forked from apache/cassandra
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathnative_protocol_v5.spec
1417 lines (1187 loc) · 72 KB
/
native_protocol_v5.spec
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
CQL BINARY PROTOCOL v5
Table of Contents
1. Overview
2. Frame format
2.1. Uncompressed Format
2.2. Compressed Format
2.3 Protocol Negotiation
2.3.1 Initial Handshake
2.3.2 Compression
2.4. Frame Payload
2.4.1 Frame Header
2.4.1.1. version
2.4.1.2. flags
2.4.1.3. stream
2.4.1.4. opcode
2.4.1.5. length
3. Notations
4. Messages
4.1. Requests
4.1.1. STARTUP
4.1.2. AUTH_RESPONSE
4.1.3. OPTIONS
4.1.4. QUERY
4.1.5. PREPARE
4.1.6. EXECUTE
4.1.7. BATCH
4.1.8. REGISTER
4.2. Responses
4.2.1. ERROR
4.2.2. READY
4.2.3. AUTHENTICATE
4.2.4. SUPPORTED
4.2.5. RESULT
4.2.5.1. Void
4.2.5.2. Rows
4.2.5.3. Set_keyspace
4.2.5.4. Prepared
4.2.5.5. Schema_change
4.2.6. EVENT
4.2.7. AUTH_CHALLENGE
4.2.8. AUTH_SUCCESS
5. Data Type Serialization Formats
6. User Defined Type Serialization
7. Result paging
8. Error codes
9. Changes from v4
1. Overview
The CQL binary protocol is a frame based protocol with a frame comprises a header, payload
and trailer. In v5 there are two distinct frame formats, compressed and uncompressed, in
both cases, the payload is a stream of CQL envelopes (Section 2.4). Each envelope contains
a single CQL message, along with a metadata header. In effect, the v5 framing format is a
simple wrapper around protocol v5.
In either format, a frame may or may not be self contained. If self contained, then the
payload includes one or more complete envelopes and can be fully processed immediately.
Otherwise, the payload contains some part of a large envelope, which has been split into
its own sequence of frames. These are expected to be transmitted/received in order, so
the receiver can accumulate them as they arrive and process them once all have been received.
The frame header contains length information for the payload, a flag to indicate whether
or not the frame is self contained and a CRC24 to assert the integrity of the header itself.
There are slight variations in the header format between the compressed and uncompressed
variants.
The payload is opaque as far as the framing format is concerned, modulo the self
contained variation.
The trailer contains a CRC32 to protect the integrity of the payload, covering all envelopes
(whole or partial) contained therein.
2. Frame Format
2.1 Uncompressed Format
The uncompressed variant uses a 6 byte header containing payload length, self contained
flag and CRC24 for the header itself. The max size for the payload is 128KiB, and is
followed by its CRC32.
1. Payload length (17 bits)
2. isSelfContained flag (1 bit)
3. Header padding (6 bits)
4. CRC24 of the header (24 bits)
5. Payload (up to 2 ^ 17 - 1 bits)
6. Payload CRC32 (32 bits)
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Payload Length |C| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
CRC24 of Header | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +
| |
+ +
| Payload |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| CRC32 of Payload |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2.2 LZ4 Compressed Format
The variant with LZ4 compression uses an 8 byte header, containing both the compressed
and uncompressed lengths of the payload, the self contained flag and a CRC24 for the
header. As with uncompressed frames, the max payload size is 128KiB and is followed
by a CRC32 trailer. This is the CRC of the compressed payload.
1. Compressed length (17 bits)
2. Uncompressed length (17 bits)
3. isSelfContained flag (1 bit)
4. Header padding (5 bits)
5. CRC24 of Header contents (24 bits)
6. Compressed Payload (up to 2 ^ 17 - 1 bits)
7. CRC32 of Compressed Payload (32 bits)
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Compressed Length | Uncompressed Length
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|C| | CRC24 of Header |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
+ +
| Compressed Payload |
+ +
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| CRC32 of Compressed Payload |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2.3 Protocol Negotiation
2.3.1 Initial Handshake
In order to support both v5 and earlier formats, the v5 framing format is not
applied to message exchanges before an initial handshake is completed. Practically,
this means that the initial STARTUP message and any OPTIONS messages which precede
it are expected to be unframed. Likewise, the responses returned by the server,
SUPPORTED in response to OPTIONS and either READY or AUTHENTICATE in response to
STARTUP are transmitted unframed.
After sending the READY or AUTHENTICATE response to a STARTUP message, the server
will begin encoding and decoding all further transmissions according to the protocol
version of that STARTUP message. Compression of the frames is dictated by the
COMPRESSION option sent in the STARTUP message. Only LZ4 compression is currently
supported for v5.
Note: OPTIONS requests may be sent by the client at any time in the connection
lifecycle, both before and after the STARTUP exchange. As mentioned, those
transmitted before STARTUP, as well as the SUPPORTED responses the server returns
are unframed. Any OPTIONS/SUPPORTED exchanges after the STARTUP handshake are
formatted according to the negotiated protocol version, so for v5 these must be
framed.
