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
|
// /db/repl/rs.h
/**
* Copyright (C) 2008 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.
*/
#pragma once
#include "mongo/db/commands.h"
#include "mongo/db/index.h"
#include "mongo/db/repl/oplogreader.h"
#include "mongo/bson/optime.h"
#include "mongo/db/repl/rs_config.h"
#include "mongo/db/repl/rs_exception.h"
#include "mongo/db/repl/rs_member.h"
#include "mongo/db/repl/rs_sync.h"
#include "mongo/db/repl/sync_source_feedback.h"
#include "mongo/util/concurrency/list.h"
#include "mongo/util/concurrency/msg.h"
#include "mongo/util/concurrency/thread_pool.h"
#include "mongo/util/concurrency/value.h"
#include "mongo/util/net/hostandport.h"
/**
* Order of Events
*
* On startup, if the --replSet option is present, startReplSets is called.
* startReplSets forks off a new thread for replica set activities. It creates
* the global theReplSet variable and calls go() on it.
*
* theReplSet's constructor changes the replica set's state to RS_STARTUP,
* starts the replica set manager, and loads the config (if the replica set
* has been initialized).
*/
namespace mongo {
class Cloner;
class DBClientConnection;
struct HowToFixUp;
class ReplSetImpl;
struct Target;
extern bool replSet; // true if using repl sets
extern class ReplSet *theReplSet; // null until initialized
extern Tee *rsLog;
class ReplSetCmdline;
// Main entry point for replica sets
void startReplSets(ReplSetCmdline *replSetCmdline);
class ReplicationStartSynchronizer {
public:
ReplicationStartSynchronizer() : indexRebuildDone(false) {}
boost::mutex mtx;
bool indexRebuildDone;
boost::condition cond;
};
/* member of a replica set */
class Member : public List1<Member>::Base {
private:
~Member(); // intentionally unimplemented as should never be called -- see List1<>::Base.
Member(const Member&);
public:
Member(HostAndPort h, unsigned ord, const ReplSetConfig::MemberCfg *c, bool self);
string fullName() const { return h().toString(); }
const ReplSetConfig::MemberCfg& config() const { return _config; }
ReplSetConfig::MemberCfg& configw() { return _config; }
const HeartbeatInfo& hbinfo() const { return _hbinfo; }
HeartbeatInfo& get_hbinfo() { return _hbinfo; }
string lhb() const { return _hbinfo.lastHeartbeatMsg; }
MemberState state() const { return _hbinfo.hbstate; }
const HostAndPort& h() const { return _h; }
unsigned id() const { return _hbinfo.id(); }
bool potentiallyHot() const { return _config.potentiallyHot(); } // not arbiter, not priority 0
void summarizeMember(stringstream& s) const;
// If we could sync from this member. This doesn't tell us anything about the quality of
// this member, just if they are a possible sync target.
bool syncable() const;
private:
friend class ReplSetImpl;
ReplSetConfig::MemberCfg _config;
const HostAndPort _h;
HeartbeatInfo _hbinfo;
};
class Manager : public task::Server {
ReplSetImpl *rs;
bool busyWithElectSelf;
int _primary;
/** @param two - if true two primaries were seen. this can happen transiently, in addition to our
polling being only occasional. in this case null is returned, but the caller should
not assume primary itself in that situation.
*/
const Member* findOtherPrimary(bool& two);
void noteARemoteIsPrimary(const Member *);
void checkElectableSet();
void checkAuth();
virtual void starting();
public:
Manager(ReplSetImpl *rs);
virtual ~Manager();
void msgReceivedNewConfig(BSONObj);
void msgCheckNewState();
};
class GhostSync : public task::Server {
struct GhostSlave : boost::noncopyable {
GhostSlave() : last(0), slave(0), init(false) { }
OplogReader reader;
OpTime last;
Member* slave;
bool init;
};
/**
* This is a cache of ghost slaves
*/
typedef map< mongo::OID,shared_ptr<GhostSlave> > MAP;
MAP _ghostCache;
RWLock _lock; // protects _ghostCache
ReplSetImpl *rs;
virtual void starting();
public:
GhostSync(ReplSetImpl *_rs) : task::Server("rsGhostSync"), _lock("GhostSync"), rs(_rs) {}
~GhostSync() {
log() << "~GhostSync() called" << rsLog;
}
/**
* Replica sets can sync in a hierarchical fashion, which throws off w
* calculation on the master. percolate() faux-syncs from an upstream
* node so that the primary will know what the slaves are up to.
