(_mutex);
invariant(_rsConfigState == kConfigInitiating);
invariant(!_rsConfig.isInitialized());
auto action = _setCurrentRSConfig_inlock(newConfig, myIndex);
lk.unlock();
_performPostMemberStateUpdateAction(action);
}
void ReplicationCoordinatorImpl::_setConfigState_inlock(ConfigState newState) {
if (newState != _rsConfigState) {
_rsConfigState = newState;
_rsConfigStateChange.notify_all();
}
}
ReplicationCoordinatorImpl::PostMemberStateUpdateAction
ReplicationCoordinatorImpl::_updateMemberStateFromTopologyCoordinator_inlock() {
const MemberState newState = _topCoord->getMemberState();
if (newState == _memberState) {
if (_topCoord->getRole() == TopologyCoordinator::Role::candidate) {
invariant(_rsConfig.getNumMembers() == 1 && _selfIndex == 0 &&
_rsConfig.getMemberAt(0).isElectable());
if (isV1ElectionProtocol()) {
// Start election in protocol version 1
return kActionStartSingleNodeElection;
}
return kActionWinElection;
}
return kActionNone;
}
PostMemberStateUpdateAction result;
if (_memberState.primary() || newState.removed() || newState.rollback()) {
// Wake up any threads blocked in awaitReplication, close connections, etc.
_replicationWaiterList.signalAndRemoveAll_inlock();
// Wake up the optime waiter that is waiting for primary catch-up to finish.
_opTimeWaiterList.signalAndRemoveAll_inlock();
_canAcceptNonLocalWrites = false;
_stepDownPending = false;
serverGlobalParams.featureCompatibility.validateFeaturesAsMaster.store(false);
result = kActionCloseAllConnections;
} else {
result = kActionFollowerModeStateChange;
}
// Enable replication producer and applier on stepdown.
if (_memberState.primary()) {
_applierState = ApplierState::Running;
_externalState->startProducerIfStopped();
}
if (_memberState.secondary() && !newState.primary()) {
// Switching out of SECONDARY, but not to PRIMARY.
_canServeNonLocalReads.store(0U);
} else if (!_memberState.primary() && newState.secondary()) {
// Switching into SECONDARY, but not from PRIMARY.
_canServeNonLocalReads.store(1U);
}
if (newState.secondary() && _topCoord->getRole() == TopologyCoordinator::Role::candidate) {
// When transitioning to SECONDARY, the only way for _topCoord to report the candidate
// role is if the configuration represents a single-node replica set. In that case, the
// overriding requirement is to elect this singleton node primary.
invariant(_rsConfig.getNumMembers() == 1 && _selfIndex == 0 &&
_rsConfig.getMemberAt(0).isElectable());
if (isV1ElectionProtocol()) {
// Start election in protocol version 1
result = kActionStartSingleNodeElection;
} else {
result = kActionWinElection;
}
}
if (newState.readable() && !_memberState.readable()) {
// When we transition to a readable state from a non-readable one, force the SnapshotThread
// to take a snapshot, if it is running. This is because it never takes snapshots when not
// in readable states.
_externalState->forceSnapshotCreation();
}
if (newState.rollback()) {
// When we start rollback, we need to drop all snapshots since we may need to create
// out-of-order snapshots. This would be necessary even if the SnapshotName was completely
// monotonically increasing because we don't necessarily have a snapshot of every write.
// If we didn't drop all snapshots on rollback it could lead to the following situation:
//
// |--------|-------------|-------------|
// | OpTime | HasSnapshot | Committed |
// |--------|-------------|-------------|
// | (0, 1) | * | * |
// | (0, 2) | * | ROLLED BACK |
// | (1, 2) | | * |
// |--------|-------------|-------------|
//
// When we try to make (1,2) the commit point, we'd find (0,2) as the newest snapshot
// before the commit point, but it would be invalid to mark it as the committed snapshot
// since it was never committed.
//
// TODO SERVER-19209 We also need to clear snapshots before a resync.
_dropAllSnapshots_inlock();
}
if (_memberState.rollback()) {
// Ensure that no snapshots were created while we were in rollback.
invariant(!_currentCommittedSnapshot);
invariant(_uncommittedSnapshots.empty());
}
_memberState = newState;
log() << "transition to " << newState.toString() << rsLog;
_cancelAndRescheduleElectionTimeout_inlock();
// Notifies waiters blocked in waitForMemberState().
// For testing only.
_memberStateChange.notify_all();
return result;
}
void ReplicationCoordinatorImpl::_performPostMemberStateUpdateAction(
PostMemberStateUpdateAction action) {
switch (action) {
case kActionNone:
break;
case kActionFollowerModeStateChange:
// In follower mode, or sub-mode so ensure replication is active
_externalState->signalApplierToChooseNewSyncSource();
break;
case kActionCloseAllConnections:
_externalState->closeConnections();
_externalState->shardingOnStepDownHook();
_externalState->stopNoopWriter();
break;
case kActionWinElection: {
stdx::unique_lock<stdx::mutex> lk(_mutex);
if (isV1ElectionProtocol()) {
invariant(_topCoord->getTerm() != OpTime::kUninitializedTerm);
_electionId = OID::fromTerm(_topCoord->getTerm());
} else {
_electionId = OID::gen();
}
auto ts = LogicalClock::get(getServiceContext())->reserveTicks(1).asTimestamp();
_topCoord->processWinElection(_electionId, ts);
const PostMemberStateUpdateAction nextAction =
_updateMemberStateFromTopologyCoordinator_inlock();
invariant(nextAction != kActionWinElection);
lk.unlock();
_performPostMemberStateUpdateAction(nextAction);
// Notify all secondaries of the election win.