2.3.2 Compression
Before being used, client and server must agree on a compression algorithm to
use, which is done in the STARTUP message. As a consequence, a STARTUP message
must never be compressed. However, once the STARTUP frame has been received
by the server, messages can be compressed (including the response to the STARTUP
request). Frames do not have to be compressed, however, even if compression has
been agreed upon (a sender may only compress frames above a certain size at its
discretion). Where compression has been agreed, the sender signals that the payload
is not compressed by setting the compressed length to 0.
As of v5 of the protocol, the only compression available is lz4
(https://code.google.com/p/lz4/).
2.4. Frame Payload
Envelopes are defined as:
0 8 16 24 32 40
+---------+---------+---------+---------+---------+
| version | flags | stream | opcode |
+---------+---------+---------+---------+---------+
| length |
+---------+---------+---------+---------+
| |
. ... body ... .
. .
. .
+----------------------------------------
The protocol is big-endian (network byte order).
Each envelope contains a fixed size header (9 bytes) followed by a variable size
body. The header is described in Section 2.4.1. The content of the body depends
on the header opcode value (the body can in particular be empty for some
opcode values). The list of allowed opcodes is defined in Section 2.4.1.4 and the
details of each corresponding message are described Section 4.
The protocol distinguishes two types of envelope: requests and responses. Requests
are those envelopes sent by the client to the server. Responses are those envelopes sent
by the server to the client. Note, however, that the protocol supports server pushes
(events) so a response does not necessarily come right after a client request.
Note to client implementors: client libraries should always assume that the
body of a given envelope may contain more data than what is described in this
document. It will however always be safe to ignore the remainder of the body
in such cases. The reason is that this may enable extending the protocol
with optional features without needing to change the protocol version.
Envelope headers are designed to support backwards compatibility with earlier
protocol versions. For that reason, they include an unused leading byte in place
of the version field from previous protocol versions. This was always to some extent
redundant as the version is set and enforced at the connection level. It was also
previously possible to enable compression for an individual envelope. This is no
longer possible, as the framing format is responsible for compression, which is set for
the lifetime of a connection and applies to all messages transmitted throughout it
(see Section 2.2.1 for caveats). The compression flag is therefore deprecated and
ignored in protocol v5.
2.4.1.1. version
The version is a single byte that indicates both the direction of the message
(request or response) and the version of the protocol in use. The most
significant bit of version is used to define the direction of the message:
0 indicates a request, 1 indicates a response. This can be useful for protocol
analyzers to distinguish the nature of the packet from the direction in which
it is moving. The rest of that byte is the protocol version (5 for the protocol
defined in this document). In other words, for this version of the protocol,
version will be one of:
0x05 Request frame for this protocol version
0x85 Response frame for this protocol version
Please note that while every message ships with the version, only one version
of messages is accepted on a given connection. In other words, the first message
exchanged (STARTUP) sets the version for the connection for the lifetime of this
connection.
This document describes version 5 of the protocol. For the changes made since
version 4, see Section 10.
2.4.1.2. flags
Flags applying to this envelope. The flags have the following meaning (described
by the mask that allows selecting them):
0x01: Compression flag. In protocol v5 this flag is deprecated and ignored.
0x02: Tracing flag. For a request, this indicates the client requires tracing
of the request. Note that only QUERY, PREPARE and EXECUTE queries
support tracing. Other requests will simply ignore the tracing flag if
set. If a request supports tracing and the tracing flag is set, the
response to this request will have the tracing flag set and contain
tracing information.
If a response has the tracing flag set, its body contains a tracing ID.
The tracing ID is a [uuid] and is the first thing in the body.
0x04: Custom payload flag. For a request or response, this indicates that a
generic key-value custom payload for a custom QueryHandler implementation
is present. Such a custom payload is simply ignored by the default
QueryHandler implementation. Currently, only QUERY, PREPARE, EXECUTE and
BATCH requests support custom payloads.
Type of custom payload is [bytes map] (see below). If either or both
of the tracing and warning flags are set, the custom payload will follow
those indicated elements in the body. If neither are set, the custom
payload will be the first value in the body.
0x08: Warning flag. The response contains warnings which were generated by the
server to go along with this response.
If a response has the warning flag set, its body will contain the text of
the warnings. The warnings are a [string list] and will be the first value
in the body if the tracing flag is not set, or directly after the tracing
ID if it is.
0x10: Use beta flag. Indicates that the client opts in to use protocol version
that is currently in beta. Server will respond with ERROR if protocol
version is marked as beta on server and client does not provide this flag.
The rest of flags is currently unused and ignored.
2.4.1.3. stream
An envelope has a stream id (a [short] value). When sending request messages, this
stream id must be set by the client to a non-negative value (negative stream id
are reserved for streams initiated by the server; currently all EVENT messages
(section 4.2.6) have a streamId of -1). If a client sends a request message
with the stream id X, it is guaranteed that the stream id of the response to
that message will be X.