*
* We can't just directly sync to the primary because it could be
* unreachable, e.g., S1--->S2--->S3--->P. S2 should ghost sync from S3
* and S3 can ghost sync from the primary.
*
* Say we have an S1--->S2--->P situation and this node is S2. rid
* would refer to S1. S2 would create a ghost slave of S1 and connect
* it to P (_currentSyncTarget). Then it would use this connection to
* pretend to be S1, replicating off of P.
*/
void percolate(const mongo::OID& rid, const OpTime& last);
void associateSlave(const BSONObj& rid, const int memberId);
void updateSlave(const mongo::OID& id, const OpTime& last);
void clearCache();
};
class Consensus {
ReplSetImpl &rs;
struct LastYea {
LastYea() : when(0), who(0xffffffff) { }
time_t when;
unsigned who;
};
static SimpleMutex lyMutex;
Guarded<LastYea,lyMutex> ly;
unsigned yea(unsigned memberId); // throws VoteException
void electionFailed(unsigned meid);
void _electSelf();
bool weAreFreshest(bool& allUp, int& nTies);
bool sleptLast; // slept last elect() pass
public:
Consensus(ReplSetImpl *t) : rs(*t) {
sleptLast = false;
steppedDown = 0;
}
/* if we've stepped down, this is when we are allowed to try to elect ourself again.
todo: handle possible weirdnesses at clock skews etc.
*/
time_t steppedDown;
int totalVotes() const;
bool aMajoritySeemsToBeUp() const;
bool shouldRelinquish() const;
void electSelf();
void electCmdReceived(BSONObj, BSONObjBuilder*);
void multiCommand(BSONObj cmd, list<Target>& L);
};
/**
* most operations on a ReplSet object should be done while locked. that
* logic implemented here.
*
* Order of locking: lock the replica set, then take a rwlock.
*/
class RSBase : boost::noncopyable {
public:
const unsigned magic;
void assertValid() { verify( magic == 0x12345677 ); }
private:
mongo::mutex m;
int _locked;
ThreadLocalValue<bool> _lockedByMe;
protected:
RSBase() : magic(0x12345677), m("RSBase"), _locked(0) { }
~RSBase() {
/* this can happen if we throw in the constructor; otherwise never happens. thus we log it as it is quite unusual. */
log() << "replSet ~RSBase called" << rsLog;
}
public:
class lock {
RSBase& rsbase;
auto_ptr<scoped_lock> sl;
public:
lock(RSBase* b) : rsbase(*b) {
if( rsbase._lockedByMe.get() )
return; // recursive is ok...
sl.reset( new scoped_lock(rsbase.m) );
DEV verify(rsbase._locked == 0);
rsbase._locked++;
rsbase._lockedByMe.set(true);
}
~lock() {
if( sl.get() ) {
verify( rsbase._lockedByMe.get() );
DEV verify(rsbase._locked == 1);
rsbase._lockedByMe.set(false);
rsbase._locked--;
}
}
};
/* for asserts */
bool locked() const { return _locked != 0; }
/* if true, is locked, and was locked by this thread. note if false, it could be in the lock or not for another
just for asserts & such so we can make the contracts clear on who locks what when.
we don't use these locks that frequently, so the little bit of overhead is fine.