_scheduleElectionWinNotification();
lk.lock();
if (isV1ElectionProtocol()) {
_scanOpTimeForCatchUp_inlock();
} else {
_finishCatchingUpOplog_inlock();
}
break;
}
case kActionStartSingleNodeElection:
// In protocol version 1, single node replset will run an election instead of
// kActionWinElection as in protocol version 0.
_startElectSelfV1();
break;
default:
severe() << "Unknown post member state update action " << static_cast<int>(action);
fassertFailed(26010);
}
}
void ReplicationCoordinatorImpl::_scanOpTimeForCatchUp_inlock() {
auto scanner = std::make_shared<FreshnessScanner>();
auto scanStartTime = _replExecutor.now();
auto evhStatus =
scanner->start(&_replExecutor, _rsConfig, _selfIndex, _rsConfig.getCatchUpTimeoutPeriod());
if (evhStatus == ErrorCodes::ShutdownInProgress) {
_finishCatchingUpOplog_inlock();
return;
}
fassertStatusOK(40254, evhStatus.getStatus());
long long term = _cachedTerm;
_replExecutor.onEvent(
evhStatus.getValue(), [this, scanner, scanStartTime, term](const CallbackArgs& cbData) {
LockGuard lk(_mutex);
if (cbData.status == ErrorCodes::CallbackCanceled) {
_finishCatchingUpOplog_inlock();
return;
}
auto totalTimeout = _rsConfig.getCatchUpTimeoutPeriod();
auto catchUpTimeout = totalTimeout - (_replExecutor.now() - scanStartTime);
_catchUpOplogToLatest_inlock(*scanner, catchUpTimeout, term);
});
}
void ReplicationCoordinatorImpl::_catchUpOplogToLatest_inlock(const FreshnessScanner& scanner,
Milliseconds timeout,
long long originalTerm) {
// On stepping down, the node doesn't update its term immediately due to SERVER-21425.
// Term is also checked in case the catchup timeout is so long that the node becomes primary
// again.
if (!_memberState.primary() || originalTerm != _cachedTerm) {
// If the node steps down during the catch-up, we don't go into drain mode.
log() << "Stopped transition to primary of term " << originalTerm
<< " because I've already stepped down.";
return;
}
auto result = scanner.getResult();
// Cannot access any nodes within timeout.
if (result.size() == 0) {
log() << "Could not access any nodes within timeout when checking for "
<< "additional ops to apply before finishing transition to primary. "
<< "Will move forward with becoming primary anyway.";
_finishCatchingUpOplog_inlock();
return;
}
// I'm most up-to-date as far as I know.
auto freshnessInfo = result.front();
if (freshnessInfo.opTime <= _getMyLastAppliedOpTime_inlock()) {
log() << "My optime is most up-to-date, skipping catch-up "
<< "and completing transition to primary.";
_finishCatchingUpOplog_inlock();
return;
}
// Wait for the replication level to reach the latest opTime within timeout.
auto latestOpTime = freshnessInfo.opTime;
auto finishCB = [this, latestOpTime]() {
if (latestOpTime > _getMyLastAppliedOpTime_inlock()) {
log() << "Cannot catch up oplog after becoming primary.";
} else {
log() << "Finished catch-up oplog after becoming primary.";
}
_finishCatchingUpOplog_inlock();
};
auto waiterInfo = std::make_shared<WaiterInfo>(freshnessInfo.opTime, finishCB);
_opTimeWaiterList.add_inlock(waiterInfo.get());
auto timeoutCB = [this, waiterInfo, finishCB](const CallbackArgs& cbData) {
LockGuard lk(_mutex);
if (_opTimeWaiterList.remove_inlock(waiterInfo.get())) {
finishCB();
}
};
// Schedule the timeout callback. It may signal after we have already caught up.
_replExecutor.scheduleWorkAt(_replExecutor.now() + timeout, timeoutCB);
}
void ReplicationCoordinatorImpl::_finishCatchingUpOplog_inlock() {
_applierState = ApplierState::Draining;
_externalState->stopProducer();
}
Status ReplicationCoordinatorImpl::processReplSetGetRBID(BSONObjBuilder* resultObj) {
stdx::lock_guard<stdx::mutex> lk(_mutex);
resultObj->append("rbid", _rbid);
return Status::OK();
}
void ReplicationCoordinatorImpl::incrementRollbackID() {
stdx::lock_guard<stdx::mutex> lk(_mutex);
++_rbid;
}
Status ReplicationCoordinatorImpl::processReplSetFresh(const ReplSetFreshArgs& args,
BSONObjBuilder* resultObj) {
LockGuard topoLock(_topoMutex);
Status result(ErrorCodes::InternalError, "didn't set status in prepareFreshResponse");
_topCoord->prepareFreshResponse(
args, _replExecutor.now(), getMyLastAppliedOpTime(), resultObj, &result);
return result;
}
Status ReplicationCoordinatorImpl::processReplSetElect(const ReplSetElectArgs& args,
BSONObjBuilder* responseObj) {
LockGuard topoLock(_topoMutex);
Status result = Status(ErrorCodes::InternalError, "status not set by callback");
_topCoord->prepareElectResponse(
args, _replExecutor.now(), getMyLastAppliedOpTime(), responseObj, &result);
return result;
}
ReplicationCoordinatorImpl::PostMemberStateUpdateAction
ReplicationCoordinatorImpl::_setCurrentRSConfig_inlock(const ReplSetConfig& newConfig,
int myIndex) {
invariant(_settings.usingReplSets());
_cancelHeartbeats_inlock();
_setConfigState_inlock(kConfigSteady);
// Must get this before changing our config.
