This repository was archived by the owner on Nov 17, 2025. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 162
Expand file tree
/
Copy path0239-profiles-data-model.md
More file actions
1604 lines (1216 loc) · 68.5 KB
/
0239-profiles-data-model.md
File metadata and controls
1604 lines (1216 loc) · 68.5 KB
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
# Profiles Data Format
Introduces Data Model for Profiles signal to OpenTelemetry.
<!-- toc -->
* [Motivation](#motivation)
* [Design Notes](#design-notes)
* [Design Goals](#design-goals)
* [Data Model](#data-model)
* [Relationships Diagram](#relationships-diagram)
* [Relationships With Other Signals](#relationships-with-other-signals)
* [From Profiles to Other Signals](#from-profiles-to-other-signals)
* [From Other Signals to Profiles](#from-other-signals-to-profiles)
* [Compatibility With Original pprof](#compatibility-with-original-pprof)
* [Proto Definition](#proto-definition)
* [Message Descriptions](#message-descriptions)
* [Message `ProfilesData`](#message-profilesdata)
* [Message `ResourceProfiles`](#message-resourceprofiles)
* [Message `ScopeProfiles`](#message-scopeprofiles)
* [Message `ProfileContainer`](#message-profilecontainer)
* [Message `Profile`](#message-profile)
* [Message `ValueType`](#message-valuetype)
* [Message `Sample`](#message-sample)
* [Message `AttributeUnit`](#message-attributeunit)
* [Message `Link`](#message-link)
* [Message `Location`](#message-location)
* [Message `Line`](#message-line)
* [Message `Mapping`](#message-mapping)
* [Message `Function`](#message-function)
* [Example Payloads](#example-payloads)
* [Simple Example](#simple-example)
* [Notable Differences Compared to Other Signals](#notable-differences-compared-to-other-signals)
* [Relationships Between Messages](#relationships-between-messages)
* [Relationship Between Samples and Locations](#relationship-between-samples-and-locations)
* [Trade-Offs and Mitigations](#trade-offs-and-mitigations)
* [Prior Art and Alternatives](#prior-art-and-alternatives)
* [Other Popular Formats](#other-popular-formats)
* [Folded Stacks](#folded-stacks)
* [Chromium's Trace Event Format](#chromiums-trace-event-format)
* [Linux perf.data](#linux-perfdata)
* [Java Flight Recorder (JFR)](#java-flight-recorder-jfr)
* [Alternative Representations](#alternative-representations)
* [Benchmarking](#benchmarking)
* ["average" Profile](#average-profile)
* ["average" Profile With Timestamps Added to Each Sample](#average-profile-with-timestamps-added-to-each-sample)
* ["ruby" Profile With Very Deep Stacktraces](#ruby-profile-with-very-deep-stacktraces)
* ["large" Profile](#large-profile)
* [Conclusions](#conclusions)
* [Semantic Conventions](#semantic-conventions)
* [Attributes](#attributes)
* [Profile Types](#profile-types)
* [Profile Units](#profile-units)
* [Decision Log](#decision-log)
* [Open Questions](#open-questions)
* [Units in Attributes](#units-in-attributes)
* [Timestamps](#timestamps)
* [Repetition of Attribute Keys](#repetition-of-attribute-keys)
* [Locations Optimization](#locations-optimization)
* [Future Possibilities](#future-possibilities)
<!-- toc -->
## Motivation
This is a proposal of a data model and semantic conventions that allow to represent profiles coming from a variety of different applications or systems. Existing profiling formats can be unambiguously mapped to this data model. Reverse mapping from this data model is also possible to the extent that the target profiling format has equivalent capabilities.
The purpose of the data model is to have a common understanding of what a profile is, what data needs to be recorded, transferred, stored and interpreted by a profiling system.
## Design Notes
### Design Goals
These goals are based on the vision set out in [Profiling Vision OTEP](./0212-profiling-vision.md):
* Make profiling compatible with other signals.
* Standardize profiling data model for industry-wide sharing and reuse.
* Profilers must be implementable with low overhead and conforming to OpenTelemetry-wide runtime overhead/intrusiveness and wire data size requirements.
The last point is particularly important in the context of profiling. Profilers generate large amounts of data, and users of profiling technology are very sensitive to the overhead that profiling introduces. In the past high overhead has been a blocker for wider adoption of continuous profiling and was one of the reasons why profiling was not used in production environments. Therefore, it is important to make sure that the overhead of handling the profiling data on the client side as well as in intermediaries (e.g collector) is minimal.
## Data Model
This section describes various protobuf messages that are used to represent profiles data.
### Relationships Diagram
The following diagram shows the relationships between the messages. Relationships between messages are represented by either embedding one message in another (red arrows), or by referencing a message by index in a lookup table (blue arrows). More on that in [Relationships Between Messages](#relationships-between-messages) section below.
In addition to that, relationship between `samples` and `locations` is further optimized for better performance. More on that in [Relationship Between Samples and Locations](#relationship-between-samples-and-locations) section below.

