-
Notifications
You must be signed in to change notification settings - Fork 3.2k
/
Copy pathrdkafka_msgset_writer.c
1446 lines (1207 loc) · 55.1 KB
/
rdkafka_msgset_writer.c
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
/*
* librdkafka - Apache Kafka C library
*
* Copyright (c) 2017-2022, Magnus Edenhill
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* 1. Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
#include "rd.h"
#include "rdkafka_int.h"
#include "rdkafka_msg.h"
#include "rdkafka_msgset.h"
#include "rdkafka_topic.h"
#include "rdkafka_partition.h"
#include "rdkafka_header.h"
#include "rdkafka_lz4.h"
#if WITH_ZSTD
#include "rdkafka_zstd.h"
#endif
#include "snappy.h"
#include "rdvarint.h"
#include "crc32c.h"
/** @brief The maxium ProduceRequestion ApiVersion supported by librdkafka */
static const int16_t rd_kafka_ProduceRequest_max_version = 8;
typedef struct rd_kafka_msgset_writer_s {
rd_kafka_buf_t *msetw_rkbuf; /* Backing store buffer (refcounted)*/
int16_t msetw_ApiVersion; /* ProduceRequest ApiVersion */
int msetw_MsgVersion; /* MsgVersion to construct */
int msetw_features; /* Protocol features to use */
rd_kafka_compression_t msetw_compression; /**< Compression type */
int msetw_msgcntmax; /* Max number of messages to send
* in a batch. */
size_t msetw_messages_len; /* Total size of Messages, with Message
* framing but without
* MessageSet header */
size_t msetw_messages_kvlen; /* Total size of Message keys
* and values */
size_t msetw_MessageSetSize; /* Current MessageSetSize value */
size_t msetw_of_MessageSetSize; /* offset of MessageSetSize */
size_t msetw_of_start; /* offset of MessageSet */
int msetw_relative_offsets; /* Bool: use relative offsets */
/* For MessageSet v2 */
int msetw_Attributes; /* MessageSet Attributes */
int64_t msetw_MaxTimestamp; /* Maximum timestamp in batch */
size_t msetw_of_CRC; /* offset of MessageSet.CRC */
rd_kafka_msgbatch_t *msetw_batch; /**< Convenience pointer to
* rkbuf_u.Produce.batch */
/* First message information */
struct {
size_t of; /* rkbuf's first message position */
int64_t timestamp;
} msetw_firstmsg;
rd_kafka_pid_t msetw_pid; /**< Idempotent producer's
* current Producer Id */
rd_kafka_broker_t *msetw_rkb; /* @warning Not a refcounted
* reference! */
rd_kafka_toppar_t *msetw_rktp; /* @warning Not a refcounted
* reference! */
rd_kafka_msgq_t *msetw_msgq; /**< Input message queue */
} rd_kafka_msgset_writer_t;
/**
* @brief Select ApiVersion and MsgVersion to use based on broker's
* feature compatibility.
*
* @returns -1 if a MsgVersion (or ApiVersion) could not be selected, else 0.
* @locality broker thread
*/
static RD_INLINE int
rd_kafka_msgset_writer_select_MsgVersion(rd_kafka_msgset_writer_t *msetw) {
rd_kafka_broker_t *rkb = msetw->msetw_rkb;
rd_kafka_toppar_t *rktp = msetw->msetw_rktp;
const int16_t max_ApiVersion = rd_kafka_ProduceRequest_max_version;
int16_t min_ApiVersion = 0;
int feature;
/* Map compression types to required feature and ApiVersion */
static const struct {
int feature;
int16_t ApiVersion;
} compr_req[RD_KAFKA_COMPRESSION_NUM] = {
[RD_KAFKA_COMPRESSION_LZ4] = {RD_KAFKA_FEATURE_LZ4, 0},
#if WITH_ZSTD
[RD_KAFKA_COMPRESSION_ZSTD] = {RD_KAFKA_FEATURE_ZSTD, 7},
#endif
};
if ((feature = rkb->rkb_features & RD_KAFKA_FEATURE_MSGVER2)) {
min_ApiVersion = 3;
msetw->msetw_MsgVersion = 2;
msetw->msetw_features |= feature;
} else if ((feature = rkb->rkb_features & RD_KAFKA_FEATURE_MSGVER1)) {
min_ApiVersion = 2;
msetw->msetw_MsgVersion = 1;
msetw->msetw_features |= feature;
} else {
if ((feature =
rkb->rkb_features & RD_KAFKA_FEATURE_THROTTLETIME)) {
min_ApiVersion = 1;
msetw->msetw_features |= feature;
} else
min_ApiVersion = 0;
msetw->msetw_MsgVersion = 0;
}
msetw->msetw_compression = rktp->rktp_rkt->rkt_conf.compression_codec;
/*
* Check that the configured compression type is supported
* by both client and broker, else disable compression.
