-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathPG.h
1522 lines (1296 loc) · 44.4 KB
/
PG.h
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
// -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
// vim: ts=8 sw=2 smarttab
/*
* Ceph - scalable distributed file system
*
* Copyright (C) 2004-2006 Sage Weil <[email protected]>
*
* This is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License version 2.1, as published by the Free Software
* Foundation. See file COPYING.
*
*/
#ifndef CEPH_PG_H
#define CEPH_PG_H
#include <boost/statechart/custom_reaction.hpp>
#include <boost/statechart/event.hpp>
#include <boost/statechart/simple_state.hpp>
#include <boost/statechart/state.hpp>
#include <boost/statechart/state_machine.hpp>
#include <boost/statechart/transition.hpp>
#include <boost/statechart/event_base.hpp>
#include <boost/scoped_ptr.hpp>
#include <boost/container/flat_set.hpp>
#include "include/mempool.h"
// re-include our assert to clobber boost's
#include "include/ceph_assert.h"
#include "include/types.h"
#include "include/stringify.h"
#include "osd_types.h"
#include "include/xlist.h"
#include "SnapMapper.h"
#include "Session.h"
#include "common/Timer.h"
#include "PGLog.h"
#include "OSDMap.h"
#include "messages/MOSDPGLog.h"
#include "include/str_list.h"
#include "PGBackend.h"
#include "PGPeeringEvent.h"
#include "PeeringState.h"
#include "MissingLoc.h"
#include "mgr/OSDPerfMetricTypes.h"
#include <atomic>
#include <list>
#include <memory>
#include <string>
#include <tuple>
//#define DEBUG_RECOVERY_OIDS // track set of recovering oids explicitly, to find counting bugs
//#define PG_DEBUG_REFS // track provenance of pg refs, helpful for finding leaks
class OSD;
class OSDService;
class OSDShard;
class OSDShardPGSlot;
class MOSDOp;
class MOSDPGScan;
class MOSDPGBackfill;
class MOSDPGInfo;
class PG;
struct OpRequest;
typedef OpRequest::Ref OpRequestRef;
class MOSDPGLog;
class CephContext;
class DynamicPerfStats;
namespace Scrub {
class Store;
}
#ifdef PG_DEBUG_REFS
#include "common/tracked_int_ptr.hpp"
uint64_t get_with_id(PG *pg);
void put_with_id(PG *pg, uint64_t id);
typedef TrackedIntPtr<PG> PGRef;
#else
typedef boost::intrusive_ptr<PG> PGRef;
#endif
class PGRecoveryStats {
struct per_state_info {
uint64_t enter, exit; // enter/exit counts
uint64_t events;
utime_t event_time; // time spent processing events
utime_t total_time; // total time in state
utime_t min_time, max_time;
// cppcheck-suppress unreachableCode
per_state_info() : enter(0), exit(0), events(0) {}
};
map<const char *,per_state_info> info;
ceph::mutex lock = ceph::make_mutex("PGRecoverStats::lock");
public:
PGRecoveryStats() = default;
void reset() {
std::lock_guard l(lock);
info.clear();
}
void dump(ostream& out) {
std::lock_guard l(lock);
for (map<const char *,per_state_info>::iterator p = info.begin(); p != info.end(); ++p) {
per_state_info& i = p->second;
out << i.enter << "\t" << i.exit << "\t"
<< i.events << "\t" << i.event_time << "\t"
<< i.total_time << "\t"
<< i.min_time << "\t" << i.max_time << "\t"
<< p->first << "\n";
}
}
void dump_formatted(Formatter *f) {
std::lock_guard l(lock);
f->open_array_section("pg_recovery_stats");
for (map<const char *,per_state_info>::iterator p = info.begin();
p != info.end(); ++p) {
per_state_info& i = p->second;
f->open_object_section("recovery_state");
f->dump_int("enter", i.enter);
f->dump_int("exit", i.exit);
f->dump_int("events", i.events);
f->dump_stream("event_time") << i.event_time;
f->dump_stream("total_time") << i.total_time;
f->dump_stream("min_time") << i.min_time;
f->dump_stream("max_time") << i.max_time;
vector<string> states;
get_str_vec(p->first, "/", states);
f->open_array_section("nested_states");
for (vector<string>::iterator st = states.begin();
st != states.end(); ++st) {
f->dump_string("state", *st);
}
f->close_section();
f->close_section();
}