*/
bool lockedByMe() { return _lockedByMe.get(); }
};
class ReplSetHealthPollTask;
/* safe container for our state that keeps member pointer and state variables always aligned */
class StateBox : boost::noncopyable {
public:
struct SP { // SP is like pair<MemberState,const Member *> but nicer
SP() : state(MemberState::RS_STARTUP), primary(0) { }
MemberState state;
const Member *primary;
};
const SP get() {
rwlock lk(m, false);
return sp;
}
MemberState getState() const {
rwlock lk(m, false);
return sp.state;
}
const Member* getPrimary() const {
rwlock lk(m, false);
return sp.primary;
}
void change(MemberState s, const Member *self) {
rwlock lk(m, true);
if( sp.state != s ) {
log() << "replSet " << s.toString() << rsLog;
}
sp.state = s;
if( s.primary() ) {
sp.primary = self;
}
else {
if( self == sp.primary )
sp.primary = 0;
}
}
void set(MemberState s, const Member *p) {
rwlock lk(m, true);
sp.state = s;
sp.primary = p;
}
void setSelfPrimary(const Member *self) { change(MemberState::RS_PRIMARY, self); }
void setOtherPrimary(const Member *mem) {
rwlock lk(m, true);
verify( !sp.state.primary() );
sp.primary = mem;
}
void noteRemoteIsPrimary(const Member *remote) {
rwlock lk(m, true);
verify(!sp.state.primary());
sp.primary = remote;
}
StateBox() : m("StateBox") { }
private:
RWLock m;
SP sp;
};
void parseReplsetCmdLine(const std::string& cfgString,
string& setname,
vector<HostAndPort>& seeds,
set<HostAndPort>& seedSet);
/** Parameter given to the --replSet command line option (parsed).
Syntax is "<setname>/<seedhost1>,<seedhost2>"
where setname is a name and seedhost is "<host>[:<port>]" */
class ReplSetCmdline {
public:
ReplSetCmdline(const std::string& cfgString) { parseReplsetCmdLine(cfgString, setname, seeds, seedSet); }
string setname;
vector<HostAndPort> seeds;
set<HostAndPort> seedSet;
};
/* information about the entire repl set, such as the various servers in the set, and their state */
/* note: We currently do not free mem when the set goes away - it is assumed the replset is a
singleton and long lived.
*/
class ReplSetImpl : protected RSBase {
public:
/** info on our state if the replset isn't yet "up". for example, if we are pre-initiation. */
enum StartupStatus {
PRESTART=0, LOADINGCONFIG=1, BADCONFIG=2, EMPTYCONFIG=3,
EMPTYUNREACHABLE=4, STARTED=5, SOON=6
};
static StartupStatus startupStatus;
static DiagStr startupStatusMsg;
static string stateAsHtml(MemberState state);
static ReplicationStartSynchronizer rss;
/* todo thread */
void msgUpdateHBInfo(HeartbeatInfo);
/**
* Updates the lastHeartbeatRecv of Member with the given id.
*/
void msgUpdateHBRecv(unsigned id, time_t newTime);
StateBox box;
SyncSourceFeedback syncSourceFeedback;
OpTime lastOpTimeWritten;
long long lastH; // hash we use to make sure we are reading the right flow of ops and aren't on an out-of-date "fork"
bool forceSyncFrom(const string& host, string& errmsg, BSONObjBuilder& result);
// Check if the current sync target is suboptimal. This must be called while holding a mutex
// that prevents the sync source from changing.
bool shouldChangeSyncTarget(const OpTime& target) const;
/**
* Find the closest member (using ping time) with a higher latest optime.
*/
const Member* getMemberToSyncTo();
void veto(const string& host, unsigned secs=10);
bool gotForceSync();
void goStale(const Member* m, const BSONObj& o);
private:
set<ReplSetHealthPollTask*> healthTasks;
void endOldHealthTasks();
void startHealthTaskFor(Member *m);
Consensus elect;
void relinquish();
void forgetPrimary();
protected:
bool _stepDown(int secs);
bool _freeze(int secs);
private:
void assumePrimary();
void loadLastOpTimeWritten(bool quiet=false);
void changeState(MemberState s);
Member* _forceSyncTarget;
bool _blockSync;
void blockSync(bool block);
// set of electable members' _ids
set<unsigned> _electableSet;
protected:
// "heartbeat message"
// sent in requestHeartbeat respond in field "hbm"
char _hbmsg[256]; // we change this unlocked, thus not an stl::string
time_t _hbmsgTime; // when it was logged
public:
void sethbmsg(const std::string& s, int logLevel = 0);
/**
* Election with Priorities
*
* Each node (n) keeps a set of nodes that could be elected primary.
