-
-
Notifications
You must be signed in to change notification settings - Fork 494
/
searchdha.cpp
4610 lines (3895 loc) · 123 KB
/
searchdha.cpp
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
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
//
// Copyright (c) 2001-2016, Andrew Aksyonoff
// Copyright (c) 2008-2016, Sphinx Technologies Inc
// Copyright (c) 2017-2018, Manticore Software LTD (http://manticoresearch.com)
// All rights reserved
//
// This program is free software; you can redistribute it and/or modify
// it under the terms of the GNU General Public License. You should have
// received a copy of the GPL license along with this program; if you
// did not, you can find it at http://www.gnu.org/
//
#include "sphinx.h"
#include "sphinxstd.h"
#include "sphinxrt.h"
#include "sphinxint.h"
#include <errno.h>
#include "searchdaemon.h"
#include "searchdha.h"
#include <utility>
#if !USE_WINDOWS
#include <netinet/in.h>
#include <netdb.h>
#else
#include <WinSock2.h>
#include <MSWSock.h>
#include <WS2tcpip.h>
#pragma comment(lib, "Ws2_32.lib")
#endif
#if HAVE_EVENTFD
#include <sys/eventfd.h>
#endif
#if HAVE_GETADDRINFO_A
#include <signal.h>
#endif
int g_iPingInterval = 0; // by default ping HA agents every 1 second
DWORD g_uHAPeriodKarma = 60; // by default use the last 1 minute statistic to determine the best HA agent
int g_iPersistentPoolSize = 0;
static auto& g_bShutdown = sphGetShutdown();
CSphString HostDesc_t::GetMyUrl() const
{
CSphString sName;
switch ( m_iFamily )
{
case AF_INET: sName.SetSprintf ( "%s:%u", m_sAddr.cstr(), m_iPort ); break;
case AF_UNIX: sName = m_sAddr; break;
}
return sName;
}
static DWORD g_uTimePrefix = 0;
void StartLogTime()
{
g_uTimePrefix = sphMicroTimer () / 1000;
}
void sphLogDebugTimeredPrefix ( const char* sPrefix, const char * sFmt, ... )
{
StringBuilder_c sMyPrefix;
sMyPrefix+=sPrefix;
sMyPrefix.Appendf ("[%04d] ", (int)(sphMicroTimer ()/1000-g_uTimePrefix) );
sMyPrefix+=sFmt;
va_list ap;
va_start ( ap, sFmt );
sphLogVa ( sMyPrefix.cstr(), ap, SPH_LOG_VERBOSE_DEBUG );
va_end ( ap );
}
#define VERBOSE_NETLOOP 0
#if VERBOSE_NETLOOP
#define sphLogDebugA( ... ) sphLogDebugTimeredPrefix ("A ", __VA_ARGS__)
#define sphLogDebugL( ... ) sphLogDebugTimeredPrefix ("L ", __VA_ARGS__)
#else
#if USE_WINDOWS
#pragma warning(disable:4390)
#endif
#define sphLogDebugA( ... )
#define sphLogDebugL( ... )
#endif
/////////////////////////////////////////////////////////////////////////////
// HostDashboard_t
//
// Dashboard contains generic statistic for a host like query time, answer time,
// num of errors a row, and all this data in retrospective on time
/////////////////////////////////////////////////////////////////////////////
HostDashboard_t::HostDashboard_t ( const HostDesc_t & tHost )
{
assert ( !tHost.m_pDash );
m_tHost.CloneFromHost ( tHost );
m_iLastQueryTime = m_iLastAnswerTime = sphMicroTimer () - g_iPingInterval*1000;
for ( auto & dStat : m_dStats )
dStat.m_dData.Reset();
}
HostDashboard_t::~HostDashboard_t ()
{
SafeDelete ( m_pPersPool );
}
bool HostDashboard_t::IsOlder ( int64_t iTime ) const
{
return ( (iTime-m_iLastAnswerTime)>g_iPingInterval*1000LL );
}
DWORD HostDashboard_t::GetCurSeconds()
{
int64_t iNow = sphMicroTimer()/1000000;
return DWORD ( iNow & 0xFFFFFFFF );
}
bool HostDashboard_t::IsHalfPeriodChanged ( DWORD * pLast )
{
assert ( pLast );
DWORD uSeconds = GetCurSeconds();
if ( ( uSeconds - *pLast )>( g_uHAPeriodKarma / 2 ) )
{
*pLast = uSeconds;
return true;
}
return false;
}
AgentDash_t& HostDashboard_t::GetCurrentStat ()
{