*/
if (msetw->msetw_compression &&
(rd_kafka_broker_ApiVersion_supported(
rkb, RD_KAFKAP_Produce, 0,
compr_req[msetw->msetw_compression].ApiVersion, NULL) == -1 ||
(compr_req[msetw->msetw_compression].feature &&
!(msetw->msetw_rkb->rkb_features &
compr_req[msetw->msetw_compression].feature)))) {
if (unlikely(
rd_interval(&rkb->rkb_suppress.unsupported_compression,
/* at most once per day */
(rd_ts_t)86400 * 1000 * 1000, 0) > 0))
rd_rkb_log(
rkb, LOG_NOTICE, "COMPRESSION",
"%.*s [%" PRId32
"]: "
"Broker does not support compression "
"type %s: not compressing batch",
RD_KAFKAP_STR_PR(rktp->rktp_rkt->rkt_topic),
rktp->rktp_partition,
rd_kafka_compression2str(msetw->msetw_compression));
else
rd_rkb_dbg(
rkb, MSG, "PRODUCE",
"%.*s [%" PRId32
"]: "
"Broker does not support compression "
"type %s: not compressing batch",
RD_KAFKAP_STR_PR(rktp->rktp_rkt->rkt_topic),
rktp->rktp_partition,
rd_kafka_compression2str(msetw->msetw_compression));
msetw->msetw_compression = RD_KAFKA_COMPRESSION_NONE;
} else {
/* Broker supports this compression type. */
msetw->msetw_features |=
compr_req[msetw->msetw_compression].feature;
if (min_ApiVersion <
compr_req[msetw->msetw_compression].ApiVersion)
min_ApiVersion =
compr_req[msetw->msetw_compression].ApiVersion;
}
/* MsgVersion specific setup. */
switch (msetw->msetw_MsgVersion) {
case 2:
msetw->msetw_relative_offsets = 1; /* OffsetDelta */
break;
case 1:
if (msetw->msetw_compression != RD_KAFKA_COMPRESSION_NONE)
msetw->msetw_relative_offsets = 1;
break;
}
/* Set the highest ApiVersion supported by us and broker */
msetw->msetw_ApiVersion = rd_kafka_broker_ApiVersion_supported(
rkb, RD_KAFKAP_Produce, min_ApiVersion, max_ApiVersion, NULL);
if (msetw->msetw_ApiVersion == -1) {
rd_kafka_msg_t *rkm;
/* This will only happen if the broker reports none, or
* no matching ProduceRequest versions, which should never
* happen. */
rd_rkb_log(rkb, LOG_ERR, "PRODUCE",
"%.*s [%" PRId32
"]: "
"No viable ProduceRequest ApiVersions (v%d..%d) "
"supported by broker: unable to produce",
RD_KAFKAP_STR_PR(rktp->rktp_rkt->rkt_topic),
rktp->rktp_partition, min_ApiVersion,
max_ApiVersion);
/* Back off and retry in 5s */
rkm = rd_kafka_msgq_first(msetw->msetw_msgq);
rd_assert(rkm);
rkm->rkm_u.producer.ts_backoff = rd_clock() + (5 * 1000 * 1000);
return -1;
}
/* It should not be possible to get a lower version than requested,
* otherwise the logic in this function is buggy. */
rd_assert(msetw->msetw_ApiVersion >= min_ApiVersion);
return 0;
}
/**
* @brief Allocate buffer for messageset writer based on a previously set
* up \p msetw.
*
* Allocate iovecs to hold all headers and messages,
* and allocate enough space to allow copies of small messages.
* The allocated size is the minimum of message.max.bytes
* or queued_bytes + msgcntmax * msg_overhead
*/
static void rd_kafka_msgset_writer_alloc_buf(rd_kafka_msgset_writer_t *msetw) {
rd_kafka_t *rk = msetw->msetw_rkb->rkb_rk;
size_t msg_overhead = 0;
size_t hdrsize = 0;
size_t msgsetsize = 0;
size_t bufsize;
rd_kafka_assert(NULL, !msetw->msetw_rkbuf);
/* Calculate worst-case buffer size, produce header size,
* message size, etc, this isn't critical but avoids unnecesary
* extra allocations. The buffer will grow as needed if we get
* this wrong.
