summaryrefslogtreecommitdiff
path: root/src/mongo/db/index_builder.cpp
blob: 7559d91959142d1ee3245c5281fb2bd5f1c405b4 (plain)
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
/**
 *    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.
 */

#define MONGO_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kIndex

#include "mongo/platform/basic.h"

#include "mongo/db/index_builder.h"

#include "mongo/bson/bsonobjbuilder.h"
#include "mongo/db/auth/authorization_session.h"
#include "mongo/db/catalog/database.h"
#include "mongo/db/catalog/database_holder.h"
#include "mongo/db/catalog/index_timestamp_helper.h"
#include "mongo/db/catalog/multi_index_block.h"
#include "mongo/db/client.h"
#include "mongo/db/concurrency/write_conflict_exception.h"
#include "mongo/db/curop.h"
#include "mongo/db/db_raii.h"
#include "mongo/db/logical_clock.h"
#include "mongo/db/op_observer.h"
#include "mongo/db/repl/timestamp_block.h"
#include "mongo/db/server_options.h"
#include "mongo/util/assert_util.h"
#include "mongo/util/log.h"
#include "mongo/util/mongoutils/str.h"

namespace mongo {

using std::endl;

AtomicWord<unsigned> IndexBuilder::_indexBuildCount;

namespace {

const StringData kIndexesFieldName = "indexes"_sd;
const StringData kCommandName = "createIndexes"_sd;

// Synchronization tools when replication spawns a background index in a new thread.
// The bool is 'true' when a new background index has started in a new thread but the
// parent thread has not yet synchronized with it.
bool _bgIndexStarting(false);
stdx::mutex _bgIndexStartingMutex;
stdx::condition_variable _bgIndexStartingCondVar;

void _setBgIndexStarting() {
    stdx::lock_guard<stdx::mutex> lk(_bgIndexStartingMutex);
    invariant(_bgIndexStarting == false);
    _bgIndexStarting = true;
    _bgIndexStartingCondVar.notify_one();
}
}  // namespace

IndexBuilder::IndexBuilder(const BSONObj& index,
                           IndexConstraints indexConstraints,
                           ReplicatedWrites replicatedWrites,
                           Timestamp initIndexTs)
    : BackgroundJob(true /* self-delete */),
      _index(index.getOwned()),
      _indexConstraints(indexConstraints),
      _replicatedWrites(replicatedWrites),
      _initIndexTs(initIndexTs),
      _name(str::stream() << "repl-index-builder-" << _indexBuildCount.addAndFetch(1)) {}

IndexBuilder::~IndexBuilder() {}

bool IndexBuilder::canBuildInBackground() {
    return MultiIndexBlock::areHybridIndexBuildsEnabled();
}

std::string IndexBuilder::name() const {
    return _name;
}

void IndexBuilder::run() {
    ThreadClient tc(name(), getGlobalServiceContext());
    LOG(2) << "IndexBuilder building index " << _index;

    auto opCtx = cc().makeOperationContext();
    ShouldNotConflictWithSecondaryBatchApplicationBlock shouldNotConflictBlock(opCtx->lockState());

    // If the calling thread is not replicating writes, neither should this thread.
    boost::optional<repl::UnreplicatedWritesBlock> unreplicatedWrites;
    if (_replicatedWrites == ReplicatedWrites::kUnreplicated) {
        unreplicatedWrites.emplace(opCtx.get());
    }

    AuthorizationSession::get(opCtx->getClient())->grantInternalAuthorization();

    {
        stdx::lock_guard<Client> lk(*(opCtx->getClient()));
        CurOp::get(opCtx.get())->setNetworkOp_inlock(dbInsert);
    }
    NamespaceString ns(_index["ns"].String());

    Lock::DBLock dlk(opCtx.get(), ns.db(), MODE_X);
    auto databaseHolder = DatabaseHolder::get(opCtx.get());
    auto db = databaseHolder->getDb(opCtx.get(), ns.db().toString());