DWORD uCurrentPeriod = GetCurSeconds()/g_uHAPeriodKarma;
auto & dCurrentStats = m_dStats[uCurrentPeriod % STATS_DASH_PERIODS];
if ( dCurrentStats.m_uPeriod!=uCurrentPeriod ) // we have new or reused stat
{
dCurrentStats.m_dData.Reset ();
dCurrentStats.m_uPeriod = uCurrentPeriod;
}
return dCurrentStats.m_dData;
}
void HostDashboard_t::GetCollectedStat ( HostStatSnapshot_t& dResult, int iPeriods ) const
{
DWORD uSeconds = GetCurSeconds();
if ( (uSeconds % g_uHAPeriodKarma) < (g_uHAPeriodKarma/2) )
++iPeriods;
iPeriods = Min ( iPeriods, STATS_DASH_PERIODS );
DWORD uCurrentPeriod = uSeconds/g_uHAPeriodKarma;
AgentDash_t tAccum;
tAccum.Reset ();
CSphScopedRLock tRguard ( m_dDataLock );
for ( ; iPeriods>0; --iPeriods, --uCurrentPeriod )
// it might be no queries at all in the fixed time
if ( m_dStats[uCurrentPeriod % STATS_DASH_PERIODS].m_uPeriod==uCurrentPeriod )
tAccum.Add ( m_dStats[uCurrentPeriod % STATS_DASH_PERIODS].m_dData );
for ( int i = 0; i<eMaxAgentStat; ++i )
dResult[i] = tAccum.m_dCounters[i];
for ( int i = 0; i<ehMaxStat; ++i )
dResult[i+eMaxAgentStat] = tAccum.m_dMetrics[i];
}
/////////////////////////////////////////////////////////////////////////////
// PersistentConnectionsPool_c
//
// This is pool of sockets which can be rented/returned by workers.
// returned socket may be in connected state, so later usage will not require
// connection, but can work immediately (in persistent mode).
/////////////////////////////////////////////////////////////////////////////
// modify size of the pool
// (actually we just need to close any extra sockets not fit in new limit. And do nothing otherwize).
void PersistentConnectionsPool_c::ReInit ( int iPoolSize )
{
assert ( iPoolSize >0 );
// close any extra sockets which are not fit in new limit
ScopedMutex_t tGuard ( m_dDataLock );
m_iLimit = iPoolSize;
m_dSockets.Reserve ( m_iLimit );
while ( m_iFreeWindow>m_iLimit )
{
--m_iFreeWindow;
int iSock = m_dSockets[Step ( &m_iRit )];
if ( iSock>=0 )
sphSockClose ( iSock );
}
}
// move the iterator to the next item, or loop the ring.
int PersistentConnectionsPool_c::Step ( int* pVar )
{
assert ( pVar );
int iRes = *pVar++;
if ( *pVar>=m_dSockets.GetLength () )
*pVar = 0;
return iRes;
}
// Rent first try to return the sockets which already were in work (i.e. which are connected)
// If no such socket available and the limit is not reached, it will add the new one.
int PersistentConnectionsPool_c::RentConnection ()
{
ScopedMutex_t tGuard ( m_dDataLock );
if ( m_iFreeWindow>0 )
{
--m_iFreeWindow;
return m_dSockets[Step ( &m_iRit )];
}
if ( m_dSockets.GetLength () >= m_iLimit )
return -2; // no more slots available;
// this branch will be executed only during initial 'heating'
m_dSockets.Add ( -1 );
return -1;
}
// write given socket into the pool.
void PersistentConnectionsPool_c::ReturnConnection ( int iSocket )
{
ScopedMutex_t tGuard ( m_dDataLock );
// overloaded pool
if ( m_iFreeWindow >= m_dSockets.GetLength () )
{
// no place at all (i.e. if pool was resized, but some sockets rented before,
// and now they are returned, but we have no place for them)
if ( m_dSockets.GetLength () >= m_iLimit )
{
sphSockClose ( iSocket );
return;
}
// add the place for one more returned socket
m_dSockets.Add ();
m_iWit = m_dSockets.GetLength ()-1;
}
++m_iFreeWindow;
if ( m_bShutdown )
{
sphSockClose ( iSocket );
iSocket = -1;
}
// if there were no free sockets until now, point the reading iterator to just released socket
if ( m_iFreeWindow==1 )
m_iRit = m_iWit;
m_dSockets[Step ( &m_iWit )] = iSocket;
}
// close all the sockets in the pool.