*
* ProduceRequest headers go in one iovec:
* ProduceRequest v0..2:
* RequiredAcks + Timeout +
* [Topic + [Partition + MessageSetSize]]
*
* ProduceRequest v3:
* TransactionalId + RequiredAcks + Timeout +
* [Topic + [Partition + MessageSetSize + MessageSet]]
*/
/*
* ProduceRequest header sizes
*/
switch (msetw->msetw_ApiVersion) {
case 8:
case 7:
case 6:
case 5:
case 4:
case 3:
/* Add TransactionalId */
hdrsize += RD_KAFKAP_STR_SIZE(rk->rk_eos.transactional_id);
/* FALLTHRU */
case 0:
case 1:
case 2:
hdrsize +=
/* RequiredAcks + Timeout + TopicCnt */
2 + 4 + 4 +
/* Topic */
RD_KAFKAP_STR_SIZE(msetw->msetw_rktp->rktp_rkt->rkt_topic) +
/* PartitionCnt + Partition + MessageSetSize */
4 + 4 + 4;
msgsetsize += 4; /* MessageSetSize */
break;
default:
RD_NOTREACHED();
}
/*
* MsgVersion specific sizes:
* - (Worst-case) Message overhead: message fields
* - MessageSet header size
*/
switch (msetw->msetw_MsgVersion) {
case 0:
/* MsgVer0 */
msg_overhead = RD_KAFKAP_MESSAGE_V0_OVERHEAD;
break;
case 1:
/* MsgVer1 */
msg_overhead = RD_KAFKAP_MESSAGE_V1_OVERHEAD;
break;
case 2:
/* MsgVer2 uses varints, we calculate for the worst-case. */
msg_overhead += RD_KAFKAP_MESSAGE_V2_MAX_OVERHEAD;
/* MessageSet header fields */
msgsetsize += 8 /* BaseOffset */ + 4 /* Length */ +
4 /* PartitionLeaderEpoch */ +
1 /* Magic (MsgVersion) */ +
4 /* CRC (CRC32C) */ + 2 /* Attributes */ +
4 /* LastOffsetDelta */ + 8 /* BaseTimestamp */ +
8 /* MaxTimestamp */ + 8 /* ProducerId */ +
2 /* ProducerEpoch */ + 4 /* BaseSequence */ +
4 /* RecordCount */;
break;
default:
RD_NOTREACHED();
}
/*
* Calculate total buffer size to allocate
*/
bufsize = hdrsize + msgsetsize;
/* If copying for small payloads is enabled, allocate enough
* space for each message to be copied based on this limit.
*/
if (rk->rk_conf.msg_copy_max_size > 0) {
size_t queued_bytes = rd_kafka_msgq_size(msetw->msetw_msgq);
bufsize +=
RD_MIN(queued_bytes, (size_t)rk->rk_conf.msg_copy_max_size *
msetw->msetw_msgcntmax);
}
/* Add estimed per-message overhead */
bufsize += msg_overhead * msetw->msetw_msgcntmax;
/* Cap allocation at message.max.bytes */
if (bufsize > (size_t)rk->rk_conf.max_msg_size)
bufsize = (size_t)rk->rk_conf.max_msg_size;
/*
* Allocate iovecs to hold all headers and messages,
* and allocate auxilliery space for message headers, etc.
*/
msetw->msetw_rkbuf =
rd_kafka_buf_new_request(msetw->msetw_rkb, RD_KAFKAP_Produce,
msetw->msetw_msgcntmax / 2 + 10, bufsize);
rd_kafka_buf_ApiVersion_set(msetw->msetw_rkbuf, msetw->msetw_ApiVersion,
msetw->msetw_features);
}
/**
* @brief Write the MessageSet header.
* @remark Must only be called for MsgVersion 2
*/
static void rd_kafka_msgset_writer_write_MessageSet_v2_header(
rd_kafka_msgset_writer_t *msetw) {
rd_kafka_buf_t *rkbuf = msetw->msetw_rkbuf;
rd_kafka_assert(NULL, msetw->msetw_ApiVersion >= 3);
rd_kafka_assert(NULL, msetw->msetw_MsgVersion == 2);
/* BaseOffset (also store the offset to the start of
* the messageset header fields) */
msetw->msetw_of_start = rd_kafka_buf_write_i64(rkbuf, 0);
/* Length: updated later */
rd_kafka_buf_write_i32(rkbuf, 0);
/* PartitionLeaderEpoch (KIP-101) */
rd_kafka_buf_write_i32(rkbuf, 0);
/* Magic (MsgVersion) */
rd_kafka_buf_write_i8(rkbuf, msetw->msetw_MsgVersion);
/* CRC (CRC32C): updated later.