    // This background index build can only be interrupted at shutdown.
    // For the duration of the OperationContext::runWithoutInterruption() invocation, any kill
    // status set by the killOp command will be ignored.
    // After OperationContext::runWithoutInterruption() returns, any call to
    // OperationContext::checkForInterrupt() will see the kill status and respond accordingly
    // (checkForInterrupt() will throw an exception while checkForInterruptNoAssert() returns
    // an error Status).
    Status status = opCtx->runWithoutInterruption([&, this] {
        return _buildAndHandleErrors(opCtx.get(), db, true /* buildInBackground */, &dlk);
    });
    if (!status.isOK()) {
        error() << "IndexBuilder could not build index: " << redact(status);
        fassert(28555, ErrorCodes::isInterruption(status.code()));
    }
}

Status IndexBuilder::buildInForeground(OperationContext* opCtx, Database* db) const {
    return _buildAndHandleErrors(opCtx, db, false /*buildInBackground */, nullptr);
}

void IndexBuilder::waitForBgIndexStarting() {
    stdx::unique_lock<stdx::mutex> lk(_bgIndexStartingMutex);
    while (_bgIndexStarting == false) {
        _bgIndexStartingCondVar.wait(lk);
    }
    // Reset for next time.
    _bgIndexStarting = false;
}

Status IndexBuilder::_buildAndHandleErrors(OperationContext* opCtx,
                                           Database* db,
                                           bool buildInBackground,
                                           Lock::DBLock* dbLock) const {
    const NamespaceString ns(_index["ns"].String());

    Collection* coll = db->getCollection(opCtx, ns);
    // Collections should not be implicitly created by the index builder.
    fassert(40409, coll);

    MultiIndexBlock indexer(opCtx, coll);

    auto status = _build(opCtx, buildInBackground, coll, indexer, dbLock);
    // Background index builds are not allowed to return errors because they run in a background
    // thread.
    if (status.isOK() || !buildInBackground) {
        invariant(!dbLock || dbLock->mode() == MODE_X);
        return status;
    }

    // The MultiIndexBlock destructor may only be called when an X lock is held on the database.
    if (dbLock->mode() != MODE_X) {
        UninterruptibleLockGuard noInterrupt(opCtx->lockState());
        dbLock->relockWithMode(MODE_X);
    }

    invariant(status.code() != ErrorCodes::WriteConflict);

    if (status.code() == ErrorCodes::InterruptedAtShutdown) {
        // leave it as-if kill -9 happened. This will be handled on restart.
        indexer.abortWithoutCleanup();
        return status;
    }

    error() << "Background index build failed. Status: " << redact(status);
    fassertFailed(50769);
}

Status IndexBuilder::_build(OperationContext* opCtx,
                            bool buildInBackground,
                            Collection* coll,
                            MultiIndexBlock& indexer,
                            Lock::DBLock* dbLock) const try {
    auto ns = coll->ns();

    {
        BSONObjBuilder builder;
        builder.append(kCommandName, ns.coll());
        {
            BSONArrayBuilder indexesBuilder;
            indexesBuilder.append(_index);
            builder.append(kIndexesFieldName, indexesBuilder.arr());
        }
        auto opDescObj = builder.obj();

        stdx::lock_guard<Client> lk(*opCtx->getClient());
        // Show which index we're building in the curop display.
        auto curOp = CurOp::get(opCtx);
        curOp->setLogicalOp_inlock(LogicalOp::opCommand);
        curOp->setNS_inlock(ns.ns());
        curOp->setOpDescription_inlock(opDescObj);
    }

    // Ignore uniqueness constraint violations when relaxed (on secondaries). Secondaries can
    // complete index builds in the middle of batches, which creates the potential for finding
    // duplicate key violations where there otherwise would be none at consistent states.
    if (_indexConstraints == IndexConstraints::kRelax) {
        indexer.ignoreUniqueConstraint();
    }

    Status status = Status::OK();