OpTime myOptime = _getMyLastAppliedOpTime_inlock();
_topCoord->updateConfig(newConfig, myIndex, _replExecutor.now(), myOptime);
_cachedTerm = _topCoord->getTerm();
const ReplSetConfig oldConfig = _rsConfig;
_rsConfig = newConfig;
_protVersion.store(_rsConfig.getProtocolVersion());
log() << "New replica set config in use: " << _rsConfig.toBSON() << rsLog;
_selfIndex = myIndex;
if (_selfIndex >= 0) {
log() << "This node is " << _rsConfig.getMemberAt(_selfIndex).getHostAndPort()
<< " in the config";
} else {
log() << "This node is not a member of the config";
}
_cancelPriorityTakeover_inlock();
_cancelAndRescheduleElectionTimeout_inlock();
const PostMemberStateUpdateAction action = _updateMemberStateFromTopologyCoordinator_inlock();
_updateSlaveInfoFromConfig_inlock();
if (_selfIndex >= 0) {
// Don't send heartbeats if we're not in the config, if we get re-added one of the
// nodes in the set will contact us.
_startHeartbeats_inlock();
}
_updateLastCommittedOpTime_inlock();
// Set election id if we're primary.
if (oldConfig.isInitialized() && _memberState.primary()) {
if (oldConfig.getProtocolVersion() > newConfig.getProtocolVersion()) {
// Downgrade
invariant(newConfig.getProtocolVersion() == 0);
_electionId = OID::gen();
auto ts = LogicalClock::get(getServiceContext())->reserveTicks(1).asTimestamp();
_topCoord->setElectionInfo(_electionId, ts);
} else if (oldConfig.getProtocolVersion() < newConfig.getProtocolVersion()) {
// Upgrade
invariant(newConfig.getProtocolVersion() == 1);
invariant(_topCoord->getTerm() != OpTime::kUninitializedTerm);
_electionId = OID::fromTerm(_topCoord->getTerm());
auto ts = LogicalClock::get(getServiceContext())->reserveTicks(1).asTimestamp();
_topCoord->setElectionInfo(_electionId, ts);
}
}
_wakeReadyWaiters_inlock();
return action;
}
void ReplicationCoordinatorImpl::_wakeReadyWaiters_inlock() {
_replicationWaiterList.signalAndRemoveIf_inlock([this](WaiterInfo* waiter) {
return _doneWaitingForReplication_inlock(
waiter->opTime, SnapshotName::min(), *waiter->writeConcern);
});
}
Status ReplicationCoordinatorImpl::processReplSetUpdatePosition(
const OldUpdatePositionArgs& updates, long long* configVersion) {
stdx::unique_lock<stdx::mutex> lock(_mutex);
Status status = Status::OK();
bool somethingChanged = false;
for (OldUpdatePositionArgs::UpdateIterator update = updates.updatesBegin();
update != updates.updatesEnd();
++update) {
status = _setLastOptime_inlock(*update, configVersion);
if (!status.isOK()) {
break;
}
somethingChanged = true;
}
if (somethingChanged && !_getMemberState_inlock().primary()) {
lock.unlock();
// Must do this outside _mutex
_externalState->forwardSlaveProgress();
}
return status;
}
Status ReplicationCoordinatorImpl::processReplSetUpdatePosition(const UpdatePositionArgs& updates,
long long* configVersion) {
stdx::unique_lock<stdx::mutex> lock(_mutex);
Status status = Status::OK();
bool somethingChanged = false;
for (UpdatePositionArgs::UpdateIterator update = updates.updatesBegin();
update != updates.updatesEnd();
++update) {
status = _setLastOptime_inlock(*update, configVersion);
if (!status.isOK()) {
break;
}
somethingChanged = true;
}
if (somethingChanged && !_getMemberState_inlock().primary()) {
lock.unlock();
// Must do this outside _mutex
_externalState->forwardSlaveProgress();
}
return status;
}
Status ReplicationCoordinatorImpl::processHandshake(OperationContext* opCtx,
const HandshakeArgs& handshake) {
LOG(2) << "Received handshake " << handshake.toBSON();
stdx::unique_lock<stdx::mutex> lock(_mutex);
if (getReplicationMode() != modeMasterSlave) {
return Status(ErrorCodes::IllegalOperation,
"The handshake command is only used for master/slave replication");
}
SlaveInfo* slaveInfo = _findSlaveInfoByRID_inlock(handshake.getRid());
if (slaveInfo) {
return Status::OK(); // nothing to do
}
SlaveInfo newSlaveInfo;
newSlaveInfo.rid = handshake.getRid();
newSlaveInfo.memberId = -1;
newSlaveInfo.hostAndPort = _externalState->getClientHostAndPort(opCtx);
// Don't call _addSlaveInfo_inlock as that would wake sleepers unnecessarily.
_slaveInfo.push_back(newSlaveInfo);
return Status::OK();
}
bool ReplicationCoordinatorImpl::buildsIndexes() {
stdx::lock_guard<stdx::mutex> lk(_mutex);
if (_selfIndex == -1) {
return true;
}
const MemberConfig& self = _rsConfig.getMemberAt(_selfIndex);
return self.shouldBuildIndexes();
}
std::vector<HostAndPort> ReplicationCoordinatorImpl::getHostsWrittenTo(const OpTime& op,
bool durablyWritten) {
std::vector<HostAndPort> hosts;
stdx::lock_guard<stdx::mutex> lk(_mutex);
for (size_t i = 0; i < _slaveInfo.size(); ++i) {
const SlaveInfo& slaveInfo = _slaveInfo[i];
if (getReplicationMode() == modeMasterSlave && slaveInfo.rid == _getMyRID_inlock()) {
// Master-slave doesn't know the HostAndPort for itself at this point.