* CRC needs to be done after the entire messageset+messages has
* been constructed and the following header fields updated. :(
* Save the offset for this position. so it can be udpated later. */
msetw->msetw_of_CRC = rd_kafka_buf_write_i32(rkbuf, 0);
/* Attributes: updated later */
rd_kafka_buf_write_i16(rkbuf, 0);
/* LastOffsetDelta: updated later */
rd_kafka_buf_write_i32(rkbuf, 0);
/* BaseTimestamp: updated later */
rd_kafka_buf_write_i64(rkbuf, 0);
/* MaxTimestamp: updated later */
rd_kafka_buf_write_i64(rkbuf, 0);
/* ProducerId */
rd_kafka_buf_write_i64(rkbuf, msetw->msetw_pid.id);
/* ProducerEpoch */
rd_kafka_buf_write_i16(rkbuf, msetw->msetw_pid.epoch);
/* BaseSequence: updated later in case of Idempotent Producer */
rd_kafka_buf_write_i32(rkbuf, -1);
/* RecordCount: udpated later */
rd_kafka_buf_write_i32(rkbuf, 0);
}
/**
* @brief Write ProduceRequest headers.
* When this function returns the msgset is ready for
* writing individual messages.
* msetw_MessageSetSize will have been set to the messageset header.
*/
static void
rd_kafka_msgset_writer_write_Produce_header(rd_kafka_msgset_writer_t *msetw) {
rd_kafka_buf_t *rkbuf = msetw->msetw_rkbuf;
rd_kafka_t *rk = msetw->msetw_rkb->rkb_rk;
rd_kafka_topic_t *rkt = msetw->msetw_rktp->rktp_rkt;
/* V3: TransactionalId */
if (msetw->msetw_ApiVersion >= 3)
rd_kafka_buf_write_kstr(rkbuf, rk->rk_eos.transactional_id);
/* RequiredAcks */
rd_kafka_buf_write_i16(rkbuf, rkt->rkt_conf.required_acks);
/* Timeout */
rd_kafka_buf_write_i32(rkbuf, rkt->rkt_conf.request_timeout_ms);
/* TopicArrayCnt */
rd_kafka_buf_write_i32(rkbuf, 1);
/* Insert topic */
rd_kafka_buf_write_kstr(rkbuf, rkt->rkt_topic);
/* PartitionArrayCnt */
rd_kafka_buf_write_i32(rkbuf, 1);
/* Partition */
rd_kafka_buf_write_i32(rkbuf, msetw->msetw_rktp->rktp_partition);
/* MessageSetSize: Will be finalized later*/
msetw->msetw_of_MessageSetSize = rd_kafka_buf_write_i32(rkbuf, 0);
if (msetw->msetw_MsgVersion == 2) {
/* MessageSet v2 header */
rd_kafka_msgset_writer_write_MessageSet_v2_header(msetw);
msetw->msetw_MessageSetSize = RD_KAFKAP_MSGSET_V2_SIZE;
} else {
/* Older MessageSet */
msetw->msetw_MessageSetSize = RD_KAFKAP_MSGSET_V0_SIZE;
}
}
/**
* @brief Initialize a ProduceRequest MessageSet writer for
* the given broker and partition.
*
* A new buffer will be allocated to fit the pending messages in queue.
*
* @returns the number of messages to enqueue
*
* @remark This currently constructs the entire ProduceRequest, containing
* a single outer MessageSet for a single partition.
*
* @locality broker thread
*/
static int rd_kafka_msgset_writer_init(rd_kafka_msgset_writer_t *msetw,
rd_kafka_broker_t *rkb,
rd_kafka_toppar_t *rktp,
rd_kafka_msgq_t *rkmq,
rd_kafka_pid_t pid,
uint64_t epoch_base_msgid) {
int msgcnt = rd_kafka_msgq_len(rkmq);
if (msgcnt == 0)
return 0;
memset(msetw, 0, sizeof(*msetw));
msetw->msetw_rktp = rktp;
msetw->msetw_rkb = rkb;
msetw->msetw_msgq = rkmq;
msetw->msetw_pid = pid;
/* Max number of messages to send in a batch,
* limited by current queue size or configured batch size,
* whichever is lower. */
msetw->msetw_msgcntmax =
RD_MIN(msgcnt, rkb->rkb_rk->rk_conf.batch_num_messages);
rd_dassert(msetw->msetw_msgcntmax > 0);
/* Select MsgVersion to use */
if (rd_kafka_msgset_writer_select_MsgVersion(msetw) == -1)
return -1;
/* Allocate backing buffer */
rd_kafka_msgset_writer_alloc_buf(msetw);
/* Construct first part of Produce header + MessageSet header */
rd_kafka_msgset_writer_write_Produce_header(msetw);
/* The current buffer position is now where the first message
* is located.
* Record the current buffer position so it can be rewound later
* in case of compression. */
msetw->msetw_firstmsg.of =
rd_buf_write_pos(&msetw->msetw_rkbuf->rkbuf_buf);
rd_kafka_msgbatch_init(&msetw->msetw_rkbuf->rkbuf_u.Produce.batch, rktp,
pid, epoch_base_msgid);
msetw->msetw_batch = &msetw->msetw_rkbuf->rkbuf_u.Produce.batch;
return msetw->msetw_msgcntmax;
}
/**
* @brief Copy or link message payload to buffer.