void PersistentConnectionsPool_c::Shutdown ()
{
ScopedMutex_t tGuard ( m_dDataLock );
m_bShutdown = true;
for ( int i = 0; i<m_iFreeWindow; ++i )
{
int& iSock = m_dSockets[Step ( &m_iRit )];
if ( iSock>=0 )
{
sphSockClose ( iSock );
iSock = -1;
}
}
}
void ClosePersistentSockets()
{
VecRefPtrs_t<HostDashboard_t *> dHosts;
g_tDashes.GetActiveDashes ( dHosts );
dHosts.Apply ( [] ( HostDashboard_t * &pHost ) { SafeDelete ( pHost->m_pPersPool ); } );
}
// check whether sURL contains plain ip-address, and so, m.b. no need to resolve it many times.
// ip-address assumed as string from four .-separated positive integers, each in range 0..255
static bool IsIpAddress ( const char * sURL )
{
StrVec_t dParts;
sphSplit ( dParts, sURL, "." );
if ( dParts.GetLength ()!=4 )
return false;
for ( auto &dPart : dParts )
{
dPart.Trim ();
auto iLen = dPart.Length ();
for ( auto i = 0; i<iLen; ++i )
if ( !isdigit ( dPart.cstr ()[i] ) )
return false;
auto iRes = atoi ( dPart.cstr () );
if ( iRes<0 || iRes>255 )
return false;
}
return true;
}
/// Set flag m_bNeedResolve if address is AF_INET, host is not empty and not plain IP address,
/// and also if global flag for postponed resolving is set.
/// Otherwise resolve address now (if appliable) and then forbid any name resolving in future.
bool ResolveAddress ( AgentDesc_t &tAgent, const WarnInfo_t &tInfo )
{
tAgent.m_bNeedResolve = false;
if ( tAgent.m_iFamily!=AF_INET )
return true;
if ( tAgent.m_sAddr.IsEmpty () )
return tInfo.ErrSkip ( "invalid host name 'empty'" );
if ( IsIpAddress ( tAgent.m_sAddr.cstr () ) )
{
tAgent.m_uAddr = sphGetAddress ( tAgent.m_sAddr.cstr (), false, true );
if ( tAgent.m_uAddr )
return true;
// got plain ip and couldn't resolve it? Hm, strange...
if ( !g_bHostnameLookup )
return tInfo.ErrSkip ( "failed to lookup host name '%s' (error=%s)", tAgent.m_sAddr.cstr ()
, sphSockError () );
}
// if global flag set, don't resolve right now, it will be done on place.
tAgent.m_bNeedResolve = g_bHostnameLookup;
if ( tAgent.m_bNeedResolve )
return true;
tAgent.m_uAddr = sphGetAddress ( tAgent.m_sAddr.cstr () );
if ( tAgent.m_uAddr )
return true;
return tInfo.ErrSkip ( "failed to lookup host name '%s' (error=%s)", tAgent.m_sAddr.cstr (), sphSockError () );
}
/// Async resolving
using CallBack_f = std::function<void ( DWORD )>;
class DNSResolver_c
{
struct addrinfo m_tHints { 0 };
CallBack_f m_pCallback = nullptr;
bool m_bCallbackInvoked = false;
CSphString m_sHost;
DNSResolver_c ( const char * sHost, CallBack_f &&pFunc )
: m_pCallback ( std::move ( pFunc ) )
, m_sHost ( sHost )
{
m_tHints.ai_family = AF_INET;
m_tHints.ai_socktype = SOCK_STREAM;
}
~DNSResolver_c ()
{
assert ( m_pCallback );
if ( !m_bCallbackInvoked )
m_pCallback ( 0 );
};
void FinishResolve ( struct addrinfo * pResult )
{
assert ( m_pCallback );
if ( !pResult )
return;
auto * pSockaddr_ipv4 = ( struct sockaddr_in * ) pResult->ai_addr;
auto uIP = pSockaddr_ipv4->sin_addr.s_addr;
freeaddrinfo ( pResult );
m_pCallback ( uIP );
m_bCallbackInvoked = true;
};
static void ResolvingRoutine ( void * pResolver )
{
auto * pThis = ( DNSResolver_c * ) pResolver;
if ( pThis )
pThis->Resolve ();
SafeDelete ( pThis );
}
// platform-specific part starts here.