continue;
}
if (durablyWritten) {
if (slaveInfo.lastDurableOpTime < op) {
continue;
}
} else if (slaveInfo.lastAppliedOpTime < op) {
continue;
}
hosts.push_back(slaveInfo.hostAndPort);
}
return hosts;
}
std::vector<HostAndPort> ReplicationCoordinatorImpl::getOtherNodesInReplSet() const {
stdx::lock_guard<stdx::mutex> lk(_mutex);
invariant(_settings.usingReplSets());
std::vector<HostAndPort> nodes;
if (_selfIndex == -1) {
return nodes;
}
for (int i = 0; i < _rsConfig.getNumMembers(); ++i) {
if (i == _selfIndex)
continue;
nodes.push_back(_rsConfig.getMemberAt(i).getHostAndPort());
}
return nodes;
}
Status ReplicationCoordinatorImpl::checkIfWriteConcernCanBeSatisfied(
const WriteConcernOptions& writeConcern) const {
stdx::lock_guard<stdx::mutex> lock(_mutex);
return _checkIfWriteConcernCanBeSatisfied_inlock(writeConcern);
}
Status ReplicationCoordinatorImpl::_checkIfWriteConcernCanBeSatisfied_inlock(
const WriteConcernOptions& writeConcern) const {
if (getReplicationMode() == modeNone) {
return Status(ErrorCodes::NoReplicationEnabled,
"No replication enabled when checking if write concern can be satisfied");
}
if (getReplicationMode() == modeMasterSlave) {
if (!writeConcern.wMode.empty()) {
return Status(ErrorCodes::UnknownReplWriteConcern,
"Cannot use named write concern modes in master-slave");
}
// No way to know how many slaves there are, so assume any numeric mode is possible.
return Status::OK();
}
invariant(getReplicationMode() == modeReplSet);
return _rsConfig.checkIfWriteConcernCanBeSatisfied(writeConcern);
}
WriteConcernOptions ReplicationCoordinatorImpl::getGetLastErrorDefault() {
stdx::lock_guard<stdx::mutex> lock(_mutex);
if (_rsConfig.isInitialized()) {
return _rsConfig.getDefaultWriteConcern();
}
return WriteConcernOptions();
}
Status ReplicationCoordinatorImpl::checkReplEnabledForCommand(BSONObjBuilder* result) {
if (!_settings.usingReplSets()) {
if (serverGlobalParams.clusterRole == ClusterRole::ConfigServer) {
result->append("info", "configsvr"); // for shell prompt
}
return Status(ErrorCodes::NoReplicationEnabled, "not running with --replSet");
}
if (getMemberState().startup()) {
result->append("info", "run rs.initiate(...) if not yet done for the set");
return Status(ErrorCodes::NotYetInitialized, "no replset config has been received");
}
return Status::OK();
}
bool ReplicationCoordinatorImpl::isReplEnabled() const {
return getReplicationMode() != modeNone;
}
HostAndPort ReplicationCoordinatorImpl::chooseNewSyncSource(const OpTime& lastOpTimeFetched) {
LockGuard topoLock(_topoMutex);
HostAndPort oldSyncSource = _topCoord->getSyncSourceAddress();
// Always allow chaining while in catchup and drain mode.
auto chainingPreference = getMemberState().primary()
? TopologyCoordinator::ChainingPreference::kAllowChaining
: TopologyCoordinator::ChainingPreference::kUseConfiguration;
HostAndPort newSyncSource =
_topCoord->chooseNewSyncSource(_replExecutor.now(), lastOpTimeFetched, chainingPreference);
stdx::lock_guard<stdx::mutex> lock(_mutex);
// If we lost our sync source, schedule new heartbeats immediately to update our knowledge
// of other members's state, allowing us to make informed sync source decisions.
if (newSyncSource.empty() && !oldSyncSource.empty() && _selfIndex >= 0 &&
!_getMemberState_inlock().primary()) {
_restartHeartbeats_inlock();
}
return newSyncSource;
}
void ReplicationCoordinatorImpl::_unblacklistSyncSource(
const ReplicationExecutor::CallbackArgs& cbData, const HostAndPort& host) {
if (cbData.status == ErrorCodes::CallbackCanceled)
return;
LockGuard topoLock(_topoMutex);
_topCoord->unblacklistSyncSource(host, _replExecutor.now());
}
void ReplicationCoordinatorImpl::blacklistSyncSource(const HostAndPort& host, Date_t until) {
LockGuard topoLock(_topoMutex);
_topCoord->blacklistSyncSource(host, until);
_scheduleWorkAt(until,
stdx::bind(&ReplicationCoordinatorImpl::_unblacklistSyncSource,
this,
stdx::placeholders::_1,
host));
}
void ReplicationCoordinatorImpl::resetLastOpTimesFromOplog(OperationContext* opCtx) {
StatusWith<OpTime> lastOpTimeStatus = _externalState->loadLastOpTime(opCtx);
OpTime lastOpTime;
if (!lastOpTimeStatus.isOK()) {
warning() << "Failed to load timestamp of most recently applied operation; "
<< lastOpTimeStatus.getStatus();
} else {
lastOpTime = lastOpTimeStatus.getValue();
}
stdx::unique_lock<stdx::mutex> lock(_mutex);
_setMyLastAppliedOpTime_inlock(lastOpTime, true);
_setMyLastDurableOpTime_inlock(lastOpTime, true);
_reportUpstream_inlock(std::move(lock));
// Unlocked below.