*/
static RD_INLINE void
rd_kafka_msgset_writer_write_msg_payload(rd_kafka_msgset_writer_t *msetw,
const rd_kafka_msg_t *rkm,
void (*free_cb)(void *)) {
const rd_kafka_t *rk = msetw->msetw_rkb->rkb_rk;
rd_kafka_buf_t *rkbuf = msetw->msetw_rkbuf;
/* If payload is below the copy limit and there is still
* room in the buffer we'll copy the payload to the buffer,
* otherwise we push a reference to the memory. */
if (rkm->rkm_len <= (size_t)rk->rk_conf.msg_copy_max_size &&
rd_buf_write_remains(&rkbuf->rkbuf_buf) > rkm->rkm_len) {
rd_kafka_buf_write(rkbuf, rkm->rkm_payload, rkm->rkm_len);
if (free_cb)
free_cb(rkm->rkm_payload);
} else
rd_kafka_buf_push(rkbuf, rkm->rkm_payload, rkm->rkm_len,
free_cb);
}
/**
* @brief Write message headers to buffer.
*
* @remark The enveloping HeaderCount varint must already have been written.
* @returns the number of bytes written to msetw->msetw_rkbuf
*/
static size_t
rd_kafka_msgset_writer_write_msg_headers(rd_kafka_msgset_writer_t *msetw,
const rd_kafka_headers_t *hdrs) {
rd_kafka_buf_t *rkbuf = msetw->msetw_rkbuf;
const rd_kafka_header_t *hdr;
int i;
size_t start_pos = rd_buf_write_pos(&rkbuf->rkbuf_buf);
size_t written;
RD_LIST_FOREACH(hdr, &hdrs->rkhdrs_list, i) {
rd_kafka_buf_write_varint(rkbuf, hdr->rkhdr_name_size);
rd_kafka_buf_write(rkbuf, hdr->rkhdr_name,
hdr->rkhdr_name_size);
rd_kafka_buf_write_varint(
rkbuf,
hdr->rkhdr_value ? (int64_t)hdr->rkhdr_value_size : -1);
rd_kafka_buf_write(rkbuf, hdr->rkhdr_value,
hdr->rkhdr_value_size);
}
written = rd_buf_write_pos(&rkbuf->rkbuf_buf) - start_pos;
rd_dassert(written == hdrs->rkhdrs_ser_size);
return written;
}
/**
* @brief Write message to messageset buffer with MsgVersion 0 or 1.
* @returns the number of bytes written.
*/
static size_t
rd_kafka_msgset_writer_write_msg_v0_1(rd_kafka_msgset_writer_t *msetw,
rd_kafka_msg_t *rkm,
int64_t Offset,
int8_t MsgAttributes,
void (*free_cb)(void *)) {
rd_kafka_buf_t *rkbuf = msetw->msetw_rkbuf;
size_t MessageSize;
size_t of_Crc;
/*
* MessageSet's (v0 and v1) per-Message header.
*/
/* Offset (only relevant for compressed messages on MsgVersion v1) */
rd_kafka_buf_write_i64(rkbuf, Offset);
/* MessageSize */
MessageSize = 4 + 1 + 1 + /* Crc+MagicByte+Attributes */
4 /* KeyLength */ + rkm->rkm_key_len +
4 /* ValueLength */ + rkm->rkm_len;
if (msetw->msetw_MsgVersion == 1)
MessageSize += 8; /* Timestamp i64 */
rd_kafka_buf_write_i32(rkbuf, (int32_t)MessageSize);
/*
* Message
*/
/* Crc: will be updated later */
of_Crc = rd_kafka_buf_write_i32(rkbuf, 0);
/* Start Crc calculation of all buf writes. */
rd_kafka_buf_crc_init(rkbuf);
/* MagicByte */
rd_kafka_buf_write_i8(rkbuf, msetw->msetw_MsgVersion);
/* Attributes */
rd_kafka_buf_write_i8(rkbuf, MsgAttributes);
/* V1: Timestamp */
if (msetw->msetw_MsgVersion == 1)
rd_kafka_buf_write_i64(rkbuf, rkm->rkm_timestamp);
/* Message Key */
rd_kafka_buf_write_bytes(rkbuf, rkm->rkm_key, rkm->rkm_key_len);
/* Write or copy Value/payload */
if (rkm->rkm_payload) {
rd_kafka_buf_write_i32(rkbuf, (int32_t)rkm->rkm_len);
rd_kafka_msgset_writer_write_msg_payload(msetw, rkm, free_cb);
} else
rd_kafka_buf_write_i32(rkbuf, RD_KAFKAP_BYTES_LEN_NULL);
/* Finalize Crc */
rd_kafka_buf_update_u32(rkbuf, of_Crc,
rd_kafka_buf_crc_finalize(rkbuf));
/* Return written message size */
return 8 /*Offset*/ + 4 /*MessageSize*/ + MessageSize;
}
/**
* @brief Write message to messageset buffer with MsgVersion 2.