f->close_section();
}
void log_enter(const char *s) {
std::lock_guard l(lock);
info[s].enter++;
}
void log_exit(const char *s, utime_t dur, uint64_t events, utime_t event_dur) {
std::lock_guard l(lock);
per_state_info &i = info[s];
i.exit++;
i.total_time += dur;
if (dur > i.max_time)
i.max_time = dur;
if (dur < i.min_time || i.min_time == utime_t())
i.min_time = dur;
i.events += events;
i.event_time += event_dur;
}
};
/** PG - Replica Placement Group
*
*/
class PG : public DoutPrefixProvider, public PeeringState::PeeringListener {
friend class NamedState;
friend class PeeringState;
public:
const pg_shard_t pg_whoami;
const spg_t pg_id;
public:
// -- members --
const coll_t coll;
ObjectStore::CollectionHandle ch;
// -- methods --
std::ostream& gen_prefix(std::ostream& out) const override;
CephContext *get_cct() const override {
return cct;
}
unsigned get_subsys() const override {
return ceph_subsys_osd;
}
const char* const get_current_state() const {
return recovery_state.get_current_state();
}
const OSDMapRef& get_osdmap() const {
ceph_assert(is_locked());
return recovery_state.get_osdmap();
}
epoch_t get_osdmap_epoch() const override final {
return recovery_state.get_osdmap()->get_epoch();
}
PerfCounters &get_peering_perf() override;
PerfCounters &get_perf_logger() override;
void log_state_enter(const char *state) override;
void log_state_exit(
const char *state_name, utime_t enter_time,
uint64_t events, utime_t event_dur) override;
void lock_suspend_timeout(ThreadPool::TPHandle &handle) {
handle.suspend_tp_timeout();
lock();
handle.reset_tp_timeout();
}
void lock(bool no_lockdep = false) const;
void unlock() const;
bool is_locked() const;
const spg_t& get_pgid() const {
return pg_id;
}
const PGPool& get_pool() const {
return pool;
}
uint64_t get_last_user_version() const {
return info.last_user_version;
}
const pg_history_t& get_history() const {
return info.history;
}
bool get_need_up_thru() const {
return recovery_state.get_need_up_thru();
}
epoch_t get_same_interval_since() const {
return info.history.same_interval_since;
}
static void set_last_scrub_stamp(
utime_t t, pg_history_t &history, pg_stat_t &stats) {
stats.last_scrub_stamp = t;
history.last_scrub_stamp = t;
}
void set_last_scrub_stamp(utime_t t) {
recovery_state.update_stats(
[=](auto &history, auto &stats) {
set_last_scrub_stamp(t, history, stats);
return true;
});
}
static void set_last_deep_scrub_stamp(
utime_t t, pg_history_t &history, pg_stat_t &stats) {
stats.last_deep_scrub_stamp = t;
history.last_deep_scrub_stamp = t;
}
void set_last_deep_scrub_stamp(utime_t t) {
recovery_state.update_stats(
[=](auto &history, auto &stats) {
set_last_deep_scrub_stamp(t, history, stats);
return true;
});
}
bool is_deleting() const {
return recovery_state.is_deleting();
}
bool is_deleted() const {
return recovery_state.is_deleted();
}
bool is_nonprimary() const {
return recovery_state.is_nonprimary();
}
bool is_primary() const {
return recovery_state.is_primary();
}
bool pg_has_reset_since(epoch_t e) {
ceph_assert(is_locked());
return recovery_state.pg_has_reset_since(e);
}
bool is_ec_pg() const {
return recovery_state.is_ec_pg();
}
int get_role() const {
return recovery_state.get_role();
}
const vector<int> get_acting() const {
return recovery_state.get_acting();
}
const set<pg_shard_t> &get_actingset() const {
return recovery_state.get_actingset();
}
int get_acting_primary() const {
return recovery_state.get_acting_primary();
}
pg_shard_t get_primary() const {
return recovery_state.get_primary();
}
const vector<int> get_up() const {
return recovery_state.get_up();
}
int get_up_primary() const {
return recovery_state.get_up_primary();
}
const PastIntervals& get_past_intervals() const {
return recovery_state.get_past_intervals();
}
bool is_acting_recovery_backfill(pg_shard_t osd) const {
return recovery_state.is_acting_recovery_backfill(osd);