This helps to enable the asynchronous nature of the protocol. If a client
sends multiple messages simultaneously (without waiting for responses), there
is no guarantee on the order of the responses. For instance, if the client
writes REQ_1, REQ_2, REQ_3 on the wire (in that order), the server might
respond to REQ_3 (or REQ_2) first. Assigning different stream ids to these 3
requests allows the client to distinguish to which request a received answer
responds to. As there can only be 32768 different simultaneous streams, it is up
to the client to reuse stream id.
Note that clients are free to use the protocol synchronously (i.e. wait for
the response to REQ_N before sending REQ_N+1). In that case, the stream id
can be safely set to 0 as long as each frame contains only a single envelope.
Clients should also feel free to use only a subset of the 32768 maximum possible stream
ids if it is simpler for its implementation.
2.4.1.4. opcode
An integer byte that distinguishes the actual message:
0x00 ERROR
0x01 STARTUP
0x02 READY
0x03 AUTHENTICATE
0x05 OPTIONS
0x06 SUPPORTED
0x07 QUERY
0x08 RESULT
0x09 PREPARE
0x0A EXECUTE
0x0B REGISTER
0x0C EVENT
0x0D BATCH
0x0E AUTH_CHALLENGE
0x0F AUTH_RESPONSE
0x10 AUTH_SUCCESS
Messages are described in Section 4.
(Note that there is no 0x04 message in this version of the protocol)
2.4.1.5. length
A 4 byte integer representing the length of the body of the envelope (note:
currently an envelope body is limited to 256MB in length).
3. Notations
To describe the layout of the envelope body for the messages in Section 4, we
define the following:
[int] A 4 bytes integer
[long] A 8 bytes integer
[byte] A 1 byte unsigned integer
[short] A 2 bytes unsigned integer
[string] A [short] n, followed by n bytes representing an UTF-8
string.
[long string] An [int] n, followed by n bytes representing an UTF-8 string.
[uuid] A 16 bytes long uuid.
[string list] A [short] n, followed by n [string].
[bytes] A [int] n, followed by n bytes if n >= 0. If n < 0,
no byte should follow and the value represented is `null`.
[value] A [int] n, followed by n bytes if n >= 0.
If n == -1 no byte should follow and the value represented is `null`.
If n == -2 no byte should follow and the value represented is
`not set` not resulting in any change to the existing value.
n < -2 is an invalid value and results in an error.
[short bytes] A [short] n, followed by n bytes if n >= 0.
[unsigned vint] An unsigned variable length integer. A vint is encoded with the most significant byte (MSB) first.
The most significant byte will contains the information about how many extra bytes need to be read
as well as the most significant bits of the integer.
The number of extra bytes to read is encoded as 1 bits on the left side.
For example, if we need to read 2 more bytes the first byte will start with 110
(e.g. 256 000 will be encoded on 3 bytes as [110]00011 11101000 00000000)
If the encoded integer is 8 bytes long the vint will be encoded on 9 bytes and the first
byte will be: 11111111
[vint] A signed variable length integer. This is encoded using zig-zag encoding and then sent
like an [unsigned vint]. Zig-zag encoding converts numbers as follows:
0 = 0, -1 = 1, 1 = 2, -2 = 3, 2 = 4, -3 = 5, 3 = 6 and so forth.
The purpose is to send small negative values as small unsigned values, so that we save bytes on the wire.
To encode a value n use "(n >> 31) ^ (n << 1)" for 32 bit values, and "(n >> 63) ^ (n << 1)"
for 64 bit values where "^" is the xor operation, "<<" is the left shift operation and ">>" is
the arithemtic right shift operation (highest-order bit is replicated).
Decode with "(n >> 1) ^ -(n & 1)".
[option] A pair of <id><value> where <id> is a [short] representing
the option id and <value> depends on that option (and can be
of size 0). The supported id (and the corresponding <value>)
will be described when this is used.
[option list] A [short] n, followed by n [option].
[inet] An address (ip and port) to a node. It consists of one
[byte] n, that represents the address size, followed by n
[byte] representing the IP address (in practice n can only be
either 4 (IPv4) or 16 (IPv6)), following by one [int]
representing the port.
[inetaddr] An IP address (without a port) to a node. It consists of one
[byte] n, that represents the address size, followed by n
[byte] representing the IP address.
[consistency] A consistency level specification. This is a [short]
representing a consistency level with the following
correspondance:
0x0000 ANY
0x0001 ONE
0x0002 TWO
0x0003 THREE
0x0004 QUORUM
0x0005 ALL
0x0006 LOCAL_QUORUM
0x0007 EACH_QUORUM
0x0008 SERIAL
0x0009 LOCAL_SERIAL
0x000A LOCAL_ONE
[string map] A [short] n, followed by n pair <k><v> where <k> and <v>
are [string].
[string multimap] A [short] n, followed by n pair <k><v> where <k> is a
[string] and <v> is a [string list].
[bytes map] A [short] n, followed by n pair <k><v> where <k> is a
[string] and <v> is a [bytes].