* @returns the number of bytes written.
*/
static size_t
rd_kafka_msgset_writer_write_msg_v2(rd_kafka_msgset_writer_t *msetw,
rd_kafka_msg_t *rkm,
int64_t Offset,
int8_t MsgAttributes,
void (*free_cb)(void *)) {
rd_kafka_buf_t *rkbuf = msetw->msetw_rkbuf;
size_t MessageSize = 0;
char varint_Length[RD_UVARINT_ENC_SIZEOF(int32_t)];
char varint_TimestampDelta[RD_UVARINT_ENC_SIZEOF(int64_t)];
char varint_OffsetDelta[RD_UVARINT_ENC_SIZEOF(int64_t)];
char varint_KeyLen[RD_UVARINT_ENC_SIZEOF(int32_t)];
char varint_ValueLen[RD_UVARINT_ENC_SIZEOF(int32_t)];
char varint_HeaderCount[RD_UVARINT_ENC_SIZEOF(int32_t)];
size_t sz_Length;
size_t sz_TimestampDelta;
size_t sz_OffsetDelta;
size_t sz_KeyLen;
size_t sz_ValueLen;
size_t sz_HeaderCount;
int HeaderCount = 0;
size_t HeaderSize = 0;
if (rkm->rkm_headers) {
HeaderCount = rkm->rkm_headers->rkhdrs_list.rl_cnt;
HeaderSize = rkm->rkm_headers->rkhdrs_ser_size;
}
/* All varints, except for Length, needs to be pre-built
* so that the Length field can be set correctly and thus have
* correct varint encoded width. */
sz_TimestampDelta = rd_uvarint_enc_i64(
varint_TimestampDelta, sizeof(varint_TimestampDelta),
rkm->rkm_timestamp - msetw->msetw_firstmsg.timestamp);
sz_OffsetDelta = rd_uvarint_enc_i64(varint_OffsetDelta,
sizeof(varint_OffsetDelta), Offset);
sz_KeyLen = rd_uvarint_enc_i32(varint_KeyLen, sizeof(varint_KeyLen),
rkm->rkm_key
? (int32_t)rkm->rkm_key_len
: (int32_t)RD_KAFKAP_BYTES_LEN_NULL);
sz_ValueLen = rd_uvarint_enc_i32(
varint_ValueLen, sizeof(varint_ValueLen),
rkm->rkm_payload ? (int32_t)rkm->rkm_len
: (int32_t)RD_KAFKAP_BYTES_LEN_NULL);
sz_HeaderCount =
rd_uvarint_enc_i32(varint_HeaderCount, sizeof(varint_HeaderCount),
(int32_t)HeaderCount);
/* Calculate MessageSize without length of Length (added later)
* to store it in Length. */
MessageSize = 1 /* MsgAttributes */ + sz_TimestampDelta +
sz_OffsetDelta + sz_KeyLen + rkm->rkm_key_len +
sz_ValueLen + rkm->rkm_len + sz_HeaderCount + HeaderSize;
/* Length */
sz_Length = rd_uvarint_enc_i64(varint_Length, sizeof(varint_Length),
MessageSize);
rd_kafka_buf_write(rkbuf, varint_Length, sz_Length);
MessageSize += sz_Length;
/* Attributes: The MsgAttributes argument is losely based on MsgVer0
* which don't apply for MsgVer2 */
rd_kafka_buf_write_i8(rkbuf, 0);
/* TimestampDelta */
rd_kafka_buf_write(rkbuf, varint_TimestampDelta, sz_TimestampDelta);
/* OffsetDelta */
rd_kafka_buf_write(rkbuf, varint_OffsetDelta, sz_OffsetDelta);
/* KeyLen */
rd_kafka_buf_write(rkbuf, varint_KeyLen, sz_KeyLen);
/* Key (if any) */
if (rkm->rkm_key)
rd_kafka_buf_write(rkbuf, rkm->rkm_key, rkm->rkm_key_len);
/* ValueLen */
rd_kafka_buf_write(rkbuf, varint_ValueLen, sz_ValueLen);
/* Write or copy Value/payload */
if (rkm->rkm_payload)
rd_kafka_msgset_writer_write_msg_payload(msetw, rkm, free_cb);
/* HeaderCount */
rd_kafka_buf_write(rkbuf, varint_HeaderCount, sz_HeaderCount);
/* Headers array */
if (rkm->rkm_headers)
rd_kafka_msgset_writer_write_msg_headers(msetw,
rkm->rkm_headers);
/* Return written message size */
return MessageSize;
}
/**
* @brief Write message to messageset buffer.