#if HAVE_GETADDRINFO_A
struct gaicb m_dReq, *m_pReq;
static void ResolvingWrapper ( sigval dSigval )
{
ResolvingRoutine ( dSigval.sival_ptr );
};
void ResolveBegin ()
{
m_dReq = { 0 };
m_dReq.ar_name = m_sHost.cstr ();
m_dReq.ar_request = &m_tHints;
m_pReq = &m_dReq;
sigevent_t dCallBack = {{ 0 }};
dCallBack.sigev_notify = SIGEV_THREAD;
dCallBack.sigev_value.sival_ptr = this;
dCallBack.sigev_notify_function = ResolvingWrapper;
getaddrinfo_a ( GAI_NOWAIT, &m_pReq, 1, &dCallBack );
}
void Resolve ()
{
auto * pResult = m_dReq.ar_result;
FinishResolve ( pResult );
}
#else
SphThread_t m_dResolverThread;
void ResolveBegin ()
{
sphThreadCreate ( &m_dResolverThread, ResolvingRoutine, this, true );
}
void Resolve ()
{
struct addrinfo * pResult = nullptr;
if ( !getaddrinfo ( m_sHost.cstr (), nullptr, &m_tHints, &pResult ) )
FinishResolve ( pResult );
}
#endif
public:
//! Non-blocking DNS resolver
//! Will call provided callback with single DWORD parameter - resolved IP. (or 0 on fail)
//! If there is no need to go background (i.e. we have plain ip address which not need to be
//! resolved any complex way), will call the callback immediately.
//! \param sHost - host address, as 'www.google.com', or '127.0.0.1'
//! \param pFunc - callback func. May be simple functor, or even lambda with acquired resources.
static void GetAddress_a ( const char * sHost, CallBack_f && pFunc )
{
if ( IsIpAddress ( sHost ) )
{
DWORD uResult = sphGetAddress ( sHost, false, true );
if ( uResult )
{
if ( pFunc )
pFunc ( uResult );
return;
}
}
auto * pPayload = new DNSResolver_c ( sHost, std::forward<CallBack_f>(pFunc) );
pPayload->ResolveBegin();
}
};
/// initialize address resolving,
/// add to agent records of stats,
/// add agent into global dashboard hash
static bool ValidateAndAddDashboard ( AgentDesc_t& dAgent, const WarnInfo_t &tInfo )
{
assert ( !dAgent.m_pDash && !dAgent.m_pStats );
if ( !ResolveAddress ( dAgent, tInfo ) )
return false;
g_tDashes.LinkHost ( dAgent );
dAgent.m_pStats = new AgentDash_t;
assert ( dAgent.m_pDash );
return true;
}
/////////////////////////////////////////////////////////////////////////////
// MultiAgentDesc_c
//
// That is set of hosts serving one and same agent (i.e., 'mirrors').
// class also provides mirror choosing using different strategies
/////////////////////////////////////////////////////////////////////////////
GuardedHash_c & g_MultiAgents()
{
static GuardedHash_c dGlobalHash;
return dGlobalHash;
}
void MultiAgentDesc_c::CleanupOrphaned()
{
// cleanup global
auto &gAgents = g_MultiAgents ();
for ( WLockedHashIt_c it ( &gAgents ); it.Next (); )
{
auto pAgent = it.Get ();
if ( pAgent )
{
pAgent->Release (); // need release since it.Get() just made AddRef().
if ( pAgent->IsLast () )
{
it.Delete ();
SafeRelease ( pAgent );
}
}
}
}
// calculate uniq key for holding MultiAgent instance in global hash
CSphString MultiAgentDesc_c::GetKey ( const CSphVector<AgentDesc_t *> &dTemplateHosts, const AgentOptions_t &tOpt )
{
StringBuilder_c sKey;
for ( const auto* dHost : dTemplateHosts )
{
sKey+=dHost->GetMyUrl ().cstr();
sKey+=":";
sKey+=dHost->m_sIndexes.cstr();
sKey+="|";
}
sKey.Appendf ("[%d,%d,%d,%d,%d]",
tOpt.m_bBlackhole?1:0,
tOpt.m_bPersistent?1:0,
(int)tOpt.m_eStrategy,
tOpt.m_iRetryCount,
tOpt.m_iRetryCountMultiplier);
return sKey.cstr();
}
MultiAgentDesc_c* MultiAgentDesc_c::GetAgent ( const CSphVector<AgentDesc_t*> &dHosts, const AgentOptions_t &tOpt,
const WarnInfo_t &tWarn ) NO_THREAD_SAFETY_ANALYSIS
{
auto sKey = GetKey ( dHosts, tOpt );
auto &gHash = g_MultiAgents ();
// if an elem exists, return it addreffed.