4. Messages
Dependant on the flags specified in the header, the layout of the message body must be:
[<tracing_id>][<warnings>][<custom_payload>]<message>
where:
- <tracing_id> is a UUID tracing ID, present if this is a request message and the Tracing flag is set.
- <warnings> is a string list of warnings (if this is a request message and the Warning flag is set.
- <custom_payload> is bytes map for the serialised custom payload present if this is one of the message types
which support custom payloads (QUERY, PREPARE, EXECUTE and BATCH) and the Custom payload flag is set.
- <message> as defined below through sections 4 and 5.
4.1. Requests
Note that outside of their normal responses (described below), all requests
can get an ERROR message (Section 4.2.1) as response.
4.1.1. STARTUP
Initialize the connection. The server will respond by either a READY message
(in which case the connection is ready for queries) or an AUTHENTICATE message
(in which case credentials will need to be provided using AUTH_RESPONSE).
This must be the first message of the connection, except for OPTIONS that can
be sent before to find out the options supported by the server. Once the
connection has been initialized, a client should not send any more STARTUP
messages.
The body is a [string map] of options. Possible options are:
- "CQL_VERSION": the version of CQL to use. This option is mandatory and
currently the only version supported is "3.0.0". Note that this is
different from the protocol version.
- "COMPRESSION": the compression algorithm to use for frames (See section 2.3.2).
This is optional; if not specified no compression will be used.
- "DRIVER_NAME": allows clients to supply a free-form label representing the driver
implementation. This is displayed in the output of `nodetool clientstats`
- "DRIVER_VERSION": allows clients to supply a free-form label represting the driver
version. This is displayed in the output of `nodetool clientstats`
- "THROW_ON_OVERLOAD": flag to specify server behaviour where the incoming message
rate is too high. An [string] value of "1" instructs the server to respond with
and Error when its resources are exhausted. Any other value, or if the the key
is not present, and the server will apply backpressure to the connection until it
has cleared its backlog of inbound messages.
As mentioned in Section 2.3, STARTUP messages must not be sent in the framed format. STARTUP,
any OPTIONS requests which precede them, as well as the server's responses to those messages
must be unframed to support protocol negotiation with older clients.
4.1.2. AUTH_RESPONSE
Answers a server authentication challenge.
Authentication in the protocol is SASL based. The server sends authentication
challenges (a bytes token) to which the client answers with this message. Those
exchanges continue until the server accepts the authentication by sending a
AUTH_SUCCESS message after a client AUTH_RESPONSE. Note that the exchange
begins with the client sending an initial AUTH_RESPONSE in response to a
server AUTHENTICATE request.
The body of this message is a single [bytes] token. The details of what this
token contains (and when it can be null/empty, if ever) depends on the actual
authenticator used.
The response to a AUTH_RESPONSE is either a follow-up AUTH_CHALLENGE message,
an AUTH_SUCCESS message or an ERROR message.
4.1.3. OPTIONS
Asks the server to return which STARTUP options are supported. The body of an
OPTIONS message should be empty and the server will respond with a SUPPORTED
message.
4.1.4. QUERY
Performs a CQL query. The body of the message must be:
<query><query_parameters>
where <query> is a [long string] representing the query and
<query_parameters> must be
<consistency><flags>[<n>[name_1]<value_1>...[name_n]<value_n>][<result_page_size>][<paging_state>][<serial_consistency>][<timestamp>][<keyspace>][<now_in_seconds>]
where:
- <consistency> is the [consistency] level for the operation.
- <flags> is a [int] whose bits define the options for this query and
in particular influence what the remainder of the message contains.
A flag is set if the bit corresponding to its `mask` is set. Supported
flags are, given their mask:
0x0001: Values. If set, a [short] <n> followed by <n> [value]
values are provided. Those values are used for bound variables in
the query. Optionally, if the 0x40 flag is present, each value
will be preceded by a [string] name, representing the name of
the marker the value must be bound to.
0x0002: Skip_metadata. If set, the Result Set returned as a response
to the query (if any) will have the NO_METADATA flag (see
Section 4.2.5.2).
0x0004: Page_size. If set, <result_page_size> is an [int]
controlling the desired page size of the result (in CQL3 rows).
See the section on paging (Section 7) for more details.
0x0008: With_paging_state. If set, <paging_state> should be present.
<paging_state> is a [bytes] value that should have been returned
in a result set (Section 4.2.5.2). The query will be
executed but starting from a given paging state. This is also to
continue paging on a different node than the one where it
started (See Section 7 for more details).
0x0010: With serial consistency. If set, <serial_consistency> should be
present. <serial_consistency> is the [consistency] level for the
serial phase of conditional updates. That consitency can only be
either SERIAL or LOCAL_SERIAL and if not present, it defaults to
SERIAL. This option will be ignored for anything else other than a
conditional update/insert.
0x0020: With default timestamp. If set, <timestamp> must be present.