### Relationships With Other Signals
There are two types of relationships between profiles and other signals:
* from other signals to profiles (e.g from log records, exemplars or trace spans)
* from profiles to other signals
#### From Profiles to Other Signals
[Link](#message-link) is a message that is used to represent connections between profile [Samples](#message-sample) and trace spans. It uses `trace_id` and `span_id` as identifiers.
For other signals, such as logs or metrics, because other signals use the same way of linking between such signals and traces (`trace_id` and `span_id`), it is possible to correlate profiles with other signals using this same information.
#### From Other Signals to Profiles
Other signals can use `profile_id` to reference a profile. For example, a log record can reference a profile that was collected at the time when the log record was generated by using `profile_id` as one of the attributes. This allows to correlate logs with profiles.
Additionally, `trace_id`, `span_id` can be used to reference groups of [Samples](#message-sample) (but not individual [Samples](#message-sample)) in a Profile, since [Samples](#message-sample) are linked to traces with these same identifiers using [Links](#message-link).
The exact details of such linking are out of scope for this OTEP. It is expected that the exact details will be defined in Profiles part of [opentelemetry-specification](https://github.com/open-telemetry/opentelemetry-specification).
### Compatibility With Original pprof
The proposed data model is backward compatible with original pprof in a sense that a pprof file generated by existing software can be parsed using the new proto. All fields in the original pprof are preserved, so that original pprof files can still be parsed using the new proto, and no data is lost.
It is not forward compatible, meaning that a pprof file generated by the new proto cannot be parsed by existing software. This is mainly due to the sharing of the call stacks between samples + new format for labels (more on these differences below).
### Proto Definition
Proto definition is based on [pprof format](https://github.com/google/pprof/blob/main/proto/profile.proto).
In the landscape of performance profiling tools, pprof's data format stands as a clear industry standard. Its evolution and enduring relevance are a reflection of its effectiveness in addressing diverse and complex performance profiling needs. Major technology firms and open-source projects alike routinely employ pprof, underscoring its universal applicability and reliability.
According to the [data from Profilerpedia](https://docs.google.com/spreadsheets/d/1UM-WFQhNf4GcyXmluSUGnMbOenvN-TqP2HQC9-Y50Lc/edit?usp=sharing), pprof is one of the most widely used formats. Compared to other formats it has the highest number of profilers, UIs, formats it can be converted to and from.
The original pprof data model underwent enhancements to more effectively manage profiling data within the scope of OpenTelemetry, and certain upgrades were implemented to overcome a few of the original format's constraints.
Here's a [link to a diff between original pprof and modified pprof](https://github.com/open-telemetry/opentelemetry-proto-profile/compare/2cf711b3cfcc1edd4e46f9b82d19d016d6d0aa2a...petethepig:opentelemetry-proto:pprof-experiments#diff-9cb689ea05ecfd2edffc39869eca3282a3f2f45a8e1aa21624b452fa5362d1d2) and here's a list of main differences between pprof and OTLP profiles:
* Sharing of the call stacks between samples.
* Sharing of labels (now called attributes) between samples.
* Reuse of OpenTelemetry conventions and message types.
* Semantic conventions for linking to other signals via `trace_id`s and `span_id`s.
* First-class timestamp support.
* Expanded metadata attach points (Sample / Location / Mapping).
Below you will find the proto for the new Profiles signal. It is split into two parts: the first part is the OpenTelemetry specific part, and the second part is the modified pprof proto. Intention here is to make it easier to compare modified pprof proto to the original pprof proto.
OpenTelemetry specific part:
<!-- proto1 -->
```proto
syntax = "proto3";
package opentelemetry.proto.profiles.v1;
import "opentelemetry/proto/common/v1/common.proto";
import "opentelemetry/proto/resource/v1/resource.proto";
import "opentelemetry/proto/profiles/v1/alternatives/pprofextended/pprofextended.proto";
option csharp_namespace = "OpenTelemetry.Proto.Profiles.V1";
option java_multiple_files = true;
option java_package = "io.opentelemetry.proto.profiles.v1";
option java_outer_classname = "ProfilesProto";
option go_package = "go.opentelemetry.io/proto/otlp/profiles/v1";
// Relationships Diagram
//
// ┌──────────────────┐ LEGEND
// │ ProfilesData │
// └──────────────────┘ ─────▶ embedded
// │
// │ 1-n ─────▷ referenced by index
// ▼
// ┌──────────────────┐
// │ ResourceProfiles │
// └──────────────────┘
// │
// │ 1-n
// ▼
// ┌──────────────────┐
// │ ScopeProfiles │
// └──────────────────┘
// │
// │ 1-n
// ▼
// ┌──────────────────┐
// │ ProfileContainer │
// └──────────────────┘
// │
// │ 1-1
// ▼
// ┌──────────────────┐
// │ Profile │
// └──────────────────┘
// │ 1-n
// │ 1-n ┌───────────────────────────────────────┐
// ▼ │ ▽
// ┌──────────────────┐ 1-n ┌──────────────┐ ┌──────────┐
// │ Sample │ ──────▷ │ KeyValue │ │ Link │
// └──────────────────┘ └──────────────┘ └──────────┘
// │ 1-n △ △
// │ 1-n ┌─────────────────┘ │ 1-n
// ▽ │ │
// ┌──────────────────┐ n-1 ┌──────────────┐
// │ Location │ ──────▷ │ Mapping │
// └──────────────────┘ └──────────────┘
// │
// │ 1-n
// ▼
// ┌──────────────────┐
// │ Line │
// └──────────────────┘
// │
// │ 1-1
// ▽
// ┌──────────────────┐
// │ Function │
// └──────────────────┘
//
// ProfilesData represents the profiles data that can be stored in persistent storage,
// OR can be embedded by other protocols that transfer OTLP profiles data but do not
// implement the OTLP protocol.
