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
|
/**
* Copyright (C) 2012 10gen 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/bgsync.h"
#include "mongo/base/counter.h"
#include "mongo/db/auth/authorization_session.h"
#include "mongo/db/client.h"
#include "mongo/db/commands/fsync.h"
#include "mongo/db/commands/server_status_metric.h"
#include "mongo/db/dbhelpers.h"
#include "mongo/db/operation_context_impl.h"
#include "mongo/db/repl/oplog.h"
#include "mongo/db/repl/replication_coordinator_global.h"
#include "mongo/db/repl/replication_coordinator_impl.h"
#include "mongo/db/repl/rs_rollback.h"
#include "mongo/db/repl/rs_sync.h"
#include "mongo/db/stats/timer_stats.h"
#include "mongo/util/exit.h"
#include "mongo/util/fail_point_service.h"
#include "mongo/util/log.h"
namespace mongo {
using std::string;
namespace repl {
namespace {
const char hashFieldName[] = "h";
int SleepToAllowBatchingMillis = 2;
const int BatchIsSmallish = 40000; // bytes
} // namespace
MONGO_FP_DECLARE(rsBgSyncProduce);
BackgroundSync* BackgroundSync::s_instance = 0;
boost::mutex BackgroundSync::s_mutex;
//The number and time spent reading batches off the network
static TimerStats getmoreReplStats;
static ServerStatusMetricField<TimerStats> displayBatchesRecieved(
"repl.network.getmores",
&getmoreReplStats );
//The oplog entries read via the oplog reader
static Counter64 opsReadStats;
static ServerStatusMetricField<Counter64> displayOpsRead( "repl.network.ops",
&opsReadStats );
//The bytes read via the oplog reader
static Counter64 networkByteStats;
static ServerStatusMetricField<Counter64> displayBytesRead( "repl.network.bytes",
&networkByteStats );
//The count of items in the buffer
static Counter64 bufferCountGauge;
static ServerStatusMetricField<Counter64> displayBufferCount( "repl.buffer.count",
&bufferCountGauge );
//The size (bytes) of items in the buffer
static Counter64 bufferSizeGauge;
static ServerStatusMetricField<Counter64> displayBufferSize( "repl.buffer.sizeBytes",
&bufferSizeGauge );
//The max size (bytes) of the buffer
static int bufferMaxSizeGauge = 256*1024*1024;
static ServerStatusMetricField<int> displayBufferMaxSize( "repl.buffer.maxSizeBytes",
&bufferMaxSizeGauge );
BackgroundSyncInterface::~BackgroundSyncInterface() {}
size_t getSize(const BSONObj& o) {
// SERVER-9808 Avoid Fortify complaint about implicit signed->unsigned conversion
return static_cast<size_t>(o.objsize());
}
BackgroundSync::BackgroundSync() : _buffer(bufferMaxSizeGauge, &getSize),
_lastOpTimeFetched(std::numeric_limits<int>::max(),
0),
_lastAppliedHash(0),
_lastFetchedHash(0),
_pause(true),
_appliedBuffer(true),
_replCoord(getGlobalReplicationCoordinator()),
_initialSyncRequestedFlag(false),
_indexPrefetchConfig(PREFETCH_ALL) {
}
BackgroundSync* BackgroundSync::get() {
boost::unique_lock<boost::mutex> lock(s_mutex);
if (s_instance == NULL && !inShutdown()) {
s_instance = new BackgroundSync();
}
return s_instance;
}
void BackgroundSync::shutdown() {
boost::lock_guard<boost::mutex> lock(_mutex);
// Clear the buffer in case the producerThread is waiting in push() due to a full queue.