_externalState->setGlobalTimestamp(opCtx->getServiceContext(), lastOpTime.getTimestamp());
}
bool ReplicationCoordinatorImpl::shouldChangeSyncSource(
const HostAndPort& currentSource,
const rpc::ReplSetMetadata& replMetadata,
boost::optional<rpc::OplogQueryMetadata> oqMetadata) {
LockGuard topoLock(_topoMutex);
return _topCoord->shouldChangeSyncSource(
currentSource, getMyLastAppliedOpTime(), replMetadata, oqMetadata, _replExecutor.now());
}
void ReplicationCoordinatorImpl::_updateLastCommittedOpTime_inlock() {
if (!_getMemberState_inlock().primary() || _stepDownPending) {
return;
}
std::vector<OpTime> votingNodesOpTimes;
// Whether we use the applied or durable OpTime for the commit point is decided here.
const bool useDurableOpTime = getWriteConcernMajorityShouldJournal_inlock();
for (const auto& sI : _slaveInfo) {
auto memberConfig = _rsConfig.findMemberByID(sI.memberId);
invariant(memberConfig);
if (memberConfig->isVoter()) {
const auto opTime = useDurableOpTime ? sI.lastDurableOpTime : sI.lastAppliedOpTime;
votingNodesOpTimes.push_back(opTime);
}
}
invariant(votingNodesOpTimes.size() > 0);
if (votingNodesOpTimes.size() < static_cast<unsigned long>(_rsConfig.getWriteMajority())) {
return;
}
std::sort(votingNodesOpTimes.begin(), votingNodesOpTimes.end());
// need the majority to have this OpTime
OpTime committedOpTime =
votingNodesOpTimes[votingNodesOpTimes.size() - _rsConfig.getWriteMajority()];
_advanceCommitPoint_inlock(committedOpTime);
}
void ReplicationCoordinatorImpl::advanceCommitPoint(const OpTime& committedOpTime) {
stdx::unique_lock<stdx::mutex> lk(_mutex);
_advanceCommitPoint_inlock(committedOpTime);
}
void ReplicationCoordinatorImpl::_advanceCommitPoint_inlock(const OpTime& committedOpTime) {
if (committedOpTime == _lastCommittedOpTime) {
return; // Hasn't changed, so ignore it.
} else if (committedOpTime < _lastCommittedOpTime) {
LOG(1) << "Ignoring older committed snapshot optime: " << committedOpTime
<< ", currentCommittedOpTime: " << _lastCommittedOpTime;
return; // This may have come from an out-of-order heartbeat. Ignore it.
}
// This check is performed to ensure primaries do not commit an OpTime from a previous term.
if (_getMemberState_inlock().primary() && committedOpTime < _firstOpTimeOfMyTerm) {
LOG(1) << "Ignoring older committed snapshot from before I became primary, optime: "
<< committedOpTime << ", firstOpTimeOfMyTerm: " << _firstOpTimeOfMyTerm;
return;
}
if (_getMemberState_inlock().arbiter()) {
_setMyLastAppliedOpTime_inlock(committedOpTime, false);
}
LOG(2) << "Updating _lastCommittedOpTime to " << committedOpTime;
_lastCommittedOpTime = committedOpTime;
_externalState->notifyOplogMetadataWaiters();
auto maxSnapshotForOpTime = SnapshotInfo{committedOpTime, SnapshotName::max()};
if (!_uncommittedSnapshots.empty() && _uncommittedSnapshots.front() <= maxSnapshotForOpTime) {
// At least one uncommitted snapshot is ready to be blessed as committed.
// Seek to the first entry > the commit point. Previous element must be <=.
const auto onePastCommitPoint = std::upper_bound(
_uncommittedSnapshots.begin(), _uncommittedSnapshots.end(), maxSnapshotForOpTime);
const auto newSnapshot = *std::prev(onePastCommitPoint);
// Forget about all snapshots <= the new commit point.
_uncommittedSnapshots.erase(_uncommittedSnapshots.begin(), onePastCommitPoint);
_uncommittedSnapshotsSize.store(_uncommittedSnapshots.size());
// Update committed snapshot and wake up any threads waiting on read concern or
// write concern.
//
// This function is only called on secondaries, so only threads waiting for
// committed snapshot need to be woken up.
_updateCommittedSnapshot_inlock(newSnapshot);
}
}
void ReplicationCoordinatorImpl::_setFirstOpTimeOfMyTerm(const OpTime& newOpTime) {
stdx::unique_lock<stdx::mutex> lk(_mutex);
_firstOpTimeOfMyTerm = newOpTime;
}
OpTime ReplicationCoordinatorImpl::getLastCommittedOpTime() const {
stdx::unique_lock<stdx::mutex> lk(_mutex);
return _lastCommittedOpTime;
}
Status ReplicationCoordinatorImpl::processReplSetRequestVotes(
OperationContext* opCtx,
const ReplSetRequestVotesArgs& args,
ReplSetRequestVotesResponse* response) {
if (!isV1ElectionProtocol()) {
return {ErrorCodes::BadValue, "not using election protocol v1"};
}
auto termStatus = updateTerm(opCtx, args.getTerm());
if (!termStatus.isOK() && termStatus.code() != ErrorCodes::StaleTerm)
return termStatus;
{
LockGuard topoLock(_topoMutex);
LockGuard lk(_mutex);
_topCoord->processReplSetRequestVotes(args, response, _getMyLastAppliedOpTime_inlock());
}
if (!args.isADryRun() && response->getVoteGranted()) {
LastVote lastVote{args.getTerm(), args.getCandidateIndex()};
Status status = _externalState->storeLocalLastVoteDocument(opCtx, lastVote);
if (!status.isOK()) {
error() << "replSetRequestVotes failed to store LastVote document; " << status;
return status;
}
}
return Status::OK();
}
void ReplicationCoordinatorImpl::prepareReplMetadata(const BSONObj& metadataRequestObj,
const OpTime& lastOpTimeFromClient,
BSONObjBuilder* builder) const {
bool hasReplSetMetadata = metadataRequestObj.hasField(rpc::kReplSetMetadataFieldName);
bool hasOplogQueryMetadata = metadataRequestObj.hasField(rpc::kOplogQueryMetadataFieldName);
// Don't take any locks if we do not need to.