<timestamp> is a [long] representing the default timestamp for the query
in microseconds (negative values are forbidden). This will
replace the server side assigned timestamp as default timestamp.
Note that a timestamp in the query itself will still override
this timestamp. This is entirely optional.
0x0040: With names for values. This only makes sense if the 0x01 flag is set and
is ignored otherwise. If present, the values from the 0x01 flag will
be preceded by a name (see above). Note that this is only useful for
QUERY requests where named bind markers are used; for EXECUTE statements,
since the names for the expected values was returned during preparation,
a client can always provide values in the right order without any names
and using this flag, while supported, is almost surely inefficient.
0x0080: With keyspace. If set, <keyspace> must be present. <keyspace> is a
[string] indicating the keyspace that the query should be executed in.
It supercedes the keyspace that the connection is bound to, if any.
0x0100: With now in seconds. If set, <now_in_seconds> must be present.
<now_in_seconds> is an [int] representing the current time (now) for
the query. Affects TTL cell liveness in read queries and local deletion
time for tombstones and TTL cells in update requests. It's intended
for testing purposes and is optional.
Note that the consistency is ignored by some queries (USE, CREATE, ALTER,
TRUNCATE, ...).
The server will respond to a QUERY message with a RESULT message, the content
of which depends on the query.
4.1.5. PREPARE
Prepare a query for later execution (through EXECUTE). The body of the message must be:
<query><flags>[<keyspace>]
where:
- <query> is a [long string] representing the CQL query.
- <flags> is a [int] whose bits define the options for this statement and in particular
influence what the remainder of the message contains.
A flag is set if the bit corresponding to its `mask` is set. Supported
flags are, given their mask:
0x01: With keyspace. If set, <keyspace> must be present. <keyspace> is a
[string] indicating the keyspace that the query should be executed in.
It supercedes the keyspace that the connection is bound to, if any.
The server will respond with a RESULT message with a `prepared` kind (0x0004,
see Section 4.2.5).
4.1.6. EXECUTE
Executes a prepared query. The body of the message must be:
<id><result_metadata_id><query_parameters>
where
- <id> is the prepared query ID. It's the [short bytes] returned as a
response to a PREPARE message.
- <result_metadata_id> is the ID of the resultset metadata that was sent
along with response to PREPARE message. If a RESULT/Rows message reports
changed resultset metadata with the Metadata_changed flag, the reported new
resultset metadata must be used in subsequent executions.
- <query_parameters> has the exact same definition as in QUERY (see Section 4.1.4).
4.1.7. BATCH
Allows executing a list of queries (prepared or not) as a batch (note that
only DML statements are accepted in a batch). The body of the message must
be:
<type><n><query_1>...<query_n><consistency><flags>[<serial_consistency>][<timestamp>][<keyspace>][<now_in_seconds>]
where:
- <type> is a [byte] indicating the type of batch to use:
- If <type> == 0, the batch will be "logged". This is equivalent to a
normal CQL3 batch statement.
- If <type> == 1, the batch will be "unlogged".
- If <type> == 2, the batch will be a "counter" batch (and non-counter
statements will be rejected).
- <flags> is a [int] whose bits define the options for this query and
in particular influence what the remainder of the message contains. It is similar
to the <flags> from QUERY and EXECUTE methods, except that the 4 rightmost
bits must always be 0 as their corresponding options do not make sense for
Batch. A flag is set if the bit corresponding to its `mask` is set. Supported
flags are, given their mask:
0x0010: With serial consistency. If set, <serial_consistency> should be
present. <serial_consistency> is the [consistency] level for the
serial phase of conditional updates. That consistency can only be
either SERIAL or LOCAL_SERIAL and if not present, it defaults to
SERIAL. This option will be ignored for anything else other than a
conditional update/insert.
0x0020: With default timestamp. If set, <timestamp> should be present.
<timestamp> is a [long] representing the default timestamp for the query
in microseconds. This will replace the server side assigned
timestamp as default timestamp. Note that a timestamp in the query itself
will still override this timestamp. This is entirely optional.
0x0040: With names for values. If set, then all values for all <query_i> must be
preceded by a [string] <name_i> that have the same meaning as in QUERY
requests [IMPORTANT NOTE: this feature does not work and should not be
used. It is specified in a way that makes it impossible for the server
to implement. This will be fixed in a future version of the native
protocol. See https://issues.apache.org/jira/browse/CASSANDRA-10246 for
more details].
0x0080: With keyspace. If set, <keyspace> must be present. <keyspace> is a
[string] indicating the keyspace that the query should be executed in.
It supercedes the keyspace that the connection is bound to, if any.
0x0100: With now in seconds. If set, <now_in_seconds> must be present.
<now_in_seconds> is an [int] representing the current time (now) for
the query. Affects TTL cell liveness in read queries and local deletion
time for tombstones and TTL cells in update requests. It's intended
for testing purposes and is optional.
- <n> is a [short] indicating the number of following queries.