* Each node in this set:
*
* 1. can connect to a majority of the set
* 2. has a priority greater than 0
* 3. has an optime within 10 seconds of the most up-to-date node
* that n can reach
*
* If a node fails to meet one or more of these criteria, it is removed
* from the list. This list is updated whenever the node receives a
* heartbeat.
*
* When a node sends an "am I freshest?" query, the node receiving the
* query checks their electable list to make sure that no one else is
* electable AND higher priority. If this check passes, the node will
* return an "ok" response, if not, it will veto.
*
* If a node is primary and there is another node with higher priority
* on the electable list (i.e., it must be synced to within 10 seconds
* of the current primary), the node (or nodes) with connections to both
* the primary and the secondary with higher priority will issue
* replSetStepDown requests to the primary to allow the higher-priority
* node to take over.
*/
void addToElectable(const unsigned m) { lock lk(this); _electableSet.insert(m); }
void rmFromElectable(const unsigned m) { lock lk(this); _electableSet.erase(m); }
bool iAmElectable() { lock lk(this); return _electableSet.find(_self->id()) != _electableSet.end(); }
bool isElectable(const unsigned id) { lock lk(this); return _electableSet.find(id) != _electableSet.end(); }
Member* getMostElectable();
protected:
/**
* Load a new config as the replica set's main config.
*
* If there is a "simple" change (just adding a node), this shortcuts
* the config. Returns true if the config was changed. Returns false
* if the config doesn't include a this node. Throws an exception if
* something goes very wrong.
*
* Behavior to note:
* - locks this
* - intentionally leaks the old _cfg and any old _members (if the
* change isn't strictly additive)
*/
bool initFromConfig(ReplSetConfig& c, bool reconf=false);
void _fillIsMaster(BSONObjBuilder&);
void _fillIsMasterHost(const Member*, vector<string>&, vector<string>&, vector<string>&);
const ReplSetConfig& config() { return *_cfg; }
string name() const { return _name; } /* @return replica set's logical name */
int version() const { return _cfg->version; } /* @return replica set's config version */
MemberState state() const { return box.getState(); }
void _fatal();
void _getOplogDiagsAsHtml(unsigned server_id, stringstream& ss) const;
void _summarizeAsHtml(stringstream&) const;
void _summarizeStatus(BSONObjBuilder&) const; // for replSetGetStatus command
/* call afer constructing to start - returns fairly quickly after launching its threads */
void _go();
private:
string _name;
const vector<HostAndPort> *_seeds;
ReplSetConfig *_cfg;
/**
* Finds the configuration with the highest version number and attempts
* load it.
*/
bool _loadConfigFinish(vector<ReplSetConfig*>& v);
/**
* Gather all possible configs (from command line seeds, our own config
* doc, and any hosts listed therein) and try to initiate from the most
* recent config we find.
*/
void loadConfig();
list<HostAndPort> memberHostnames() const;
bool iAmArbiterOnly() const { return myConfig().arbiterOnly; }
bool iAmPotentiallyHot() const {
return myConfig().potentiallyHot() && // not an arbiter
elect.steppedDown <= time(0) && // not stepped down/frozen
state() == MemberState::RS_SECONDARY; // not stale
}
protected:
Member *_self;
bool _buildIndexes; // = _self->config().buildIndexes
ReplSetImpl();
/* throws exception if a problem initializing. */
void init(ReplSetCmdline&);
void setSelfTo(Member *); // use this as it sets buildIndexes var
private:
List1<Member> _members; // all members of the set EXCEPT _self.