if (!hasReplSetMetadata && !hasOplogQueryMetadata) {
return;
}
LockGuard topoLock(_topoMutex);
LockGuard lock(_mutex);
if (hasReplSetMetadata) {
_prepareReplSetMetadata_inlock(lastOpTimeFromClient, builder);
}
if (hasOplogQueryMetadata) {
_prepareOplogQueryMetadata_inlock(builder);
}
}
void ReplicationCoordinatorImpl::_prepareReplSetMetadata_inlock(const OpTime& lastOpTimeFromClient,
BSONObjBuilder* builder) const {
OpTime lastVisibleOpTime =
std::max(lastOpTimeFromClient, _getCurrentCommittedSnapshotOpTime_inlock());
auto metadata = _topCoord->prepareReplSetMetadata(lastVisibleOpTime, _lastCommittedOpTime);
metadata.writeToMetadata(builder);
}
void ReplicationCoordinatorImpl::_prepareOplogQueryMetadata_inlock(BSONObjBuilder* builder) const {
OpTime lastAppliedOpTime = _getMyLastAppliedOpTime_inlock();
auto metadata =
_topCoord->prepareOplogQueryMetadata(_lastCommittedOpTime, lastAppliedOpTime, _rbid);
metadata.writeToMetadata(builder);
}
bool ReplicationCoordinatorImpl::isV1ElectionProtocol() const {
return _protVersion.load() == 1;
}
bool ReplicationCoordinatorImpl::getWriteConcernMajorityShouldJournal() {
return getConfig().getWriteConcernMajorityShouldJournal();
}
bool ReplicationCoordinatorImpl::getWriteConcernMajorityShouldJournal_inlock() const {
return _rsConfig.getWriteConcernMajorityShouldJournal();
}
Status ReplicationCoordinatorImpl::processHeartbeatV1(const ReplSetHeartbeatArgsV1& args,
ReplSetHeartbeatResponse* response) {
{
stdx::lock_guard<stdx::mutex> lock(_mutex);
if (_rsConfigState == kConfigPreStart || _rsConfigState == kConfigStartingUp) {
return Status(ErrorCodes::NotYetInitialized,
"Received heartbeat while still initializing replication system");
}
}
Status result(ErrorCodes::InternalError, "didn't set status in prepareHeartbeatResponse");
LockGuard topoLock(_topoMutex);
auto senderHost(args.getSenderHost());
const Date_t now = _replExecutor.now();
result = _topCoord->prepareHeartbeatResponseV1(now,
args,
_settings.ourSetName(),
getMyLastAppliedOpTime(),
getMyLastDurableOpTime(),
response);
if ((result.isOK() || result == ErrorCodes::InvalidReplicaSetConfig) && _selfIndex < 0) {
// If this node does not belong to the configuration it knows about, send heartbeats
// back to any node that sends us a heartbeat, in case one of those remote nodes has
// a configuration that contains us. Chances are excellent that it will, since that
// is the only reason for a remote node to send this node a heartbeat request.
if (!senderHost.empty() && _seedList.insert(senderHost).second) {
_scheduleHeartbeatToTarget(senderHost, -1, now);
}
} else if (result.isOK() && response->getConfigVersion() < args.getConfigVersion()) {
// Schedule a heartbeat to the sender to fetch the new config.
// We cannot cancel the enqueued heartbeat, but either this one or the enqueued heartbeat
// will trigger reconfig, which cancels and reschedules all heartbeats.
if (args.hasSender()) {
int senderIndex = _rsConfig.findMemberIndexByHostAndPort(senderHost);
_scheduleHeartbeatToTarget(senderHost, senderIndex, now);
}
} else if (result.isOK()) {
// Update liveness for sending node.
stdx::lock_guard<stdx::mutex> lk(_mutex);
auto slaveInfo = _findSlaveInfoByMemberID_inlock(args.getSenderId());
if (!slaveInfo) {
return result;
}
slaveInfo->lastUpdate = _replExecutor.now();
slaveInfo->down = false;
}
return result;
}
void ReplicationCoordinatorImpl::summarizeAsHtml(ReplSetHtmlSummary* output) {
LockGuard topoLock(_topoMutex);
// TODO(dannenberg) consider putting both optimes into the htmlsummary.
output->setSelfOptime(getMyLastAppliedOpTime());
output->setSelfUptime(time(0) - serverGlobalParams.started);
output->setNow(_replExecutor.now());
_topCoord->summarizeAsHtml(output);
}
long long ReplicationCoordinatorImpl::getTerm() {
stdx::lock_guard<stdx::mutex> lock(_mutex);
return _cachedTerm;
}
EventHandle ReplicationCoordinatorImpl::updateTerm_forTest(
long long term, TopologyCoordinator::UpdateTermResult* updateResult) {
LockGuard topoLock(_topoMutex);
EventHandle finishEvh;
finishEvh = _updateTerm_incallback(term, updateResult);
if (!finishEvh) {
auto finishEvhStatus = _replExecutor.makeEvent();
invariantOK(finishEvhStatus.getStatus());
finishEvh = finishEvhStatus.getValue();
_replExecutor.signalEvent(finishEvh);
}
return finishEvh;
}
Status ReplicationCoordinatorImpl::updateTerm(OperationContext* opCtx, long long term) {
// Term is only valid if we are replicating.
if (getReplicationMode() != modeReplSet) {
return {ErrorCodes::BadValue, "cannot supply 'term' without active replication"};
}
if (!isV1ElectionProtocol()) {
// Do not update if not in V1 protocol.
return Status::OK();
}
// Check we haven't acquired any lock, because potential stepdown needs global lock.
dassert(!opCtx->lockState()->isLocked());
TopologyCoordinator::UpdateTermResult updateTermResult;
EventHandle finishEvh;
{
LockGuard topoLock(_topoMutex);
finishEvh = _updateTerm_incallback(term, &updateTermResult);
}
// Wait for potential stepdown to finish.