invariant(inShutdown());
_buffer.clear();
_pause = true;
// Wake up producerThread so it notices that we're in shutdown
_appliedBufferCondition.notify_all();
_pausedCondition.notify_all();
}
void BackgroundSync::notify(OperationContext* txn) {
boost::lock_guard<boost::mutex> lock(_mutex);
// If all ops in the buffer have been applied, unblock waitForRepl (if it's waiting)
if (_buffer.empty()) {
_appliedBuffer = true;
_appliedBufferCondition.notify_all();
}
}
void BackgroundSync::producerThread() {
Client::initThread("rsBackgroundSync");
AuthorizationSession::get(cc())->grantInternalAuthorization();
while (!inShutdown()) {
try {
_producerThread();
}
catch (const DBException& e) {
std::string msg(str::stream() << "sync producer problem: " << e.toString());
error() << msg;
_replCoord->setMyHeartbeatMessage(msg);
}
catch (const std::exception& e2) {
severe() << "sync producer exception: " << e2.what();
fassertFailed(28546);
}
}
cc().shutdown();
}
void BackgroundSync::_producerThread() {
const MemberState state = _replCoord->getMemberState();
// we want to pause when the state changes to primary
if (_replCoord->isWaitingForApplierToDrain() || state.primary()) {
if (!_pause) {
stop();
}
sleepsecs(1);
return;
}
// TODO(spencer): Use a condition variable to await loading a config.
if (state.startup()) {
// Wait for a config to be loaded
sleepsecs(1);
return;
}
OperationContextImpl txn;
// We need to wait until initial sync has started.
if (_replCoord->getMyLastOptime().isNull()) {
sleepsecs(1);
return;
}
// we want to unpause when we're no longer primary
// start() also loads _lastOpTimeFetched, which we know is set from the "if"
else if (_pause) {
start(&txn);
}
produce(&txn);
}
void BackgroundSync::produce(OperationContext* txn) {
// this oplog reader does not do a handshake because we don't want the server it's syncing
// from to track how far it has synced
{
boost::unique_lock<boost::mutex> lock(_mutex);
if (_lastOpTimeFetched.isNull()) {
// then we're initial syncing and we're still waiting for this to be set
lock.unlock();
sleepsecs(1);
// if there is no one to sync from
return;
}
// Wait until we've applied the ops we have before we choose a sync target
while (!_appliedBuffer && !inShutdownStrict()) {
_appliedBufferCondition.wait(lock);
}
if (inShutdownStrict()) {
return;
}
}
while (MONGO_FAIL_POINT(rsBgSyncProduce)) {
sleepmillis(0);
}
// find a target to sync from the last optime fetched
Timestamp lastOpTimeFetched;
{
boost::unique_lock<boost::mutex> lock(_mutex);
lastOpTimeFetched = _lastOpTimeFetched;
_syncSourceHost = HostAndPort();
}
_syncSourceReader.resetConnection();
_syncSourceReader.connectToSyncSource(txn, lastOpTimeFetched, _replCoord);
{
boost::unique_lock<boost::mutex> lock(_mutex);
// no server found
if (_syncSourceReader.getHost().empty()) {
lock.unlock();
sleepsecs(1);
// if there is no one to sync from
return;
}
lastOpTimeFetched = _lastOpTimeFetched;
_syncSourceHost = _syncSourceReader.getHost();
_replCoord->signalUpstreamUpdater();
}
_syncSourceReader.tailingQueryGTE(rsOplogName.c_str(), lastOpTimeFetched);
// if target cut connections between connecting and querying (for
// example, because it stepped down) we might not have a cursor
if (!_syncSourceReader.haveCursor()) {
return;
}
if (_rollbackIfNeeded(txn, _syncSourceReader)) {
stop();
return;
}
while (!inShutdown()) {
if (!_syncSourceReader.moreInCurrentBatch()) {
// Check some things periodically
// (whenever we run out of items in the
// current cursor batch)
int bs = _syncSourceReader.currentBatchMessageSize();
if( bs > 0 && bs < BatchIsSmallish ) {
// on a very low latency network, if we don't wait a little, we'll be
// getting ops to write almost one at a time. this will both be expensive
// for the upstream server as well as potentially defeating our parallel
// application of batches on the secondary.
//
// the inference here is basically if the batch is really small, we are
// "caught up".