}
const set<pg_shard_t> &get_acting_recovery_backfill() const {
return recovery_state.get_acting_recovery_backfill();
}
bool is_acting(pg_shard_t osd) const {
return recovery_state.is_acting(osd);
}
bool is_up(pg_shard_t osd) const {
return recovery_state.is_up(osd);
}
static bool has_shard(bool ec, const vector<int>& v, pg_shard_t osd) {
return PeeringState::has_shard(ec, v, osd);
}
/// initialize created PG
void init(
int role,
const vector<int>& up,
int up_primary,
const vector<int>& acting,
int acting_primary,
const pg_history_t& history,
const PastIntervals& pim,
bool backfill,
ObjectStore::Transaction &t);
/// read existing pg state off disk
void read_state(ObjectStore *store);
static int peek_map_epoch(ObjectStore *store, spg_t pgid, epoch_t *pepoch);
static int get_latest_struct_v() {
return pg_latest_struct_v;
}
static int get_compat_struct_v() {
return pg_compat_struct_v;
}
static int read_info(
ObjectStore *store, spg_t pgid, const coll_t &coll,
pg_info_t &info, PastIntervals &past_intervals,
__u8 &);
static bool _has_removal_flag(ObjectStore *store, spg_t pgid);
void rm_backoff(const ceph::ref_t<Backoff>& b);
void update_snap_mapper_bits(uint32_t bits) {
snap_mapper.update_bits(bits);
}
void start_split_stats(const set<spg_t>& childpgs, vector<object_stat_sum_t> *v);
virtual void split_colls(
spg_t child,
int split_bits,
int seed,
const pg_pool_t *pool,
ObjectStore::Transaction &t) = 0;
void split_into(pg_t child_pgid, PG *child, unsigned split_bits);
void merge_from(map<spg_t,PGRef>& sources, PeeringCtx &rctx,
unsigned split_bits,
const pg_merge_meta_t& last_pg_merge_meta);
void finish_split_stats(const object_stat_sum_t& stats,
ObjectStore::Transaction &t);
void scrub(epoch_t queued, ThreadPool::TPHandle &handle);
bool is_scrub_registered();
void reg_next_scrub();
void unreg_next_scrub();
void queue_want_pg_temp(const vector<int> &wanted) override;
void clear_want_pg_temp() override;
void on_new_interval() override;
void on_role_change() override;
virtual void plpg_on_role_change() = 0;
void init_collection_pool_opts();
void on_pool_change() override;
virtual void plpg_on_pool_change() = 0;
void on_info_history_change() override;
void scrub_requested(bool deep, bool repair, bool need_auto = false) override;
uint64_t get_snap_trimq_size() const override {
return snap_trimq.size();
}
void clear_publish_stats() override;
void clear_primary_state() override;
epoch_t oldest_stored_osdmap() override;
OstreamTemp get_clog_error() override;
OstreamTemp get_clog_info() override;
OstreamTemp get_clog_debug() override;
void schedule_event_after(
PGPeeringEventRef event,
float delay) override;
void request_local_background_io_reservation(
unsigned priority,
PGPeeringEventRef on_grant,
PGPeeringEventRef on_preempt) override;
void update_local_background_io_priority(
unsigned priority) override;
void cancel_local_background_io_reservation() override;
void request_remote_recovery_reservation(
unsigned priority,
PGPeeringEventRef on_grant,
PGPeeringEventRef on_preempt) override;
void cancel_remote_recovery_reservation() override;
void schedule_event_on_commit(
ObjectStore::Transaction &t,
PGPeeringEventRef on_commit) override;
void on_active_exit() override;
Context *on_clean() override {
if (is_active()) {
kick_snap_trim();
}
requeue_ops(waiting_for_clean_to_primary_repair);
return finish_recovery();
}
void on_activate(interval_set<snapid_t> snaps) override {
ceph_assert(scrubber.callbacks.empty());
ceph_assert(callbacks_for_degraded_object.empty());
snap_trimq = snaps;
release_pg_backoffs();
projected_last_update = info.last_update;
}
void on_activate_committed() override;
void on_active_actmap() override;
void on_active_advmap(const OSDMapRef &osdmap) override;
void queue_snap_retrim(snapid_t snap);
void on_backfill_reserved() override;
void on_backfill_canceled() override;