MultiAgentDescRefPtr_c pAgent ( ( MultiAgentDesc_c * ) gHash.Get ( sKey ) );
if ( pAgent )
return pAgent.Leak();
// create and init new agent
pAgent = new MultiAgentDesc_c;
if ( !pAgent->Init ( dHosts, tOpt, tWarn ) )
return nullptr;
return ( MultiAgentDesc_c * ) gHash.TryAddThenGet ( pAgent, sKey );
}
bool MultiAgentDesc_c::Init ( const CSphVector<AgentDesc_t *> &dHosts,
const AgentOptions_t &tOpt, const WarnInfo_t &tWarn ) NO_THREAD_SAFETY_ANALYSIS
{
// initialize options
m_eStrategy = tOpt.m_eStrategy;
m_iMultiRetryCount = tOpt.m_iRetryCount * tOpt.m_iRetryCountMultiplier;
// initialize hosts & weights
auto iLen = dHosts.GetLength ();
Reset ( iLen );
m_dWeights.Reset ( iLen );
if ( !iLen )
return tWarn.ErrSkip ( "Unable to initialize empty agent" );
auto fFrac = 100.0f / iLen;
ARRAY_FOREACH ( i, dHosts )
{
// we have templates parsed from config, NOT real working hosts!
assert ( !dHosts[i]->m_pDash && !dHosts[i]->m_pStats );
if ( !ValidateAndAddDashboard ( ( m_pData + i )->CloneFrom ( *dHosts[i] ), tWarn ) )
return false;
m_dWeights[i] = fFrac;
}
// require pings for HA agent
if ( IsHA () )
for ( int i = 0; i<GetLength (); ++i )
m_pData[i].m_pDash->m_bNeedPing = true;
return true;
}
const AgentDesc_t &MultiAgentDesc_c::RRAgent ()
{
if ( !IsHA() )
return *m_pData;
auto iRRCounter = (int) m_iRRCounter++;
while ( iRRCounter<0 || iRRCounter> ( GetLength ()-1 ) )
{
if ( iRRCounter+1 == (int) m_iRRCounter.CAS ( iRRCounter+1, 1 ) )
iRRCounter = 0;
else
iRRCounter = (int) m_iRRCounter++;
}
return m_pData[iRRCounter];
}
const AgentDesc_t &MultiAgentDesc_c::RandAgent ()
{
return m_pData[sphRand () % GetLength ()];
}
void MultiAgentDesc_c::ChooseWeightedRandAgent ( int * pBestAgent, CSphVector<int> & dCandidates )
{
assert ( pBestAgent );
CSphScopedRLock tLock ( m_dWeightLock );
auto fBound = m_dWeights[*pBestAgent];
auto fLimit = fBound;
for ( auto j : dCandidates )
fLimit += m_dWeights[j];
auto fChance = sphRand () * fLimit / UINT_MAX;
if ( fChance<=fBound )
return;
for ( auto j : dCandidates )
{
fBound += m_dWeights[j];
*pBestAgent = j;
if ( fChance<=fBound )
break;
}
}
static void LogAgentWeights ( const float * pOldWeights, const float * pCurWeights, const int64_t * pTimers, const CSphFixedVector<AgentDesc_t> & dAgents )
{
ARRAY_FOREACH ( i, dAgents )
sphLogDebug ( "client=%s, mirror=%d, weight=%0.2f%%, %0.2f%%, timer=" INT64_FMT
, dAgents[i].GetMyUrl ().cstr (), i, pCurWeights[i], pOldWeights[i], pTimers[i] );
}
const AgentDesc_t &MultiAgentDesc_c::StDiscardDead ()
{
if ( !IsHA() )
return m_pData[0];
// threshold errors-a-row to be counted as dead
int iDeadThr = 3;
int iBestAgent = -1;
int64_t iErrARow = -1;
CSphVector<int> dCandidates;
CSphFixedVector<int64_t> dTimers ( GetLength() );
dCandidates.Reserve ( GetLength() );
for (int i=0; i<GetLength(); ++i)
{
// no locks for g_pStats since we just reading, and read data is not critical.
const HostDashboard_t * pDash = m_pData[i].m_pDash;
HostStatSnapshot_t dDashStat;
pDash->GetCollectedStat (dDashStat);// look at last 30..90 seconds.