    {
        TimestampBlock tsBlock(opCtx, _initIndexTs);
        status = writeConflictRetry(opCtx, "Init index build", ns.ns(), [&] {
            return indexer.init(_index, MultiIndexBlock::makeTimestampedIndexOnInitFn(opCtx, coll))
                .getStatus();
        });
    }

    if (status == ErrorCodes::IndexAlreadyExists ||
        (status == ErrorCodes::IndexOptionsConflict &&
         _indexConstraints == IndexConstraints::kRelax)) {
        LOG(1) << "Ignoring indexing error: " << redact(status);

        // Must set this in case anyone is waiting for this build.
        if (dbLock) {
            _setBgIndexStarting();
        }
        return Status::OK();
    }
    if (!status.isOK()) {
        return status;
    }

    if (buildInBackground) {
        invariant(dbLock);

        _setBgIndexStarting();
        opCtx->recoveryUnit()->abandonSnapshot();

        UninterruptibleLockGuard noInterrupt(opCtx->lockState());
        dbLock->relockWithMode(MODE_IX);
    }

    {
        Lock::CollectionLock collLock(opCtx->lockState(), ns.ns(), MODE_IX);
        // WriteConflict exceptions and statuses are not expected to escape this method.
        status = indexer.insertAllDocumentsInCollection();
    }
    if (!status.isOK()) {
        return status;
    }

    if (buildInBackground) {
        {
            // Perform the first drain while holding an intent lock.
            Lock::CollectionLock collLock(opCtx->lockState(), ns.ns(), MODE_IX);

            // Read at a point in time so that the drain, which will timestamp writes at
            // lastApplied, can never commit writes earlier than its read timestamp.
            status = indexer.drainBackgroundWrites(RecoveryUnit::ReadSource::kNoOverlap);
        }
        if (!status.isOK()) {
            return status;
        }

        // Perform the second drain while stopping inserts into the collection.
        {
            Lock::CollectionLock colLock(opCtx->lockState(), ns.ns(), MODE_S);
            status = indexer.drainBackgroundWrites();
        }
        if (!status.isOK()) {
            return status;
        }

        opCtx->recoveryUnit()->abandonSnapshot();

        UninterruptibleLockGuard noInterrupt(opCtx->lockState());
        dbLock->relockWithMode(MODE_X);

        // Perform the third and final drain after releasing a shared lock and reacquiring an
        // exclusive lock on the database.
        status = indexer.drainBackgroundWrites();
        if (!status.isOK()) {
            return status;
        }

        // Only perform constraint checking when enforced (on primaries).
        if (_indexConstraints == IndexConstraints::kEnforce) {
            status = indexer.checkConstraints();
            if (!status.isOK()) {
                return status;
            }
        }
    }

    status = writeConflictRetry(opCtx, "Commit index build", ns.ns(), [opCtx, coll, &indexer, &ns] {
        WriteUnitOfWork wunit(opCtx);
        auto status = indexer.commit(
            [opCtx, coll, &ns](const BSONObj& indexSpec) {
                opCtx->getServiceContext()->getOpObserver()->onCreateIndex(
                    opCtx, ns, *(coll->uuid()), indexSpec, false);
            },
            MultiIndexBlock::kNoopOnCommitFn);
        if (!status.isOK()) {
            return status;
        }

        IndexTimestampHelper::setGhostCommitTimestampForCatalogWrite(opCtx, ns);
        wunit.commit();
        return Status::OK();
    });
    if (!status.isOK()) {
        return status;
    }

    if (buildInBackground) {
        invariant(opCtx->lockState()->isDbLockedForMode(ns.db(), MODE_X),
                  str::stream() << "Database not locked in exclusive mode after committing "
                                   "background index: "
                                << ns.ns()
                                << ": "
                                << _index);
        auto databaseHolder = DatabaseHolder::get(opCtx);
        auto reloadDb = databaseHolder->getDb(opCtx, ns.db());
        fassert(28553, reloadDb);
        fassert(28554, reloadDb->getCollection(opCtx, ns));
    }

    return Status::OK();
} catch (const DBException& e) {
    return e.toStatus();
}
}