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
|
/**
* Copyright 2015 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.
*/
#pragma once
#include <string>
#include "mongo/base/disallow_copying.h"
#include "mongo/bson/simple_bsonobj_comparator.h"
#include "mongo/db/catalog/collection.h"
#include "mongo/db/catalog/collection_options.h"
#include "mongo/db/jsobj.h"
#include "mongo/db/repl/rollback.h"
#include "mongo/db/s/collection_sharding_state.h"
namespace mongo {
struct InsertStatement;
class OperationContext;
namespace repl {
class OpTime;
} // namespace repl
/**
* Holds document update information used in logging.
*/
struct OplogUpdateEntryArgs {
enum class StoreDocOption { None, PreImage, PostImage };
// Name of the collection in which document is being updated.
NamespaceString nss;
OptionalCollectionUUID uuid;
StmtId stmtId = kUninitializedStmtId;
// The document before modifiers were applied.
boost::optional<BSONObj> preImageDoc;
// Fully updated document with damages (update modifiers) applied.
BSONObj updatedDoc;
// Document containing update modifiers -- e.g. $set and $unset
BSONObj update;
// Document containing the _id field of the doc being updated.
BSONObj criteria;
// True if this update comes from a chunk migration.
bool fromMigrate = false;
StoreDocOption storeDocOption = StoreDocOption::None;
};
struct TTLCollModInfo {
Seconds expireAfterSeconds;
Seconds oldExpireAfterSeconds;
std::string indexName;
};
/**
* The OpObserver interface contains methods that get called on certain database events. It provides
* a way for various server subsystems to be notified of other events throughout the server.
*
* In order to call any OpObserver method, you must be in a 'WriteUnitOfWork'. This means that any
* locks acquired for writes in that WUOW are still held. So, you can assume that any locks required
* to perform the operation being observed are still held. These rules should apply for all observer
* methods unless otherwise specified.
*/
class OpObserver {
public:
virtual ~OpObserver() = default;
virtual void onCreateIndex(OperationContext* opCtx,
const NamespaceString& nss,
OptionalCollectionUUID uuid,
BSONObj indexDoc,
bool fromMigrate) = 0;
virtual void onInserts(OperationContext* opCtx,
const NamespaceString& nss,
OptionalCollectionUUID uuid,
std::vector<InsertStatement>::const_iterator begin,
std::vector<InsertStatement>::const_iterator end,
bool fromMigrate) = 0;
virtual void onUpdate(OperationContext* opCtx, const OplogUpdateEntryArgs& args) = 0;
virtual void aboutToDelete(OperationContext* opCtx,
const NamespaceString& nss,
const BSONObj& doc) = 0;
/**
* Handles logging before document is deleted.
*
* "ns" name of the collection from which deleteState.idDoc will be deleted.
* "fromMigrate" indicates whether the delete was induced by a chunk migration, and
* so should be ignored by the user as an internal maintenance operation and not a
* real delete.
*/
virtual void onDelete(OperationContext* opCtx,
const NamespaceString& nss,
OptionalCollectionUUID uuid,
StmtId stmtId,
bool fromMigrate,
const boost::optional<BSONObj>& deletedDoc) = 0;
/**
* Logs a no-op with "msgObj" in the o field into oplog.
*
* This function should only be used internally. "nss", "uuid" and the o2 field should never be
* exposed to users (for instance through the appendOplogNote command).
*/
virtual void onInternalOpMessage(OperationContext* opCtx,
const NamespaceString& nss,
const boost::optional<UUID> uuid,
const BSONObj& msgObj,
const boost::optional<BSONObj> o2MsgObj) = 0;
/**
* Logs a no-op with "msgObj" in the o field into oplog.
*/
void onOpMessage(OperationContext* opCtx, const BSONObj& msgObj) {
onInternalOpMessage(opCtx, {}, boost::none, msgObj, boost::none);
}
virtual void onCreateCollection(OperationContext* opCtx,
Collection* coll,
const NamespaceString& collectionName,
const CollectionOptions& options,
const BSONObj& idIndex) = 0;
/**
* This function logs an oplog entry when a 'collMod' command on a collection is executed.
* Since 'collMod' commands can take a variety of different formats, the 'o' field of the
* oplog entry is populated with the 'collMod' command object. For TTL index updates, we
* transform key pattern index specifications into index name specifications, for uniformity.
* All other collMod fields are added to the 'o' object without modifications.
*
* To facilitate the rollback process, 'oldCollOptions' contains the previous state of all
* collection options i.e. the state prior to completion of the current collMod command.
* 'ttlInfo' contains the index name and previous expiration time of a TTL index. The old
* collection options will be stored in the 'o2.collectionOptions_old' field, and the old TTL
* expiration value in the 'o2.expireAfterSeconds_old' field.
*
* Oplog Entry Example ('o' and 'o2' fields shown):
*
* {
* ...
* o: {
* collMod: "test",
* validationLevel: "off",
* index: {name: "indexName_1", expireAfterSeconds: 600}
* }
* o2: {
* collectionOptions_old: {
* validationLevel: "strict",
* },
* expireAfterSeconds_old: 300
* }
* }
*
*/
virtual void onCollMod(OperationContext* opCtx,
const NamespaceString& nss,
OptionalCollectionUUID uuid,
const BSONObj& collModCmd,
const CollectionOptions& oldCollOptions,
boost::optional<TTLCollModInfo> ttlInfo) = 0;
virtual void onDropDatabase(OperationContext* opCtx, const std::string& dbName) = 0;
/**
* This function logs an oplog entry when a 'drop' command on a collection is executed.