- <query_1>...<query_n> are the queries to execute. A <query_i> must be of the
form:
<kind><string_or_id><n>[<name_1>]<value_1>...[<name_n>]<value_n>
where:
- <kind> is a [byte] indicating whether the following query is a prepared
one or not. <kind> value must be either 0 or 1.
- <string_or_id> depends on the value of <kind>. If <kind> == 0, it should be
a [long string] query string (as in QUERY, the query string might contain
bind markers). Otherwise (that is, if <kind> == 1), it should be a
[short bytes] representing a prepared query ID.
- <n> is a [short] indicating the number (possibly 0) of following values.
- <name_i> is the optional name of the following <value_i>. It must be present
if and only if the 0x40 flag is provided for the batch.
- <value_i> is the [value] to use for bound variable i (of bound variable <name_i>
if the 0x40 flag is used).
- <consistency> is the [consistency] level for the operation.
- <serial_consistency> is only present if the 0x10 flag is set. In that case,
<serial_consistency> is the [consistency] level for the serial phase of
conditional updates. That consitency can only be either SERIAL or
LOCAL_SERIAL and if not present will defaults to SERIAL. This option will
be ignored for anything else other than a conditional update/insert.
The server will respond with a RESULT message.
4.1.8. REGISTER
Register this connection to receive some types of events. The body of the
message is a [string list] representing the event types to register for. See
section 4.2.6 for the list of valid event types.
The response to a REGISTER message will be a READY message.
Please note that if a client driver maintains multiple connections to a
Cassandra node and/or connections to multiple nodes, it is advised to
dedicate a handful of connections to receive events, but to *not* register
for events on all connections, as this would only result in receiving
multiple times the same event messages, wasting bandwidth.
4.2. Responses
This section describes the content of the frame body for the different
responses. Please note that to make room for future evolution, clients should
support extra informations (that they should simply discard) to the one
described in this document at the end of the frame body.
4.2.1. ERROR
Indicates an error processing a request. The body of the message will be an
error code ([int]) followed by a [string] error message. Then, depending on
the exception, more content may follow. The error codes are defined in
Section 8, along with their additional content if any.
4.2.2. READY
Indicates that the server is ready to process queries. This message will be
sent by the server either after a STARTUP message if no authentication is
required (if authentication is required, the server indicates readiness by
sending a AUTH_RESPONSE message).
The body of a READY message is empty.
4.2.3. AUTHENTICATE
Indicates that the server requires authentication, and which authentication
mechanism to use.
The authentication is SASL based and thus consists of a number of server
challenges (AUTH_CHALLENGE, Section 4.2.7) followed by client responses
(AUTH_RESPONSE, Section 4.1.2). The initial exchange is however boostrapped
by an initial client response. The details of that exchange (including how
many challenge-response pairs are required) are specific to the authenticator
in use. The exchange ends when the server sends an AUTH_SUCCESS message or
an ERROR message.
This message will be sent following a STARTUP message if authentication is
required and must be answered by a AUTH_RESPONSE message from the client.
The body consists of a single [string] indicating the full class name of the
IAuthenticator in use.
4.2.4. SUPPORTED
Indicates which startup options are supported by the server. This message
comes as a response to an OPTIONS message.
The body of a SUPPORTED message is a [string multimap]. This multimap gives
for each of the supported STARTUP options, the list of supported values. It
also includes:
- "PROTOCOL_VERSIONS": the list of native protocol versions that are
supported, encoded as the version number followed by a slash and the
version description. For example: 3/v3, 4/v4, 5/v5-beta. If a version is
in beta, it will have the word "beta" in its description.
4.2.5. RESULT
The result to a query (QUERY, PREPARE, EXECUTE or BATCH messages).
The first element of the body of a RESULT message is an [int] representing the
`kind` of result. The rest of the body depends on the kind. The kind can be
one of:
0x0001 Void: for results carrying no information.
0x0002 Rows: for results to select queries, returning a set of rows.
0x0003 Set_keyspace: the result to a `use` query.
0x0004 Prepared: result to a PREPARE message.
0x0005 Schema_change: the result to a schema altering query.
The body for each kind (after the [int] kind) is defined below.
4.2.5.1. Void
The rest of the body for a Void result is empty. It indicates that a query was
successful without providing more information.
4.2.5.2. Rows
Indicates a set of rows. The rest of the body of a Rows result is:
<metadata><rows_count><rows_content>
where:
- <metadata> is composed of:
<flags><columns_count>[<paging_state>][<new_metadata_id>][<global_table_spec>?<col_spec_1>...<col_spec_n>]
where:
- <flags> is an [int]. The bits of <flags> provides information on the
formatting of the remaining information. A flag is set if the bit
corresponding to its `mask` is set. Supported flags are, given their
mask:
0x0001 Global_tables_spec: if set, only one table spec (keyspace
and table name) is provided as <global_table_spec>. If not
set, <global_table_spec> is not present.
0x0002 Has_more_pages: indicates whether this is not the last
page of results and more should be retrieved. If set, the
<paging_state> will be present. The <paging_state> is a
[bytes] value that should be used in QUERY/EXECUTE to
continue paging and retrieve the remainder of the result for
this query (See Section 7 for more details).