//
sleepmillis(SleepToAllowBatchingMillis);
}
// If we are transitioning to primary state, we need to leave
// this loop in order to go into bgsync-pause mode.
if (_replCoord->isWaitingForApplierToDrain() ||
_replCoord->getMemberState().primary()) {
return;
}
// re-evaluate quality of sync target
if (shouldChangeSyncSource()) {
return;
}
{
//record time for each getmore
TimerHolder batchTimer(&getmoreReplStats);
// This calls receiveMore() on the oplogreader cursor.
// It can wait up to five seconds for more data.
_syncSourceReader.more();
}
networkByteStats.increment(_syncSourceReader.currentBatchMessageSize());
if (!_syncSourceReader.moreInCurrentBatch()) {
// If there is still no data from upstream, check a few more things
// and then loop back for another pass at getting more data
{
boost::unique_lock<boost::mutex> lock(_mutex);
if (_pause) {
return;
}
}
_syncSourceReader.tailCheck();
if( !_syncSourceReader.haveCursor() ) {
LOG(1) << "replSet end syncTail pass";
return;
}
continue;
}
}
// If we are transitioning to primary state, we need to leave
// this loop in order to go into bgsync-pause mode.
if (_replCoord->isWaitingForApplierToDrain() ||
_replCoord->getMemberState().primary()) {
LOG(1) << "waiting for draining or we are primary, not adding more ops to buffer";
return;
}
// At this point, we are guaranteed to have at least one thing to read out
// of the oplogreader cursor.
BSONObj o = _syncSourceReader.nextSafe().getOwned();
opsReadStats.increment();
{
boost::unique_lock<boost::mutex> lock(_mutex);
_appliedBuffer = false;
}
OCCASIONALLY {
LOG(2) << "bgsync buffer has " << _buffer.size() << " bytes";
}
bufferCountGauge.increment();
bufferSizeGauge.increment(getSize(o));
_buffer.push(o);
{
boost::unique_lock<boost::mutex> lock(_mutex);
_lastFetchedHash = o["h"].numberLong();
_lastOpTimeFetched = o["ts"].timestamp();
LOG(3) << "lastOpTimeFetched: " << _lastOpTimeFetched.toStringPretty();
}
}
}
bool BackgroundSync::shouldChangeSyncSource() {
// is it even still around?
if (getSyncTarget().empty() || _syncSourceReader.getHost().empty()) {
return true;
}
// check other members: is any member's optime more than MaxSyncSourceLag seconds
// ahead of the current sync source?
return _replCoord->shouldChangeSyncSource(_syncSourceReader.getHost());
}
bool BackgroundSync::peek(BSONObj* op) {
return _buffer.peek(*op);
}
void BackgroundSync::waitForMore() {
BSONObj op;
// Block for one second before timing out.
// Ignore the value of the op we peeked at.
_buffer.blockingPeek(op, 1);
}
void BackgroundSync::consume() {
// this is just to get the op off the queue, it's been peeked at
// and queued for application already
BSONObj op = _buffer.blockingPop();
bufferCountGauge.decrement(1);
bufferSizeGauge.decrement(getSize(op));
}
bool BackgroundSync::_rollbackIfNeeded(OperationContext* txn, OplogReader& r) {
string hn = r.conn()->getServerAddress();
if (!r.more()) {
try {
BSONObj theirLastOp = r.getLastOp(rsOplogName.c_str());
if (theirLastOp.isEmpty()) {
error() << "empty query result from " << hn << " oplog";
sleepsecs(2);
return true;
}
Timestamp theirTS = theirLastOp["ts"].timestamp();
if (theirTS < _lastOpTimeFetched) {
log() << "we are ahead of the sync source, will try to roll back";
syncRollback(txn, _replCoord->getMyLastOptime(), &r, _replCoord);
return true;
}
/* we're not ahead? maybe our new query got fresher data. best to come back and try again */
log() << "syncTail condition 1";
sleepsecs(1);
}