* Returns the optime of the oplog entry successfully written to the oplog.
* Returns a null optime if an oplog entry was not written for this operation.
*/
virtual repl::OpTime onDropCollection(OperationContext* opCtx,
const NamespaceString& collectionName,
OptionalCollectionUUID uuid) = 0;
/**
* This function logs an oplog entry when an index is dropped. The namespace of the index,
* the index name, and the index info from the index descriptor are used to create a
* 'dropIndexes' op where the 'o' field is the name of the index and the 'o2' field is the
* index info. The index info can then be used to reconstruct the index on rollback.
*
* If a user specifies {dropIndexes: 'foo', index: '*'}, each index dropped will have its own
* oplog entry. This means it's possible to roll back half of the index drops.
*/
virtual void onDropIndex(OperationContext* opCtx,
const NamespaceString& nss,
OptionalCollectionUUID uuid,
const std::string& indexName,
const BSONObj& indexInfo) = 0;
/**
* This function logs an oplog entry when a 'renameCollection' command on a collection is
* executed.
* Returns the optime of the oplog entry successfully written to the oplog.
* Returns a null optime if an oplog entry was not written for this operation.
*/
virtual repl::OpTime onRenameCollection(OperationContext* opCtx,
const NamespaceString& fromCollection,
const NamespaceString& toCollection,
OptionalCollectionUUID uuid,
bool dropTarget,
OptionalCollectionUUID dropTargetUUID,
bool stayTemp) = 0;
virtual void onApplyOps(OperationContext* opCtx,
const std::string& dbName,
const BSONObj& applyOpCmd) = 0;
virtual void onEmptyCapped(OperationContext* opCtx,
const NamespaceString& collectionName,
OptionalCollectionUUID uuid) = 0;
/**
* The onTransactionCommit method is called on the commit of an atomic transaction, before the
* RecoveryUnit onCommit() is called. It must not be called when no transaction is active.
*/
virtual void onTransactionCommit(OperationContext* opCtx) = 0;
/**
* The onTransactionAbort method is called when an atomic transaction aborts, before the
* RecoveryUnit onRollback() is called. It must not be called when no transaction is active.
*/
virtual void onTransactionAbort(OperationContext* opCtx) = 0;
/**
* A structure to hold information about a replication rollback suitable to be passed along to
* any external subsystems that need to be notified of a rollback occurring.
*/
struct RollbackObserverInfo {
// Set of all namespaces from ops being rolled back.
std::set<NamespaceString> rollbackNamespaces = {};
// Set of all session ids from ops being rolled back.
std::set<UUID> rollbackSessionIds = {};
// Maps UUIDs to a set of BSONObjs containing the _ids of the documents that will be deleted
// from that collection due to rollback, and is used to populate rollback files.
// For simplicity, this BSONObj set uses the simple binary comparison, as it is never wrong
// to consider two _ids as distinct even if the collection default collation would put them
// in the same equivalence class.
stdx::unordered_map<UUID, SimpleBSONObjUnorderedSet, UUID::Hash> rollbackDeletedIdsMap;
// True if the shard identity document was rolled back.
bool shardIdentityRolledBack = false;
};
/**
* This function will get called after the replication system has completed a rollback. This
* means that all on-disk, replicated data will have been reverted to the rollback common point
* by the time this function is called. Subsystems may use this method to invalidate any in
* memory caches or, optionally, rebuild any data structures from the data that is now on disk.
* This function should not write any persistent state.
*
* When this function is called, there will be no locks held on the given OperationContext, and
* it will not be called inside an existing WriteUnitOfWork. Any work done inside this handler
* is expected to handle this on its own.
*
* This method is only applicable to the "rollback to a stable timestamp" algorithm, and is not
* called when using any other rollback algorithm i.e "rollback via refetch".
*/
virtual void onReplicationRollback(OperationContext* opCtx,
const RollbackObserverInfo& rbInfo) = 0;
struct Times;
protected:
class ReservedTimes;
};
/**
* This struct is a decoration for `OperationContext` which contains collected `repl::OpTime`
* and `Date_t` timestamps of various critical stages of an operation performed by an OpObserver
* chain.
*/
struct OpObserver::Times {
static Times& get(OperationContext*);
std::vector<repl::OpTime> reservedOpTimes;
private:
friend OpObserver::ReservedTimes;
// Because `OpObserver`s are re-entrant, it is necessary to track the recursion depth to know
// when to actually clear the `reservedOpTimes` vector, using the `ReservedTimes` scope object.
int _recursionDepth = 0;
};
/**
* This class is an RAII object to manage the state of the `OpObserver::Times` decoration on an
* operation context. Upon destruction the list of times in the decoration on the operation context
* is cleared. It is intended for use as a scope object in `OpObserverRegistry` to manage
* re-entrancy.
*/
class OpObserver::ReservedTimes {
ReservedTimes(const ReservedTimes&) = delete;
ReservedTimes& operator=(const ReservedTimes&) = delete;
public:
explicit ReservedTimes(OperationContext* const opCtx);
~ReservedTimes();
const Times& get() const {
return _times;
}
private:
Times& _times;
};
} // namespace mongo
|