0x0004 No_metadata: if set, the <metadata> is only composed of
these <flags>, the <column_count> and optionally the
<paging_state> (depending on the Has_more_pages flag) but
no other information (so no <global_table_spec> nor <col_spec_i>).
This will only ever be the case if this was requested
during the query (see QUERY and RESULT messages).
0x0008 Metadata_changed: if set, the No_metadata flag has to be unset
and <new_metadata_id> has to be supplied. This flag is to be
used to avoid a roundtrip in case of metadata changes for queries
that requested metadata to be skipped.
- <columns_count> is an [int] representing the number of columns selected
by the query that produced this result. It defines the number of <col_spec_i>
elements in and the number of elements for each row in <rows_content>.
- <new_metadata_id> is [short bytes] representing the new, changed resultset
metadata. The new metadata ID must also be used in subsequent executions of
the corresponding prepared statement, if any.
- <global_table_spec> is present if the Global_tables_spec is set in
<flags>. It is composed of two [string] representing the
(unique) keyspace name and table name the columns belong to.
- <col_spec_i> specifies the columns returned in the query. There are
<column_count> such column specifications that are composed of:
(<ksname><tablename>)?<name><type>
The initial <ksname> and <tablename> are two [string] and are only present
if the Global_tables_spec flag is not set. The <column_name> is a
[string] and <type> is an [option] that corresponds to the description
(what this description is depends a bit on the context: in results to
selects, this will be either the user chosen alias or the selection used
(often a colum name, but it can be a function call too). In results to
a PREPARE, this will be either the name of the corresponding bind variable
or the column name for the variable if it is "anonymous") and type of
the corresponding result. The option for <type> is either a native
type (see below), in which case the option has no value, or a
'custom' type, in which case the value is a [string] representing
the fully qualified class name of the type represented. Valid option
ids are:
0x0000 Custom: the value is a [string], see above.
0x0001 Ascii
0x0002 Bigint
0x0003 Blob
0x0004 Boolean
0x0005 Counter
0x0006 Decimal
0x0007 Double
0x0008 Float
0x0009 Int
0x000B Timestamp
0x000C Uuid
0x000D Varchar
0x000E Varint
0x000F Timeuuid
0x0010 Inet
0x0011 Date
0x0012 Time
0x0013 Smallint
0x0014 Tinyint
0x0015 Duration
0x0020 List: the value is an [option], representing the type
of the elements of the list.
0x0021 Map: the value is two [option], representing the types of the
keys and values of the map
0x0022 Set: the value is an [option], representing the type
of the elements of the set
0x0030 UDT: the value is <ks><udt_name><n><name_1><type_1>...<name_n><type_n>
where:
- <ks> is a [string] representing the keyspace name this
UDT is part of.
- <udt_name> is a [string] representing the UDT name.
- <n> is a [short] representing the number of fields of
the UDT, and thus the number of <name_i><type_i> pairs
following
- <name_i> is a [string] representing the name of the
i_th field of the UDT.
- <type_i> is an [option] representing the type of the
i_th field of the UDT.
0x0031 Tuple: the value is <n><type_1>...<type_n> where <n> is a [short]
representing the number of values in the type, and <type_i>
are [option] representing the type of the i_th component
of the tuple
- <rows_count> is an [int] representing the number of rows present in this
result. Those rows are serialized in the <rows_content> part.
- <rows_content> is composed of <row_1>...<row_m> where m is <rows_count>.
Each <row_i> is composed of <value_1>...<value_n> where n is
<columns_count> and where <value_j> is a [bytes] representing the value
returned for the jth column of the ith row. In other words, <rows_content>
is composed of (<rows_count> * <columns_count>) [bytes].
4.2.5.3. Set_keyspace
The result to a `use` query. The body (after the kind [int]) is a single
[string] indicating the name of the keyspace that has been set.
4.2.5.4. Prepared
The result to a PREPARE message. The body of a Prepared result is:
<id><result_metadata_id><metadata><result_metadata>
where:
- <id> is [short bytes] representing the prepared query ID.
- <result_metadata_id> is [short bytes] representing the resultset metadata ID.
- <metadata> is composed of:
<flags><columns_count><pk_count>[<pk_index_1>...<pk_index_n>][<global_table_spec>?<col_spec_1>...<col_spec_n>]
where:
- <flags> is an [int]. The bits of <flags> provides information on the
formatting of the remaining information. A flag is set if the bit
corresponding to its `mask` is set. Supported masks and their flags
are:
0x0001 Global_tables_spec: if set, only one table spec (keyspace
and table name) is provided as <global_table_spec>. If not
set, <global_table_spec> is not present.
- <columns_count> is an [int] representing the number of bind markers
in the prepared statement. It defines the number of <col_spec_i>
elements.
- <pk_count> is an [int] representing the number of <pk_index_i>
elements to follow. If this value is zero, at least one of the
partition key columns in the table that the statement acts on
did not have a corresponding bind marker (or the bind marker
was wrapped in a function call).