void on_recovery_reserved() override;
bool is_forced_recovery_or_backfill() const {
return recovery_state.is_forced_recovery_or_backfill();
}
PGLog::LogEntryHandlerRef get_log_handler(
ObjectStore::Transaction &t) override {
return std::make_unique<PG::PGLogEntryHandler>(this, &t);
}
void do_delete_work(ObjectStore::Transaction &t) override;
void clear_ready_to_merge() override;
void set_not_ready_to_merge_target(pg_t pgid, pg_t src) override;
void set_not_ready_to_merge_source(pg_t pgid) override;
void set_ready_to_merge_target(eversion_t lu, epoch_t les, epoch_t lec) override;
void set_ready_to_merge_source(eversion_t lu) override;
void send_pg_created(pg_t pgid) override;
ceph::signedspan get_mnow() override;
HeartbeatStampsRef get_hb_stamps(int peer) override;
void schedule_renew_lease(epoch_t lpr, ceph::timespan delay) override;
void queue_check_readable(epoch_t lpr, ceph::timespan delay) override;
void rebuild_missing_set_with_deletes(PGLog &pglog) override;
void queue_peering_event(PGPeeringEventRef evt);
void do_peering_event(PGPeeringEventRef evt, PeeringCtx &rcx);
void queue_null(epoch_t msg_epoch, epoch_t query_epoch);
void queue_flushed(epoch_t started_at);
void handle_advance_map(
OSDMapRef osdmap, OSDMapRef lastmap,
vector<int>& newup, int up_primary,
vector<int>& newacting, int acting_primary,
PeeringCtx &rctx);
void handle_activate_map(PeeringCtx &rctx);
void handle_initialize(PeeringCtx &rxcx);
void handle_query_state(Formatter *f);
/**
* @param ops_begun returns how many recovery ops the function started
* @returns true if any useful work was accomplished; false otherwise
*/
virtual bool start_recovery_ops(
uint64_t max,
ThreadPool::TPHandle &handle,
uint64_t *ops_begun) = 0;
// more work after the above, but with a PeeringCtx
void find_unfound(epoch_t queued, PeeringCtx &rctx);
virtual void get_watchers(std::list<obj_watch_item_t> *ls) = 0;
void dump_pgstate_history(Formatter *f);
void dump_missing(Formatter *f);
void get_pg_stats(std::function<void(const pg_stat_t&, epoch_t lec)> f);
void with_heartbeat_peers(std::function<void(int)> f);
void shutdown();
virtual void on_shutdown() = 0;
bool get_must_scrub() const {
return scrubber.must_scrub;
}
bool sched_scrub();
virtual void do_request(
OpRequestRef& op,
ThreadPool::TPHandle &handle
) = 0;
virtual void clear_cache() = 0;
virtual int get_cache_obj_count() = 0;
virtual void snap_trimmer(epoch_t epoch_queued) = 0;
virtual void do_command(
const string_view& prefix,
const cmdmap_t& cmdmap,
const bufferlist& idata,
std::function<void(int,const std::string&,bufferlist&)> on_finish) = 0;
virtual bool agent_work(int max) = 0;
virtual bool agent_work(int max, int agent_flush_quota) = 0;
virtual void agent_stop() = 0;
virtual void agent_delay() = 0;
virtual void agent_clear() = 0;
virtual void agent_choose_mode_restart() = 0;
struct C_DeleteMore : public Context {
PGRef pg;
epoch_t epoch;
C_DeleteMore(PG *p, epoch_t e) : pg(p), epoch(e) {}
void finish(int r) override {
ceph_abort();
}
void complete(int r) override;
};
void _delete_some(ObjectStore::Transaction *t);
virtual void set_dynamic_perf_stats_queries(
const std::list<OSDPerfMetricQuery> &queries) {
}
virtual void get_dynamic_perf_stats(DynamicPerfStats *stats) {
}
uint64_t get_min_alloc_size() const;
// reference counting
#ifdef PG_DEBUG_REFS
uint64_t get_with_id();
void put_with_id(uint64_t);
void dump_live_ids();
#endif
void get(const char* tag);
void put(const char* tag);
int get_num_ref() {
return ref;
}
// ctor
PG(OSDService *o, OSDMapRef curmap,
const PGPool &pool, spg_t p);
~PG() override;
// prevent copying
explicit PG(const PG& rhs) = delete;
PG& operator=(const PG& rhs) = delete;
protected:
// -------------
// protected
OSDService *osd;
public:
OSDShard *osd_shard = nullptr;
OSDShardPGSlot *pg_slot = nullptr;
protected:
CephContext *cct;
// locking and reference counting.