uint64_t uQueries = 0;
for ( int j=0; j<eMaxAgentStat; ++j )
uQueries += dDashStat[j];
if ( uQueries > 0 )
dTimers[i] = dDashStat[ehTotalMsecs]/uQueries;
else
dTimers[i] = 0;
CSphScopedRLock tRguard ( pDash->m_dDataLock );
int64_t iThisErrARow = ( pDash->m_iErrorsARow<=iDeadThr ) ? 0 : pDash->m_iErrorsARow;
if ( iErrARow < 0 )
iErrARow = iThisErrARow;
// 2. Among good nodes - select the one(s) with lowest errors/query rating
if ( iErrARow > iThisErrARow )
{
dCandidates.Reset();
iBestAgent = i;
iErrARow = iThisErrARow;
} else if ( iErrARow==iThisErrARow )
{
if ( iBestAgent>=0 )
dCandidates.Add ( iBestAgent );
iBestAgent = i;
}
}
// check if it is a time to recalculate the agent's weights
CheckRecalculateWeights ( dTimers );
// nothing to select, sorry. Just random agent...
if ( iBestAgent < 0 )
{
sphLogDebug ( "HA selector discarded all the candidates and just fall into simple Random" );
return RandAgent();
}
// only one node with lowest error rating. Return it.
if ( !dCandidates.GetLength() )
{
sphLogDebug ( "client=%s, HA selected %d node with best num of errors a row (" INT64_FMT ")"
, m_pData[iBestAgent].GetMyUrl().cstr(), iBestAgent, iErrARow );
return m_pData[iBestAgent];
}
// several nodes. Let's select the one.
ChooseWeightedRandAgent ( &iBestAgent, dCandidates );
if ( g_eLogLevel>=SPH_LOG_VERBOSE_DEBUG )
{
float fAge = 0.0;
const HostDashboard_t & dDash = *m_pData[iBestAgent].m_pDash;
CSphScopedRLock tRguard ( dDash.m_dDataLock );
fAge = ( dDash.m_iLastAnswerTime-dDash.m_iLastQueryTime ) / 1000.0f;
sphLogDebugv ("client=%s, HA selected %d node by weighted random, with best EaR ("
INT64_FMT "), last answered in %.3f milliseconds, among %d candidates"
, m_pData[iBestAgent].GetMyUrl ().cstr(), iBestAgent, iErrARow, fAge, dCandidates.GetLength()+1 );
}
return m_pData[iBestAgent];
}
// Check the time and recalculate mirror weights, if necessary.
void MultiAgentDesc_c::CheckRecalculateWeights ( const CSphFixedVector<int64_t> & dTimers )
{
if ( !dTimers.GetLength () || !HostDashboard_t::IsHalfPeriodChanged ( &m_uTimestamp ) )
return;
CSphFixedVector<float> dWeights {0};
// since we'll update values anyway, acquire w-lock.
CSphScopedWLock tWguard ( m_dWeightLock );
dWeights.CopyFrom ( m_dWeights );
RebalanceWeights ( dTimers, dWeights );
if ( g_eLogLevel>=SPH_LOG_DEBUG )
LogAgentWeights ( m_dWeights.Begin(), dWeights.Begin (), dTimers.Begin (), *this );
m_dWeights.SwapData (dWeights);
}
const AgentDesc_t &MultiAgentDesc_c::StLowErrors ()
{
if ( !IsHA() )
return *m_pData;
// how much error rating is allowed
float fAllowedErrorRating = 0.03f; // i.e. 3 errors per 100 queries is still ok
int iBestAgent = -1;
float fBestCriticalErrors = 1.0;
float fBestAllErrors = 1.0;
CSphVector<int> dCandidates;
CSphFixedVector<int64_t> dTimers ( GetLength() );
dCandidates.Reserve ( GetLength() );
for ( int i=0; i<GetLength(); ++i )
{
// no locks for g_pStats since we just reading, and read data is not critical.
const HostDashboard_t & dDash = *m_pData[i].m_pDash;
HostStatSnapshot_t dDashStat;
dDash.GetCollectedStat (dDashStat); // look at last 30..90 seconds.
uint64_t uQueries = 0;
uint64_t uCriticalErrors = 0;
uint64_t uAllErrors = 0;
uint64_t uSuccesses = 0;
for ( int j=0; j<eMaxAgentStat; ++j )
{
if ( j==eNetworkCritical )
uCriticalErrors = uQueries;
else if ( j==eNetworkNonCritical )
{
uAllErrors = uQueries;
uSuccesses = dDashStat[j];
}
uQueries += dDashStat[j];
}
if ( uQueries > 0 )
dTimers[i] = dDashStat[ehTotalMsecs]/uQueries;
else
dTimers[i] = 0;
// 1. No successes queries last period (it includes the pings). Skip such node!