if (finishEvh.isValid()) {
_replExecutor.waitForEvent(finishEvh);
}
if (updateTermResult == TopologyCoordinator::UpdateTermResult::kUpdatedTerm ||
updateTermResult == TopologyCoordinator::UpdateTermResult::kTriggerStepDown) {
return {ErrorCodes::StaleTerm, "Replication term of this node was stale; retry query"};
}
return Status::OK();
}
EventHandle ReplicationCoordinatorImpl::_updateTerm_incallback(
long long term, TopologyCoordinator::UpdateTermResult* updateTermResult) {
if (!isV1ElectionProtocol()) {
LOG(3) << "Cannot update term in election protocol version 0";
return EventHandle();
}
auto now = _replExecutor.now();
TopologyCoordinator::UpdateTermResult localUpdateTermResult = _topCoord->updateTerm(term, now);
{
stdx::lock_guard<stdx::mutex> lock(_mutex);
_cachedTerm = _topCoord->getTerm();
if (localUpdateTermResult == TopologyCoordinator::UpdateTermResult::kUpdatedTerm) {
_cancelPriorityTakeover_inlock();
_cancelAndRescheduleElectionTimeout_inlock();
}
}
if (updateTermResult) {
*updateTermResult = localUpdateTermResult;
}
if (localUpdateTermResult == TopologyCoordinator::UpdateTermResult::kTriggerStepDown) {
log() << "stepping down from primary, because a new term has begun: " << term;
_topCoord->prepareForStepDown();
return _stepDownStart(false);
}
return EventHandle();
}
SnapshotName ReplicationCoordinatorImpl::reserveSnapshotName(OperationContext* opCtx) {
auto reservedName = SnapshotName(_snapshotNameGenerator.addAndFetch(1));
dassert(reservedName > SnapshotName::min());
dassert(reservedName < SnapshotName::max());
if (opCtx) {
ReplClientInfo::forClient(opCtx->getClient()).setLastSnapshot(reservedName);
}
return reservedName;
}
void ReplicationCoordinatorImpl::forceSnapshotCreation() {
_externalState->forceSnapshotCreation();
}
void ReplicationCoordinatorImpl::waitUntilSnapshotCommitted(OperationContext* opCtx,
const SnapshotName& untilSnapshot) {
stdx::unique_lock<stdx::mutex> lock(_mutex);
while (!_currentCommittedSnapshot || _currentCommittedSnapshot->name < untilSnapshot) {
opCtx->waitForConditionOrInterrupt(_currentCommittedSnapshotCond, lock);
}
}
size_t ReplicationCoordinatorImpl::getNumUncommittedSnapshots() {
return _uncommittedSnapshotsSize.load();
}
void ReplicationCoordinatorImpl::createSnapshot(OperationContext* opCtx,
OpTime timeOfSnapshot,
SnapshotName name) {
stdx::lock_guard<stdx::mutex> lock(_mutex);
_externalState->createSnapshot(opCtx, name);
auto snapshotInfo = SnapshotInfo{timeOfSnapshot, name};
if (timeOfSnapshot <= _lastCommittedOpTime) {
// This snapshot is ready to be marked as committed.
invariant(_uncommittedSnapshots.empty());
_updateCommittedSnapshot_inlock(snapshotInfo);
return;
}
if (!_uncommittedSnapshots.empty()) {
invariant(snapshotInfo > _uncommittedSnapshots.back());
// The name must independently be newer.
invariant(snapshotInfo.name > _uncommittedSnapshots.back().name);
// Technically, we could delete older snapshots from the same optime since we will only ever
// want the newest. However, multiple snapshots on the same optime will be very rare so it
// isn't worth the effort and potential bugs that would introduce.
}
_uncommittedSnapshots.push_back(snapshotInfo);
_uncommittedSnapshotsSize.store(_uncommittedSnapshots.size());
}
void ReplicationCoordinatorImpl::_updateCommittedSnapshot_inlock(
SnapshotInfo newCommittedSnapshot) {
invariant(!newCommittedSnapshot.opTime.isNull());
invariant(newCommittedSnapshot.opTime <= _lastCommittedOpTime);
if (_currentCommittedSnapshot) {
invariant(newCommittedSnapshot.opTime >= _currentCommittedSnapshot->opTime);
invariant(newCommittedSnapshot.name > _currentCommittedSnapshot->name);
}
if (!_uncommittedSnapshots.empty())
invariant(newCommittedSnapshot < _uncommittedSnapshots.front());
_currentCommittedSnapshot = newCommittedSnapshot;
_currentCommittedSnapshotCond.notify_all();
_externalState->updateCommittedSnapshot(newCommittedSnapshot.name);
// Wake up any threads waiting for read concern or write concern.
_wakeReadyWaiters_inlock();
}
void ReplicationCoordinatorImpl::dropAllSnapshots() {
stdx::lock_guard<stdx::mutex> lock(_mutex);
_dropAllSnapshots_inlock();
}
void ReplicationCoordinatorImpl::_dropAllSnapshots_inlock() {
_uncommittedSnapshots.clear();
_uncommittedSnapshotsSize.store(_uncommittedSnapshots.size());
_currentCommittedSnapshot = boost::none;
_externalState->dropAllSnapshots();
}
void ReplicationCoordinatorImpl::waitForElectionFinish_forTest() {
if (_electionFinishedEvent.isValid()) {
_replExecutor.waitForEvent(_electionFinishedEvent);
}
}
void ReplicationCoordinatorImpl::waitForElectionDryRunFinish_forTest() {
if (_electionDryRunFinishedEvent.isValid()) {
_replExecutor.waitForEvent(_electionDryRunFinishedEvent);
}
}
EventHandle ReplicationCoordinatorImpl::_resetElectionInfoOnProtocolVersionUpgrade(
const ReplSetConfig& oldConfig, const ReplSetConfig& newConfig) {
// On protocol version upgrade, reset last vote as if I just learned the term 0 from other
// nodes.