catch(DBException& e) {
error() << "querying " << hn << ' ' << e.toString();
sleepsecs(2);
}
return true;
}
BSONObj o = r.nextSafe();
Timestamp ts = o["ts"].timestamp();
long long hash = o["h"].numberLong();
if( ts != _lastOpTimeFetched || hash != _lastFetchedHash ) {
log() << "our last op time fetched: " << _lastOpTimeFetched.toStringPretty();
log() << "source's GTE: " << ts.toStringPretty();
syncRollback(txn, _replCoord->getMyLastOptime(), &r, _replCoord);
return true;
}
return false;
}
HostAndPort BackgroundSync::getSyncTarget() {
boost::unique_lock<boost::mutex> lock(_mutex);
return _syncSourceHost;
}
void BackgroundSync::clearSyncTarget() {
boost::unique_lock<boost::mutex> lock(_mutex);
_syncSourceHost = HostAndPort();
}
void BackgroundSync::stop() {
boost::lock_guard<boost::mutex> lock(_mutex);
_pause = true;
_syncSourceHost = HostAndPort();
_lastOpTimeFetched = Timestamp(0,0);
_lastFetchedHash = 0;
_appliedBufferCondition.notify_all();
_pausedCondition.notify_all();
}
void BackgroundSync::start(OperationContext* txn) {
massert(16235, "going to start syncing, but buffer is not empty", _buffer.empty());
long long updatedLastAppliedHash = _readLastAppliedHash(txn);
boost::lock_guard<boost::mutex> lk(_mutex);
_pause = false;
// reset _last fields with current oplog data
_lastAppliedHash = updatedLastAppliedHash;
_lastOpTimeFetched = _replCoord->getMyLastOptime();
_lastFetchedHash = _lastAppliedHash;
LOG(1) << "bgsync fetch queue set to: " << _lastOpTimeFetched <<
" " << _lastFetchedHash;
}
void BackgroundSync::waitUntilPaused() {
boost::unique_lock<boost::mutex> lock(_mutex);
while (!_pause) {
_pausedCondition.wait(lock);
}
}
long long BackgroundSync::getLastAppliedHash() const {
boost::lock_guard<boost::mutex> lck(_mutex);
return _lastAppliedHash;
}
void BackgroundSync::setLastAppliedHash(long long newHash) {
boost::lock_guard<boost::mutex> lck(_mutex);
_lastAppliedHash = newHash;
}
void BackgroundSync::loadLastAppliedHash(OperationContext* txn) {
long long result = _readLastAppliedHash(txn);
boost::lock_guard<boost::mutex> lk(_mutex);
_lastAppliedHash = result;
}
long long BackgroundSync::_readLastAppliedHash(OperationContext* txn) {
BSONObj oplogEntry;
try {
ScopedTransaction transaction(txn, MODE_IX);
Lock::DBLock lk(txn->lockState(), "local", MODE_X);
bool success = Helpers::getLast(txn, rsOplogName.c_str(), oplogEntry);
if (!success) {
// This can happen when we are to do an initial sync. lastHash will be set
// after the initial sync is complete.
return 0;
}
}
catch (const DBException& ex) {
severe() << "Problem reading " << rsOplogName << ": " << ex.toStatus();
fassertFailed(18904);
}
BSONElement hashElement = oplogEntry[hashFieldName];
if (hashElement.eoo()) {
severe() << "Most recent entry in " << rsOplogName << " missing \"" << hashFieldName <<
"\" field";
fassertFailed(18902);
}
if (hashElement.type() != NumberLong) {
severe() << "Expected type of \"" << hashFieldName << "\" in most recent " <<
rsOplogName << " entry to have type NumberLong, but found " <<
typeName(hashElement.type());
fassertFailed(18903);
}
return hashElement.safeNumberLong();
}
bool BackgroundSync::getInitialSyncRequestedFlag() {
boost::lock_guard<boost::mutex> lock(_initialSyncMutex);
return _initialSyncRequestedFlag;
}
void BackgroundSync::setInitialSyncRequestedFlag(bool value) {
boost::lock_guard<boost::mutex> lock(_initialSyncMutex);
_initialSyncRequestedFlag = value;
}
} // namespace repl
} // namespace mongo
|