if ( !uSuccesses )
continue;
if ( uQueries )
{
// 2. Among good nodes - select the one(s) with lowest errors/query rating
float fCriticalErrors = (float) uCriticalErrors/uQueries;
float fAllErrors = (float) uAllErrors/uQueries;
if ( fCriticalErrors<=fAllowedErrorRating )
fCriticalErrors = 0.0f;
if ( fAllErrors<=fAllowedErrorRating )
fAllErrors = 0.0f;
if ( fCriticalErrors < fBestCriticalErrors )
{
dCandidates.Reset();
iBestAgent = i;
fBestCriticalErrors = fCriticalErrors;
fBestAllErrors = fAllErrors;
} else if ( fCriticalErrors==fBestCriticalErrors )
{
if ( fAllErrors < fBestAllErrors )
{
dCandidates.Reset();
iBestAgent = i;
fBestAllErrors = fAllErrors;
} else if ( fAllErrors==fBestAllErrors )
{
if ( iBestAgent>=0 )
dCandidates.Add ( iBestAgent );
iBestAgent = i;
}
}
}
}
// check if it is a time to recalculate the agent's weights
CheckRecalculateWeights ( dTimers );
// nothing to select, sorry. Just plain RR...
if ( iBestAgent < 0 )
{
sphLogDebug ( "HA selector discarded all the candidates and just fall into simple Random" );
return RandAgent();
}
// only one node with lowest error rating. Return it.
if ( !dCandidates.GetLength() )
{
sphLogDebug ( "client=%s, HA selected %d node with best error rating (%.2f)",
m_pData[iBestAgent].GetMyUrl().cstr(), iBestAgent, fBestCriticalErrors );
return m_pData[iBestAgent];
}
// several nodes. Let's select the one.
ChooseWeightedRandAgent ( &iBestAgent, dCandidates );
if ( g_eLogLevel>=SPH_LOG_VERBOSE_DEBUG )
{
float fAge = 0.0f;
const HostDashboard_t & dDash = *m_pData[iBestAgent].m_pDash;
CSphScopedRLock tRguard ( dDash.m_dDataLock );
fAge = ( dDash.m_iLastAnswerTime-dDash.m_iLastQueryTime ) / 1000.0f;
sphLogDebugv (
"client=%s, HA selected %d node by weighted random, "
"with best error rating (%.2f), answered %f seconds ago"
, m_pData[iBestAgent].GetMyUrl ().cstr(), iBestAgent, fBestCriticalErrors, fAge );
}
return m_pData[iBestAgent];
}
const AgentDesc_t &MultiAgentDesc_c::ChooseAgent ()
{
if ( !IsHA() )
{
assert ( m_pData && "Not initialized MultiAgent detected!");
if ( m_pData )
return *m_pData;
static AgentDesc_t dFakeHost; // avoid crash in release if not initialized.
return dFakeHost;
}
switch ( m_eStrategy )
{
case HA_AVOIDDEAD:
return StDiscardDead();
case HA_AVOIDERRORS:
return StLowErrors();
case HA_ROUNDROBIN:
return RRAgent();
default:
return RandAgent();
}
}
const char * Agent_e_Name ( Agent_e eState )
{
switch ( eState )
{
case Agent_e::HEALTHY: return "HEALTHY";
case Agent_e::CONNECTING: return "CONNECTING";
case Agent_e::RETRY: return "RETRY";
}
assert ( 0 && "UNKNOWN_STATE" );
return "UNKNOWN_STATE";
}
SearchdStats_t g_tStats;
cDashStorage g_tDashes;
// generic stats track - always to agent stats, separately to dashboard.
void agent_stats_inc ( AgentConn_t &tAgent, AgentStats_e iCountID )
{
assert ( iCountID<=eMaxAgentStat );
assert ( tAgent.m_tDesc.m_pDash );
if ( tAgent.m_tDesc.m_pStats )
++tAgent.m_tDesc.m_pStats->m_dCounters[iCountID];
HostDashboard_t &tIndexDash = *tAgent.m_tDesc.m_pDash;
CSphScopedWLock tWguard ( tIndexDash.m_dDataLock );
AgentDash_t &tAgentDash = tIndexDash.GetCurrentStat ();
tAgentDash.m_dCounters[iCountID]++;
if ( iCountID>=eNetworkNonCritical && iCountID<eMaxAgentStat )
tIndexDash.m_iErrorsARow = 0;
else
tIndexDash.m_iErrorsARow++;
tAgent.m_iEndQuery = sphMicroTimer ();
tIndexDash.m_iLastQueryTime = tAgent.m_iStartQuery;
tIndexDash.m_iLastAnswerTime = tAgent.m_iEndQuery;
// do not count query time for unlinked connections (pings)
// only count errors
if ( tAgent.m_tDesc.m_pStats )
{
tAgentDash.m_dMetrics[ehTotalMsecs] += tAgent.m_iEndQuery - tAgent.m_iStartQuery;
tAgent.m_tDesc.m_pStats->m_dMetrics[ehTotalMsecs] += tAgent.m_iEndQuery - tAgent.m_iStartQuery;
}
}
// special case of stats - all is ok, just need to track the time in dashboard.