if (!oldConfig.isInitialized() ||
oldConfig.getProtocolVersion() >= newConfig.getProtocolVersion()) {
return {};
}
invariant(newConfig.getProtocolVersion() == 1);
// Write last vote
auto evhStatus = _replExecutor.makeEvent();
if (evhStatus.getStatus() == ErrorCodes::ShutdownInProgress) {
return {};
}
invariant(evhStatus.isOK());
auto evh = evhStatus.getValue();
auto cbStatus = _replExecutor.scheduleDBWork([this, evh](const CallbackArgs& cbData) {
if (cbData.status == ErrorCodes::CallbackCanceled) {
return;
}
invariant(cbData.opCtx);
LastVote lastVote{OpTime::kInitialTerm, -1};
auto status = _externalState->storeLocalLastVoteDocument(cbData.opCtx, lastVote);
invariant(status.isOK());
_replExecutor.signalEvent(evh);
});
if (cbStatus.getStatus() == ErrorCodes::ShutdownInProgress) {
return {};
}
invariant(cbStatus.isOK());
return evh;
}
CallbackHandle ReplicationCoordinatorImpl::_scheduleWork(const CallbackFn& work) {
auto scheduleFn = [this](const CallbackFn& workWrapped) {
return _replExecutor.scheduleWork(workWrapped);
};
return _wrapAndScheduleWork(scheduleFn, work);
}
CallbackHandle ReplicationCoordinatorImpl::_scheduleWorkAt(Date_t when, const CallbackFn& work) {
auto scheduleFn = [this, when](const CallbackFn& workWrapped) {
return _replExecutor.scheduleWorkAt(when, workWrapped);
};
return _wrapAndScheduleWork(scheduleFn, work);
}
void ReplicationCoordinatorImpl::_scheduleWorkAndWaitForCompletion(const CallbackFn& work) {
if (auto handle = _scheduleWork(work)) {
_replExecutor.wait(handle);
}
}
void ReplicationCoordinatorImpl::_scheduleWorkAtAndWaitForCompletion(Date_t when,
const CallbackFn& work) {
if (auto handle = _scheduleWorkAt(when, work)) {
_replExecutor.wait(handle);
}
}
CallbackHandle ReplicationCoordinatorImpl::_scheduleDBWork(const CallbackFn& work) {
auto scheduleFn = [this](const CallbackFn& workWrapped) {
return _replExecutor.scheduleDBWork(workWrapped);
};
return _wrapAndScheduleWork(scheduleFn, work);
}
CallbackHandle ReplicationCoordinatorImpl::_wrapAndScheduleWork(ScheduleFn scheduleFn,
const CallbackFn& work) {
auto workWrapped = [this, work](const CallbackArgs& args) {
if (args.status == ErrorCodes::CallbackCanceled) {
return;
}
work(args);
};
auto cbh = scheduleFn(workWrapped);
if (cbh.getStatus() == ErrorCodes::ShutdownInProgress) {
return CallbackHandle();
}
fassert(28800, cbh.getStatus());
return cbh.getValue();
}
EventHandle ReplicationCoordinatorImpl::_makeEvent() {
auto eventResult = this->_replExecutor.makeEvent();
if (eventResult.getStatus() == ErrorCodes::ShutdownInProgress) {
return EventHandle();
}
fassert(28825, eventResult.getStatus());
return eventResult.getValue();
}
void ReplicationCoordinatorImpl::_scheduleElectionWinNotification() {
stdx::lock_guard<stdx::mutex> lock(_mutex);
if (!_getMemberState_inlock().primary()) {
return;
}
_restartHeartbeats_inlock();
}
WriteConcernOptions ReplicationCoordinatorImpl::populateUnsetWriteConcernOptionsSyncMode(
WriteConcernOptions wc) {
WriteConcernOptions writeConcern(wc);
if (writeConcern.syncMode == WriteConcernOptions::SyncMode::UNSET) {
if (writeConcern.wMode == WriteConcernOptions::kMajority &&
getWriteConcernMajorityShouldJournal()) {
writeConcern.syncMode = WriteConcernOptions::SyncMode::JOURNAL;
} else {
writeConcern.syncMode = WriteConcernOptions::SyncMode::NONE;
}
}
return writeConcern;
}
CallbackFn ReplicationCoordinatorImpl::_wrapAsCallbackFn(const stdx::function<void()>& work) {
return [work](const CallbackArgs& cbData) {
if (cbData.status == ErrorCodes::CallbackCanceled) {
return;
}
work();
};
}
Status ReplicationCoordinatorImpl::stepUpIfEligible() {
if (!isV1ElectionProtocol()) {
return Status(ErrorCodes::CommandNotSupported,
"Step-up command is only supported by Protocol Version 1");
}
_startElectSelfIfEligibleV1(false);
EventHandle finishEvent;
{
LockGuard lk(_mutex);
finishEvent = _electionFinishedEvent;
}
if (finishEvent.isValid()) {
_replExecutor.waitForEvent(finishEvent);
}
auto state = getMemberState();
if (state.primary()) {
return Status::OK();
}
return Status(ErrorCodes::CommandFailed, "Election failed.");
}
ReplSettings::IndexPrefetchConfig ReplicationCoordinatorImpl::getIndexPrefetchConfig() const {
stdx::lock_guard<stdx::mutex> lock(_indexPrefetchMutex);
return _indexPrefetchConfig;
}
void ReplicationCoordinatorImpl::setIndexPrefetchConfig(
const ReplSettings::IndexPrefetchConfig cfg) {
stdx::lock_guard<stdx::mutex> lock(_indexPrefetchMutex);
_indexPrefetchConfig = cfg;
}
} // namespace repl
} // namespace mongo
|