* @returns the number of bytes written.
*/
static size_t rd_kafka_msgset_writer_write_msg(rd_kafka_msgset_writer_t *msetw,
rd_kafka_msg_t *rkm,
int64_t Offset,
int8_t MsgAttributes,
void (*free_cb)(void *)) {
size_t outlen;
size_t (*writer[])(rd_kafka_msgset_writer_t *, rd_kafka_msg_t *,
int64_t, int8_t, void (*)(void *)) = {
[0] = rd_kafka_msgset_writer_write_msg_v0_1,
[1] = rd_kafka_msgset_writer_write_msg_v0_1,
[2] = rd_kafka_msgset_writer_write_msg_v2};
size_t actual_written;
size_t pre_pos;
if (likely(rkm->rkm_timestamp))
MsgAttributes |= RD_KAFKA_MSG_ATTR_CREATE_TIME;
pre_pos = rd_buf_write_pos(&msetw->msetw_rkbuf->rkbuf_buf);
outlen = writer[msetw->msetw_MsgVersion](msetw, rkm, Offset,
MsgAttributes, free_cb);
actual_written =
rd_buf_write_pos(&msetw->msetw_rkbuf->rkbuf_buf) - pre_pos;
rd_assert(outlen <=
rd_kafka_msg_wire_size(rkm, msetw->msetw_MsgVersion));
rd_assert(outlen == actual_written);
return outlen;
}
/**
* @brief Write as many messages from the given message queue to
* the messageset.
*
* May not write any messages.
*
* @returns 1 on success or 0 on error.
*/
static int rd_kafka_msgset_writer_write_msgq(rd_kafka_msgset_writer_t *msetw,
rd_kafka_msgq_t *rkmq) {
rd_kafka_toppar_t *rktp = msetw->msetw_rktp;
rd_kafka_broker_t *rkb = msetw->msetw_rkb;
size_t len = rd_buf_len(&msetw->msetw_rkbuf->rkbuf_buf);
size_t max_msg_size =
RD_MIN((size_t)msetw->msetw_rkb->rkb_rk->rk_conf.max_msg_size,
(size_t)msetw->msetw_rkb->rkb_rk->rk_conf.batch_size);
rd_ts_t int_latency_base;
rd_ts_t MaxTimestamp = 0;
rd_kafka_msg_t *rkm;
int msgcnt = 0;
const rd_ts_t now = rd_clock();
/* Internal latency calculation base.
* Uses rkm_ts_timeout which is enqueue time + timeout */
int_latency_base =
now + ((rd_ts_t)rktp->rktp_rkt->rkt_conf.message_timeout_ms * 1000);
/* Acquire BaseTimestamp from first message. */
rkm = TAILQ_FIRST(&rkmq->rkmq_msgs);
rd_kafka_assert(NULL, rkm);
msetw->msetw_firstmsg.timestamp = rkm->rkm_timestamp;
rd_kafka_msgbatch_set_first_msg(msetw->msetw_batch, rkm);
/*
* Write as many messages as possible until buffer is full
* or limit reached.
*/
do {
if (unlikely(msetw->msetw_batch->last_msgid &&
msetw->msetw_batch->last_msgid <
rkm->rkm_u.producer.msgid)) {
rd_rkb_dbg(rkb, MSG, "PRODUCE",
"%.*s [%" PRId32
"]: "
"Reconstructed MessageSet "
"(%d message(s), %" PRIusz
" bytes, "
"MsgIds %" PRIu64 "..%" PRIu64 ")",
RD_KAFKAP_STR_PR(rktp->rktp_rkt->rkt_topic),
rktp->rktp_partition, msgcnt, len,
msetw->msetw_batch->first_msgid,
msetw->msetw_batch->last_msgid);
break;
}
/* Check if there is enough space in the current messageset
* to add this message.
* Since calculating the total size of a request at produce()
* time is tricky (we don't know the protocol version or
* MsgVersion that will be used), we allow a messageset to
* overshoot the message.max.bytes limit by one message to
* avoid getting stuck here.