// I destroy myself when the reference count hits zero.
// lock() should be called before doing anything.
// get() should be called on pointer copy (to another thread, etc.).
// put() should be called on destruction of some previously copied pointer.
// unlock() when done with the current pointer (_most common_).
mutable ceph::mutex _lock = ceph::make_mutex("PG::_lock");
#ifndef CEPH_DEBUG_MUTEX
mutable std::thread::id locked_by;
#endif
std::atomic<unsigned int> ref{0};
#ifdef PG_DEBUG_REFS
ceph::mutex _ref_id_lock = ceph::make_mutex("PG::_ref_id_lock");
map<uint64_t, string> _live_ids;
map<string, uint64_t> _tag_counts;
uint64_t _ref_id = 0;
friend uint64_t get_with_id(PG *pg) { return pg->get_with_id(); }
friend void put_with_id(PG *pg, uint64_t id) { return pg->put_with_id(id); }
#endif
private:
friend void intrusive_ptr_add_ref(PG *pg) {
pg->get("intptr");
}
friend void intrusive_ptr_release(PG *pg) {
pg->put("intptr");
}
// =====================
protected:
OSDriver osdriver;
SnapMapper snap_mapper;
bool eio_errors_to_process = false;
virtual PGBackend *get_pgbackend() = 0;
virtual const PGBackend* get_pgbackend() const = 0;
protected:
void requeue_map_waiters();
protected:
ZTracer::Endpoint trace_endpoint;
protected:
__u8 info_struct_v = 0;
void upgrade(ObjectStore *store);
protected:
ghobject_t pgmeta_oid;
// ------------------
interval_set<snapid_t> snap_trimq;
set<snapid_t> snap_trimq_repeat;
/* You should not use these items without taking their respective queue locks
* (if they have one) */
xlist<PG*>::item stat_queue_item;
bool scrub_queued;
bool recovery_queued;
int recovery_ops_active;
set<pg_shard_t> waiting_on_backfill;
#ifdef DEBUG_RECOVERY_OIDS
multiset<hobject_t> recovering_oids;
#endif
public:
bool dne() { return info.dne(); }
virtual void send_cluster_message(
int osd, Message *m, epoch_t epoch, bool share_map_update) override;
protected:
epoch_t get_last_peering_reset() const {
return recovery_state.get_last_peering_reset();
}
/* heartbeat peers */
void set_probe_targets(const set<pg_shard_t> &probe_set) override;
void clear_probe_targets() override;
ceph::mutex heartbeat_peer_lock =
ceph::make_mutex("PG::heartbeat_peer_lock");
set<int> heartbeat_peers;
set<int> probe_targets;
public:
/**
* BackfillInterval
*
* Represents the objects in a range [begin, end)
*
* Possible states:
* 1) begin == end == hobject_t() indicates the the interval is unpopulated
* 2) Else, objects contains all objects in [begin, end)
*/
struct BackfillInterval {
// info about a backfill interval on a peer
eversion_t version; /// version at which the scan occurred
map<hobject_t,eversion_t> objects;
hobject_t begin;
hobject_t end;
/// clear content
void clear() {
*this = BackfillInterval();
}
/// clear objects list only
void clear_objects() {
objects.clear();
}
/// reinstantiate with a new start+end position and sort order
void reset(hobject_t start) {
clear();
begin = end = start;
}
/// true if there are no objects in this interval
bool empty() const {
return objects.empty();
}
/// true if interval extends to the end of the range
bool extends_to_end() const {
return end.is_max();
}
/// removes items <= soid and adjusts begin to the first object
void trim_to(const hobject_t &soid) {
trim();
while (!objects.empty() &&
objects.begin()->first <= soid) {
pop_front();
}
}
/// Adjusts begin to the first object
void trim() {
if (!objects.empty())
begin = objects.begin()->first;
else
begin = end;
}
/// drop first entry, and adjust @begin accordingly
void pop_front() {
ceph_assert(!objects.empty());
objects.erase(objects.begin());
trim();
}
/// dump
void dump(Formatter *f) const {
f->dump_stream("begin") << begin;
f->dump_stream("end") << end;
f->open_array_section("objects");
for (map<hobject_t, eversion_t>::const_iterator i =
objects.begin();
i != objects.end();
++i) {
f->open_object_section("object");
f->dump_stream("object") << i->first;
f->dump_stream("version") << i->second;
f->close_section();
}
f->close_section();
}
};
protected:
BackfillInterval backfill_info;
map<pg_shard_t, BackfillInterval> peer_backfill_info;
bool backfill_reserving;
// The primary's num_bytes and local num_bytes for this pg, only valid
// during backfill for non-primary shards.