void track_processing_time ( AgentConn_t &tAgent )
{
// first we count temporary statistic (into dashboard)
assert ( tAgent.m_tDesc.m_pDash );
uint64_t uConnTime = ( uint64_t ) sphMicroTimer () - tAgent.m_iStartQuery;
{
CSphScopedWLock tWguard ( tAgent.m_tDesc.m_pDash->m_dDataLock );
uint64_t * pMetrics = tAgent.m_tDesc.m_pDash->GetCurrentStat ().m_dMetrics;
++pMetrics[ehConnTries];
if ( uint64_t ( uConnTime )>pMetrics[ehMaxMsecs] )
pMetrics[ehMaxMsecs] = uConnTime;
if ( pMetrics[ehConnTries]>1 )
pMetrics[ehAverageMsecs] =
( pMetrics[ehAverageMsecs] * ( pMetrics[ehConnTries] - 1 ) + uConnTime ) / pMetrics[ehConnTries];
else
pMetrics[ehAverageMsecs] = uConnTime;
} // no need to hold dashboard anymore
if ( !tAgent.m_tDesc.m_pStats )
return;
// then we count permanent statistic (for show status)
uint64_t * pHStat = tAgent.m_tDesc.m_pStats->m_dMetrics;
++pHStat[ehConnTries];
if ( uint64_t ( uConnTime )>pHStat[ehMaxMsecs] )
pHStat[ehMaxMsecs] = uConnTime;
if ( pHStat[ehConnTries]>1 )
pHStat[ehAverageMsecs] =
( pHStat[ehAverageMsecs] * ( pHStat[ehConnTries] - 1 ) + uConnTime ) / pHStat[ehConnTries];
else
pHStat[ehAverageMsecs] = uConnTime;
}
/// try to parse hostname/ip/port or unixsocket on current pConfigLine.
/// fill pAgent fields on success and move ppLine pointer next after parsed instance
/// test cases and test group 'T_ParseAddressPort', are in gtest_searchdaemon.cpp.
bool ParseAddressPort ( HostDesc_t& dHost, const char ** ppLine, const WarnInfo_t &dInfo )
{
// extract host name or path
const char *&p = *ppLine;
const char * pAnchor = p;
if ( !p )
return false;
enum AddressType_e { apIP, apUNIX };
AddressType_e eState = apIP;
if ( *p=='/' ) // whether we parse inet or unix socket
eState = apUNIX;
while ( sphIsAlpha ( *p ) || *p=='.' || *p=='-' || *p=='/' )
++p;
if ( p==pAnchor )
return dInfo.ErrSkip ( "host name or path expected" );
CSphString sSub ( pAnchor, p-pAnchor );
if ( eState==apUNIX )
{
#if !USE_WINDOWS
if ( strlen ( sSub.cstr () ) + 1>sizeof(((struct sockaddr_un *)0)->sun_path) )
return dInfo.ErrSkip ( "UNIX socket path is too long" );
#endif
dHost.m_iFamily = AF_UNIX;
dHost.m_sAddr = sSub;
return true;
}
// below is only deal with inet sockets
dHost.m_iFamily = AF_INET;
dHost.m_sAddr = sSub;
// expect ':' (and then portnum) after address
if ( *p!=':' )
{
dHost.m_iPort = IANA_PORT_SPHINXAPI;
dInfo.Warn ( "colon and portnum expected before '%s' - Using default IANA %d port", p, dHost.m_iPort );
return true;
}
pAnchor = ++p;
// parse portnum
while ( isdigit(*p) )
++p;
if ( p==pAnchor )
{
dHost.m_iPort = IANA_PORT_SPHINXAPI;
dInfo.Warn ( "portnum expected before '%s' - Using default IANA %d port", p, dHost.m_iPort );
--p; /// step back to ':'
return true;