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
|
/**
* Copyright (C) 2018-present MongoDB, Inc.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the Server Side Public License, version 1,
* as published by MongoDB, Inc.
*
* 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
* Server Side Public License for more details.
*
* You should have received a copy of the Server Side Public License
* along with this program. If not, see
* <http://www.mongodb.com/licensing/server-side-public-license>.
*
* 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 Server Side 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.
*/
#pragma once
#include "mongo/base/status_with.h"
#include "mongo/db/repl/optime.h"
namespace mongo {
class OperationContext;
namespace repl {
class StorageInterface;
class ReplicationConsistencyMarkers;
/**
* This class is used by the replication system to recover after an unclean shutdown or a rollback.
*/
class ReplicationRecovery {
public:
ReplicationRecovery() = default;
virtual ~ReplicationRecovery() = default;
/**
* Recovers the data on disk from the oplog. If the provided stable timestamp is not "none",
* this function assumes the data reflects that timestamp.
*/
virtual void recoverFromOplog(OperationContext* opCtx,
boost::optional<Timestamp> stableTimestamp) = 0;
/**
* Recovers the data on disk from the oplog and puts the node in readOnly mode. If
* 'takeUnstableCheckpointOnShutdown' is specified and an unstable checkpoint is present,
* ensures that recovery can be skipped safely.
*/
virtual void recoverFromOplogAsStandalone(OperationContext* opCtx) = 0;
/**
* Recovers the data on disk from the oplog up to and including the given timestamp.
*/
virtual void recoverFromOplogUpTo(OperationContext* opCtx, Timestamp endPoint) = 0;
};
class ReplicationRecoveryImpl : public ReplicationRecovery {
ReplicationRecoveryImpl(const ReplicationRecoveryImpl&) = delete;
ReplicationRecoveryImpl& operator=(const ReplicationRecoveryImpl&) = delete;
public:
ReplicationRecoveryImpl(StorageInterface* storageInterface,
ReplicationConsistencyMarkers* consistencyMarkers);
void recoverFromOplog(OperationContext* opCtx,
boost::optional<Timestamp> stableTimestamp) override;
void recoverFromOplogAsStandalone(OperationContext* opCtx) override;
void recoverFromOplogUpTo(OperationContext* opCtx, Timestamp endPoint) override;
private:
enum class RecoveryMode {
kStartupFromStableTimestamp,
kStartupFromUnstableCheckpoint,
kRollbackFromStableTimestamp,
// There is no RecoveryMode::kRollbackFromUnstableCheckpoint, rollback can only recover from
// a stable timestamp.
};
/**
* Confirms that the data and oplog all indicate that the nodes has an unstable checkpoint
* that is fully up to date.
*/
void _assertNoRecoveryNeededOnUnstableCheckpoint(OperationContext* opCtx);
/**
* After truncating the oplog, completes recovery if we're recovering from a stable timestamp
* or a stable checkpoint.
*/
void _recoverFromStableTimestamp(OperationContext* opCtx,
Timestamp stableTimestamp,
OpTime topOfOplog,
RecoveryMode recoveryMode);
/**
* After truncating the oplog, completes recovery if we're recovering from an unstable
* checkpoint.
*/
void _recoverFromUnstableCheckpoint(OperationContext* opCtx,
OpTime appliedThrough,
OpTime topOfOplog);
/**
* Applies all oplog entries from oplogApplicationStartPoint (exclusive) to topOfOplog
* (inclusive). This fasserts if oplogApplicationStartPoint is not in the oplog.
*/
void _applyToEndOfOplog(OperationContext* opCtx,
const Timestamp& oplogApplicationStartPoint,
const Timestamp& topOfOplog,
RecoveryMode recoveryMode);
/**
* Applies all oplog entries from startPoint (exclusive) to endPoint (inclusive). Returns the
* Timestamp of the last applied operation.
*/
Timestamp _applyOplogOperations(OperationContext* opCtx,
const Timestamp& startPoint,
const Timestamp& endPoint,
RecoveryMode recoveryMode);
/**
* Gets the last applied OpTime from the end of the oplog. Returns CollectionIsEmpty if there is
* no oplog.
*/
StatusWith<OpTime> _getTopOfOplog(OperationContext* opCtx) const;
/**
* Truncates the oplog after the "truncateAfterTimestamp" entry.
* If the stableTimestamp is set, may move it backwards to the new top of oplog.
*/
void _truncateOplogTo(OperationContext* opCtx,
Timestamp truncateAfterTimestamp,
boost::optional<Timestamp>* stableTimestamp);
/**
* Uses the oplogTruncateAfterPoint, accessed via '_consistencyMarkers', to decide whether to
* truncate part of the oplog. If oplogTruncateAfterPoint has been set, then there may be holes
* in the oplog after that point. In that case, we will truncate the oplog entries starting at
* and including the entry associated with the oplogTruncateAfterPoint timestamp.
*
* If the oplogTruncateAfterPoint is earlier in time than or equal to the stable timestamp, we
* will truncate the oplog after the stable timestamp instead. There cannot be holes before a
* stable timestamp. The oplogTruncateAfterPoint can lag behind the stable timestamp because the
* oplogTruncateAfterPoint is updated on primaries by an asynchronously looping thread that can
* potentially be starved.
*
* If the stable timestamp is at a hole, this will move the stable timestamp back to the new
* top of oplog. This can happen on primaries when EMRC=false or in single-node replica sets.
*/
void _truncateOplogIfNeededAndThenClearOplogTruncateAfterPoint(
OperationContext* opCtx, boost::optional<Timestamp>* stableTimestamp);
/**
* Checks if the proposed oplog application start point (which is typically derived from the
* stable timestamp) exists in the oplog. If it does, this returns that same start point
* unchanged. If that point is not in the oplog, this function returns an entry before
* that start point.
* It is safe to do as as we make sure that we always keep an oplog entry that is less than
* or equal to the stable timestamp so such a correction always pushes the start point back and
* never forward. Applying entries from an earlier point is permissible due to oplog entry
* idempotency (and also due to the order being preserved.)
*/
Timestamp _adjustStartPointIfNecessary(OperationContext* opCtx, Timestamp startPoint);
StorageInterface* _storageInterface;
ReplicationConsistencyMarkers* _consistencyMarkers;
};
} // namespace repl
} // namespace mongo
|