// Both of these are adjusted for EC to reflect the on-disk bytes
std::atomic<int64_t> primary_num_bytes = 0;
std::atomic<int64_t> local_num_bytes = 0;
public:
// Space reserved for backfill is primary_num_bytes - local_num_bytes
// Don't care that difference itself isn't atomic
uint64_t get_reserved_num_bytes() {
int64_t primary = primary_num_bytes.load();
int64_t local = local_num_bytes.load();
if (primary > local)
return primary - local;
else
return 0;
}
bool is_remote_backfilling() {
return primary_num_bytes.load() > 0;
}
bool try_reserve_recovery_space(int64_t primary, int64_t local) override;
void unreserve_recovery_space() override;
// If num_bytes are inconsistent and local_num- goes negative
// it's ok, because it would then be ignored.
// The value of num_bytes could be negative,
// but we don't let local_num_bytes go negative.
void add_local_num_bytes(int64_t num_bytes) {
if (num_bytes) {
int64_t prev_bytes = local_num_bytes.load();
int64_t new_bytes;
do {
new_bytes = prev_bytes + num_bytes;
if (new_bytes < 0)
new_bytes = 0;
} while(!local_num_bytes.compare_exchange_weak(prev_bytes, new_bytes));
}
}
void sub_local_num_bytes(int64_t num_bytes) {
ceph_assert(num_bytes >= 0);
if (num_bytes) {
int64_t prev_bytes = local_num_bytes.load();
int64_t new_bytes;
do {
new_bytes = prev_bytes - num_bytes;
if (new_bytes < 0)
new_bytes = 0;
} while(!local_num_bytes.compare_exchange_weak(prev_bytes, new_bytes));
}
}
// The value of num_bytes could be negative,
// but we don't let info.stats.stats.sum.num_bytes go negative.
void add_num_bytes(int64_t num_bytes) {
ceph_assert(ceph_mutex_is_locked_by_me(_lock));
if (num_bytes) {
recovery_state.update_stats(
[num_bytes](auto &history, auto &stats) {
stats.stats.sum.num_bytes += num_bytes;
if (stats.stats.sum.num_bytes < 0) {
stats.stats.sum.num_bytes = 0;
}
return false;
});
}
}
void sub_num_bytes(int64_t num_bytes) {
ceph_assert(ceph_mutex_is_locked_by_me(_lock));
ceph_assert(num_bytes >= 0);
if (num_bytes) {
recovery_state.update_stats(
[num_bytes](auto &history, auto &stats) {
stats.stats.sum.num_bytes -= num_bytes;
if (stats.stats.sum.num_bytes < 0) {
stats.stats.sum.num_bytes = 0;
}
return false;
});
}
}
// Only used in testing so not worried about needing the PG lock here
int64_t get_stats_num_bytes() {
std::lock_guard l{_lock};
int num_bytes = info.stats.stats.sum.num_bytes;
if (pool.info.is_erasure()) {
num_bytes /= (int)get_pgbackend()->get_ec_data_chunk_count();
// Round up each object by a stripe
num_bytes += get_pgbackend()->get_ec_stripe_chunk_size() * info.stats.stats.sum.num_objects;
}
int64_t lnb = local_num_bytes.load();
if (lnb && lnb != num_bytes) {
lgeneric_dout(cct, 0) << this << " " << info.pgid << " num_bytes mismatch "
<< lnb << " vs stats "
<< info.stats.stats.sum.num_bytes << " / chunk "
<< get_pgbackend()->get_ec_data_chunk_count()
<< dendl;
}
return num_bytes;
}
protected:
/*
* blocked request wait hierarchy
*
* In order to preserve request ordering we need to be careful about the
* order in which blocked requests get requeued. Generally speaking, we
* push the requests back up to the op_wq in reverse order (most recent
* request first) so that they come back out again in the original order.