* The actual messageset size is enforced by the broker. */
if (unlikely(
msgcnt == msetw->msetw_msgcntmax ||
(msgcnt > 0 && len + rd_kafka_msg_wire_size(
rkm, msetw->msetw_MsgVersion) >
max_msg_size))) {
rd_rkb_dbg(rkb, MSG, "PRODUCE",
"%.*s [%" PRId32
"]: "
"No more space in current MessageSet "
"(%i message(s), %" PRIusz " bytes)",
RD_KAFKAP_STR_PR(rktp->rktp_rkt->rkt_topic),
rktp->rktp_partition, msgcnt, len);
break;
}
if (unlikely(rkm->rkm_u.producer.ts_backoff > now)) {
/* Stop accumulation when we've reached
* a message with a retry backoff in the future */
break;
}
/* Move message to buffer's queue */
rd_kafka_msgq_deq(rkmq, rkm, 1);
rd_kafka_msgq_enq(&msetw->msetw_batch->msgq, rkm);
msetw->msetw_messages_kvlen += rkm->rkm_len + rkm->rkm_key_len;
/* Add internal latency metrics */
rd_avg_add(&rkb->rkb_avg_int_latency,
int_latency_base - rkm->rkm_ts_timeout);
/* MessageSet v2's .MaxTimestamp field */
if (unlikely(MaxTimestamp < rkm->rkm_timestamp))
MaxTimestamp = rkm->rkm_timestamp;
/* Write message to buffer */
len += rd_kafka_msgset_writer_write_msg(msetw, rkm, msgcnt, 0,
NULL);
msgcnt++;
} while ((rkm = TAILQ_FIRST(&rkmq->rkmq_msgs)));
msetw->msetw_MaxTimestamp = MaxTimestamp;
/* Idempotent Producer:
* When reconstructing a batch to retry make sure
* the original message sequence span matches identically
* or we can't guarantee exactly-once delivery.
* If this check fails we raise a fatal error since
* it is unrecoverable and most likely caused by a bug
* in the client implementation.
* This should not be considered an abortable error for
* the transactional producer. */
if (msgcnt > 0 && msetw->msetw_batch->last_msgid) {
rd_kafka_msg_t *lastmsg;
lastmsg = rd_kafka_msgq_last(&msetw->msetw_batch->msgq);
rd_assert(lastmsg);
if (unlikely(lastmsg->rkm_u.producer.msgid !=
msetw->msetw_batch->last_msgid)) {
rd_kafka_set_fatal_error(
rkb->rkb_rk, RD_KAFKA_RESP_ERR__INCONSISTENT,
"Unable to reconstruct MessageSet "
"(currently with %d message(s)) "
"with msgid range %" PRIu64 "..%" PRIu64
": "
"last message added has msgid %" PRIu64
": "
"unable to guarantee consistency",
msgcnt, msetw->msetw_batch->first_msgid,
msetw->msetw_batch->last_msgid,
lastmsg->rkm_u.producer.msgid);
return 0;
}
}
return 1;
}
#if WITH_ZLIB
/**
* @brief Compress messageset using gzip/zlib
*/
static int rd_kafka_msgset_writer_compress_gzip(rd_kafka_msgset_writer_t *msetw,
rd_slice_t *slice,
struct iovec *ciov) {
rd_kafka_broker_t *rkb = msetw->msetw_rkb;
rd_kafka_toppar_t *rktp = msetw->msetw_rktp;
z_stream strm;
size_t len = rd_slice_remains(slice);
const void *p;
size_t rlen;
int r;
int comp_level =
msetw->msetw_rktp->rktp_rkt->rkt_conf.compression_level;
memset(&strm, 0, sizeof(strm));
r = deflateInit2(&strm, comp_level, Z_DEFLATED, 15 + 16, 8,
Z_DEFAULT_STRATEGY);
if (r != Z_OK) {
rd_rkb_log(rkb, LOG_ERR, "GZIP",
"Failed to initialize gzip for "
"compressing %" PRIusz
" bytes in "
"topic %.*s [%" PRId32
"]: %s (%i): "
"sending uncompressed",
len, RD_KAFKAP_STR_PR(rktp->rktp_rkt->rkt_topic),
rktp->rktp_partition, strm.msg ? strm.msg : "", r);
return -1;
}
/* Calculate maximum compressed size and
* allocate an output buffer accordingly, being
* prefixed with the Message header. */
ciov->iov_len = deflateBound(&strm, (uLong)rd_slice_remains(slice));
ciov->iov_base = rd_malloc(ciov->iov_len);
strm.next_out = (void *)ciov->iov_base;
strm.avail_out = (uInt)ciov->iov_len;
/* Iterate through each segment and compress it. */
while ((rlen = rd_slice_reader(slice, &p))) {
strm.next_in = (void *)p;
strm.avail_in = (uInt)rlen;
/* Compress message */
if ((r = deflate(&strm, Z_NO_FLUSH)) != Z_OK) {
rd_rkb_log(rkb, LOG_ERR, "GZIP",
"Failed to gzip-compress "
"%" PRIusz " bytes (%" PRIusz
" total) for "
"topic %.*s [%" PRId32
"]: "
"%s (%i): "