- <pk_index_i> is a short that represents the index of the bind marker
that corresponds to the partition key column in position i.
For example, a <pk_index> sequence of [2, 0, 1] indicates that the
table has three partition key columns; the full partition key
can be constructed by creating a composite of the values for
the bind markers at index 2, at index 0, and at index 1.
This allows implementations with token-aware routing to correctly
construct the partition key without needing to inspect table
metadata.
- <global_table_spec> is present if the Global_tables_spec is set in
<flags>. If present, it is composed of two [string]s. The first
[string] is the name of the keyspace that the statement acts on.
The second [string] is the name of the table that the columns
represented by the bind markers belong to.
- <col_spec_i> specifies the bind markers in the prepared statement.
There are <column_count> such column specifications, each with the
following format:
(<ksname><tablename>)?<name><type>
The initial <ksname> and <tablename> are two [string] that are only
present if the Global_tables_spec flag is not set. The <name> field
is a [string] that holds the name of the bind marker (if named),
or the name of the column, field, or expression that the bind marker
corresponds to (if the bind marker is "anonymous"). The <type>
field is an [option] that represents the expected type of values for
the bind marker. See the Rows documentation (section 4.2.5.2) for
full details on the <type> field.
- <result_metadata> is defined exactly the same as <metadata> in the Rows
documentation (section 4.2.5.2). This describes the metadata for the
result set that will be returned when this prepared statement is executed.
Note that <result_metadata> may be empty (have the No_metadata flag and
0 columns, See section 4.2.5.2) and will be for any query that is not a
Select. In fact, there is never a guarantee that this will be non-empty, so
implementations should protect themselves accordingly. This result metadata
is an optimization that allows implementations to later execute the
prepared statement without requesting the metadata (see the Skip_metadata
flag in EXECUTE). Clients can safely discard this metadata if they do not
want to take advantage of that optimization.
Note that the prepared query ID returned is global to the node on which the query
has been prepared. It can be used on any connection to that node
until the node is restarted (after which the query must be reprepared).
4.2.5.5. Schema_change
The result to a schema altering query (creation/update/drop of a
keyspace/table/index). The body (after the kind [int]) is the same
as the body for a "SCHEMA_CHANGE" event, so 3 strings:
<change_type><target><options>
Please refer to section 4.2.6 below for the meaning of those fields.
Note that a query to create or drop an index is considered to be a change
to the table the index is on.
4.2.6. EVENT
An event pushed by the server. A client will only receive events for the
types it has REGISTERed to. The body of an EVENT message will start with a
[string] representing the event type. The rest of the message depends on the
event type. The valid event types are:
- "TOPOLOGY_CHANGE": events related to change in the cluster topology.
Currently, events are sent when new nodes are added to the cluster, and
when nodes are removed. The body of the message (after the event type)
consists of a [string] and an [inet], corresponding respectively to the
type of change ("NEW_NODE" or "REMOVED_NODE") followed by the address of
the new/removed node.
- "STATUS_CHANGE": events related to change of node status. Currently,
up/down events are sent. The body of the message (after the event type)
consists of a [string] and an [inet], corresponding respectively to the
type of status change ("UP" or "DOWN") followed by the address of the
concerned node.
- "SCHEMA_CHANGE": events related to schema change. After the event type,
the rest of the message will be <change_type><target><options> where:
- <change_type> is a [string] representing the type of changed involved.
It will be one of "CREATED", "UPDATED" or "DROPPED".
- <target> is a [string] that can be one of "KEYSPACE", "TABLE", "TYPE",
"FUNCTION" or "AGGREGATE" and describes what has been modified
("TYPE" stands for modifications related to user types, "FUNCTION"
for modifications related to user defined functions, "AGGREGATE"
for modifications related to user defined aggregates).
- <options> depends on the preceding <target>:
- If <target> is "KEYSPACE", then <options> will be a single [string]
representing the keyspace changed.
- If <target> is "TABLE" or "TYPE", then
<options> will be 2 [string]: the first one will be the keyspace
containing the affected object, and the second one will be the name
of said affected object (either the table, user type, function, or
aggregate name).
- If <target> is "FUNCTION" or "AGGREGATE", multiple arguments follow:
- [string] keyspace containing the user defined function / aggregate
- [string] the function/aggregate name
- [string list] one string for each argument type (as CQL type)
All EVENT messages have a streamId of -1 (Section 2.4.1.3).
Please note that "NEW_NODE" and "UP" events are sent based on internal Gossip
communication and as such may be sent a short delay before the binary
protocol server on the newly up node is fully started. Clients are thus
advised to wait a short time before trying to connect to the node (1 second
should be enough), otherwise they may experience a connection refusal at
first.
4.2.7. AUTH_CHALLENGE
A server authentication challenge (see AUTH_RESPONSE (Section 4.1.2) for more
details).
The body of this message is a single [bytes] token. The details of what this
token contains (and when it can be null/empty, if ever) depends on the actual
authenticator used.