* However, because there are multiple wait queues, we need to requeue
* waitlists in order. Generally speaking, we requeue the wait lists
* that are checked first.
*
* Here are the various wait lists, in the order they are used during
* request processing, with notes:
*
* - waiting_for_map
* - may start or stop blocking at any time (depending on client epoch)
* - waiting_for_peered
* - !is_peered()
* - only starts blocking on interval change; never restarts
* - waiting_for_flush
* - flushes_in_progress
* - waiting for final flush during activate
* - waiting_for_active
* - !is_active()
* - only starts blocking on interval change; never restarts
* - waiting_for_readable
* - now > readable_until
* - unblocks when we get fresh(er) osd_pings
* - waiting_for_scrub
* - starts and stops blocking for varying intervals during scrub
* - waiting_for_unreadable_object
* - never restarts once object is readable (* except for EIO?)
* - waiting_for_degraded_object
* - never restarts once object is writeable (* except for EIO?)
* - waiting_for_blocked_object
* - starts and stops based on proxied op activity
* - obc rwlocks
* - starts and stops based on read/write activity
*
* Notes:
*
* 1. During and interval change, we requeue *everything* in the above order.
*
* 2. When an obc rwlock is released, we check for a scrub block and requeue
* the op there if it applies. We ignore the unreadable/degraded/blocked
* queues because we assume they cannot apply at that time (this is
* probably mostly true).
*
* 3. The requeue_ops helper will push ops onto the waiting_for_map list if
* it is non-empty.
*
* These three behaviors are generally sufficient to maintain ordering, with
* the possible exception of cases where we make an object degraded or
* unreadable that was previously okay, e.g. when scrub or op processing
* encounter an unexpected error. FIXME.
*/
// ops with newer maps than our (or blocked behind them)
// track these by client, since inter-request ordering doesn't otherwise
// matter.
unordered_map<entity_name_t,list<OpRequestRef>> waiting_for_map;
// ops waiting on peered
list<OpRequestRef> waiting_for_peered;
/// ops waiting on readble
list<OpRequestRef> waiting_for_readable;
// ops waiting on active (require peered as well)
list<OpRequestRef> waiting_for_active;
list<OpRequestRef> waiting_for_flush;
list<OpRequestRef> waiting_for_scrub;
list<OpRequestRef> waiting_for_cache_not_full;
list<OpRequestRef> waiting_for_clean_to_primary_repair;
map<hobject_t, list<OpRequestRef>> waiting_for_unreadable_object,
waiting_for_degraded_object,
waiting_for_blocked_object;
set<hobject_t> objects_blocked_on_cache_full;
map<hobject_t,snapid_t> objects_blocked_on_degraded_snap;
map<hobject_t,ObjectContextRef> objects_blocked_on_snap_promotion;
// Callbacks should assume pg (and nothing else) is locked
map<hobject_t, list<Context*>> callbacks_for_degraded_object;
map<eversion_t,
list<
tuple<OpRequestRef, version_t, int,
vector<pg_log_op_return_item_t>>>> waiting_for_ondisk;
void requeue_object_waiters(map<hobject_t, list<OpRequestRef>>& m);
void requeue_op(OpRequestRef op);
void requeue_ops(list<OpRequestRef> &l);
// stats that persist lazily
object_stat_collection_t unstable_stats;
// publish stats
ceph::mutex pg_stats_publish_lock =
ceph::make_mutex("PG::pg_stats_publish_lock");
bool pg_stats_publish_valid;
pg_stat_t pg_stats_publish;
friend class TestOpsSocketHook;
void publish_stats_to_osd() override;
bool needs_recovery() const {
return recovery_state.needs_recovery();
}
bool needs_backfill() const {
return recovery_state.needs_backfill();
}