ReplSetConfig::MemberCfg _config; // config of _self
unsigned _id; // _id of _self
int _maintenanceMode; // if we should stay in recovering state
public:
// this is called from within a writelock in logOpRS
unsigned selfId() const { return _id; }
Manager *mgr;
GhostSync *ghost;
/**
* This forces a secondary to go into recovering state and stay there
* until this is called again, passing in "false". Multiple threads can
* call this and it will leave maintenance mode once all of the callers
* have called it again, passing in false.
*/
bool setMaintenanceMode(const bool inc);
// Records a new slave's id in the GhostSlave map, at handshake time.
void registerSlave(const BSONObj& rid, const int memberId);
private:
Member* head() const { return _members.head(); }
public:
const Member* findById(unsigned id) const;
Member* getMutableMember(unsigned id);
Member* findByName(const std::string& hostname) const;
private:
void _getTargets(list<Target>&, int &configVersion);
void getTargets(list<Target>&, int &configVersion);
void startThreads();
friend class FeedbackThread;
friend class CmdReplSetElect;
friend class Member;
friend class Manager;
friend class GhostSync;
friend class Consensus;
private:
bool _syncDoInitialSync_clone(Cloner &cloner, const char *master,
const list<string>& dbs, bool dataPass);
bool _syncDoInitialSync_applyToHead( replset::SyncTail& syncer, OplogReader* r ,
const Member* source, const BSONObj& lastOp,
BSONObj& minValidOut);
void _syncDoInitialSync();
void syncDoInitialSync();
void _syncThread();
void syncTail();
unsigned _syncRollback(OplogReader& r);
void syncFixUp(HowToFixUp& h, OplogReader& r);
// keep a list of hosts that we've tried recently that didn't work
map<string,time_t> _veto;
// persistent pool of worker threads for writing ops to the databases
threadpool::ThreadPool _writerPool;
// persistent pool of worker threads for prefetching
threadpool::ThreadPool _prefetcherPool;
public:
// Allow index prefetching to be turned on/off
enum IndexPrefetchConfig {
PREFETCH_NONE=0, PREFETCH_ID_ONLY=1, PREFETCH_ALL=2
};
void setIndexPrefetchConfig(const IndexPrefetchConfig cfg) {
_indexPrefetchConfig = cfg;
}
IndexPrefetchConfig getIndexPrefetchConfig() {
return _indexPrefetchConfig;
}
static const int replWriterThreadCount;
static const int replPrefetcherThreadCount;
threadpool::ThreadPool& getPrefetchPool() { return _prefetcherPool; }
threadpool::ThreadPool& getWriterPool() { return _writerPool; }
static const int maxSyncSourceLagSecs;
const ReplSetConfig::MemberCfg& myConfig() const { return _config; }
bool tryToGoLiveAsASecondary(OpTime&); // readlocks
void syncRollback(OplogReader& r);
void syncThread();
const OpTime lastOtherOpTime() const;
/**
* The most up to date electable replica
*/
const OpTime lastOtherElectableOpTime() const;
/**
* When a member reaches its minValid optime it is in a consistent state. Thus, minValid is
* set as the last step in initial sync. At the beginning of initial sync, _initialSyncFlag
* is appended onto minValid to indicate that initial sync was started but has not yet
* completed.
* minValid is also used during "normal" sync: the last op in each batch is used to set
* minValid, to indicate that we are in a consistent state when the batch has been fully
* applied.
*/
static void setMinValid(BSONObj obj);
static OpTime getMinValid();
static void clearInitialSyncFlag();
static bool getInitialSyncFlag();
static void setInitialSyncFlag();
int oplogVersion;
private:
IndexPrefetchConfig _indexPrefetchConfig;
static const char* _initialSyncFlagString;
static const BSONObj _initialSyncFlag;
};
class ReplSet : public ReplSetImpl {
public:
static ReplSet* make(ReplSetCmdline& replSetCmdline);
virtual ~ReplSet() {}
// for the replSetStepDown command
bool stepDown(int secs) { return _stepDown(secs); }
// for the replSetFreeze command
bool freeze(int secs) { return _freeze(secs); }
string selfFullName() {
verify( _self );
return _self->fullName();
}
virtual bool buildIndexes() const { return _buildIndexes; }
/* call after constructing to start - returns fairly quickly after launching its threads */
void go() { _go(); }
void shutdown();
void fatal() { _fatal(); }
virtual bool isPrimary() { return box.getState().primary(); }
virtual bool isSecondary() { return box.getState().secondary(); }
MemberState state() const { return ReplSetImpl::state(); }
string name() const { return ReplSetImpl::name(); }
virtual const ReplSetConfig& config() { return ReplSetImpl::config(); }
void getOplogDiagsAsHtml(unsigned server_id, stringstream& ss) const { _getOplogDiagsAsHtml(server_id,ss); }
void summarizeAsHtml(stringstream& ss) const { _summarizeAsHtml(ss); }
void summarizeStatus(BSONObjBuilder& b) const { _summarizeStatus(b); }
void fillIsMaster(BSONObjBuilder& b) { _fillIsMaster(b); }
threadpool::ThreadPool& getPrefetchPool() { return ReplSetImpl::getPrefetchPool(); }
threadpool::ThreadPool& getWriterPool() { return ReplSetImpl::getWriterPool(); }
/**
* We have a new config (reconfig) - apply it.
* @param comment write a no-op comment to the oplog about it. only
* makes sense if one is primary and initiating the reconf.
*
* The slaves are updated when they get a heartbeat indicating the new
* config. The comment is a no-op.
*/
void haveNewConfig(ReplSetConfig& c, bool comment);
/**
* Pointer assignment isn't necessarily atomic, so this needs to assure
* locking, even though we don't delete old configs.
*/
const ReplSetConfig& getConfig() { return config(); }
bool lockedByMe() { return RSBase::lockedByMe(); }
// heartbeat msg to send to others; descriptive diagnostic info
string hbmsg() const {
if( time(0)-_hbmsgTime > 120 ) return "";
return _hbmsg;
}
protected:
ReplSet();
};
/**
* Base class for repl set commands. Checks basic things such if we're in
* rs mode before the command does its real work.
*/
class ReplSetCommand : public Command {
protected:
ReplSetCommand(const char * s, bool show=false) : Command(s, show) { }
virtual bool slaveOk() const { return true; }
virtual bool adminOnly() const { return true; }
virtual bool logTheOp() { return false; }
virtual LockType locktype() const { return NONE; }
virtual void help( stringstream &help ) const { help << "internal"; }
bool check(string& errmsg, BSONObjBuilder& result) {
if( !replSet ) {
errmsg = "not running with --replSet";
if( cmdLine.configsvr ) {
result.append("info", "configsvr"); // for shell prompt
}
return false;
}
if( theReplSet == 0 ) {
result.append("startupStatus", ReplSet::startupStatus);
string s;
errmsg = ReplSet::startupStatusMsg.empty() ? "replset unknown error 2" : ReplSet::startupStatusMsg.get();
if( ReplSet::startupStatus == 3 )
result.append("info", "run rs.initiate(...) if not yet done for the set");
return false;
}
return true;
}
};
/**
* does local authentication
* directly authorizes against AuthenticationInfo
*/
void replLocalAuth();
/** inlines ----------------- */
inline Member::Member(HostAndPort h, unsigned ord, const ReplSetConfig::MemberCfg *c, bool self) :
_config(*c), _h(h), _hbinfo(ord) {
verify(c);
if( self )
_hbinfo.health = 1.0;
}
inline bool ignoreUniqueIndex(IndexDetails& idx) {
if (!idx.unique()) {
return false;
}
if (!theReplSet) {
return false;
}
// see SERVER-6671
MemberState ms = theReplSet->state();
if (! ((ms == MemberState::RS_STARTUP2) ||
(ms == MemberState::RS_RECOVERING) ||
(ms == MemberState::RS_ROLLBACK))) {
return false;
}
// 2 is the oldest oplog version where operations
// are fully idempotent.
if (theReplSet->oplogVersion < 2) {
return false;
}
// Never ignore _id index
if (idx.isIdIndex()) {
return false;
}
return true;
}
}
|