//
// The main difference between this message and collector protocol is that
// in this message there will not be any "control" or "metadata" specific to
// OTLP protocol.
//
// When new fields are added into this message, the OTLP request MUST be updated
// as well.
message ProfilesData {
// An array of ResourceProfiles.
// For data coming from a single resource this array will typically contain
// one element. Intermediary nodes that receive data from multiple origins
// typically batch the data before forwarding further and in that case this
// array will contain multiple elements.
repeated ResourceProfiles resource_profiles = 1;
}
// A collection of ScopeProfiles from a Resource.
message ResourceProfiles {
reserved 1000;
// The resource for the profiles in this message.
// If this field is not set then no resource info is known.
opentelemetry.proto.resource.v1.Resource resource = 1;
// A list of ScopeProfiles that originate from a resource.
repeated ScopeProfiles scope_profiles = 2;
// This schema_url applies to the data in the "resource" field. It does not apply
// to the data in the "scope_profiles" field which have their own schema_url field.
string schema_url = 3;
}
// A collection of Profiles produced by an InstrumentationScope.
message ScopeProfiles {
// The instrumentation scope information for the profiles in this message.
// Semantically when InstrumentationScope isn't set, it is equivalent with
// an empty instrumentation scope name (unknown).
opentelemetry.proto.common.v1.InstrumentationScope scope = 1;
// A list of ProfileContainers that originate from an instrumentation scope.
repeated ProfileContainer profiles = 2;
// This schema_url applies to all profiles and profile events in the "profiles" field.
string schema_url = 3;
}
// A ProfileContainer represents a single profile. It wraps pprof profile with OpenTelemetry specific metadata.
message ProfileContainer {
// A globally unique identifier for a profile. The ID is a 16-byte array. An ID with
// all zeroes is considered invalid.
//
// This field is required.
bytes profile_id = 1;
// start_time_unix_nano is the start time of the profile.
// Value is UNIX Epoch time in nanoseconds since 00:00:00 UTC on 1 January 1970.
//
// This field is semantically required and it is expected that end_time >= start_time.
fixed64 start_time_unix_nano = 2;
// end_time_unix_nano is the end time of the profile.
// Value is UNIX Epoch time in nanoseconds since 00:00:00 UTC on 1 January 1970.
//
// This field is semantically required and it is expected that end_time >= start_time.
fixed64 end_time_unix_nano = 3;
// attributes is a collection of key/value pairs. Note, global attributes
// like server name can be set using the resource API. Examples of attributes:
//
// "/http/user_agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_14_2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/71.0.3578.98 Safari/537.36"
// "/http/server_latency": 300
// "abc.com/myattribute": true
// "abc.com/score": 10.239
//
// The OpenTelemetry API specification further restricts the allowed value types:
// https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/common/README.md#attribute
// Attribute keys MUST be unique (it is not allowed to have more than one
// attribute with the same key).
repeated opentelemetry.proto.common.v1.KeyValue attributes = 4;
// dropped_attributes_count is the number of attributes that were discarded. Attributes
// can be discarded because their keys are too long or because there are too many
// attributes. If this value is 0, then no attributes were dropped.
uint32 dropped_attributes_count = 5;
// Specifies format of the original payload. Common values are defined in semantic conventions. [required if original_payload is present]
string original_payload_format = 6;
// Original payload can be stored in this field. This can be useful for users who want to get the original payload.
// Formats such as JFR are highly extensible and can contain more information than what is defined in this spec.
// Inclusion of original payload should be configurable by the user. Default behavior should be to not include the original payload.
// If the original payload is in pprof format, it SHOULD not be included in this field.
// The field is optional, however if it is present `profile` MUST be present and contain the same profiling information.
bytes original_payload = 7;
// This is a reference to a pprof profile. Required, even when original_payload is present.
opentelemetry.proto.profiles.v1.alternatives.pprofextended.Profile profile = 8;
}
```
<!-- proto1 -->
Modified pprof:
<!-- proto2 -->
```proto
// Profile is a common stacktrace profile format.
//
// Measurements represented with this format should follow the
// following conventions:
//
// - Consumers should treat unset optional fields as if they had been
// set with their default value.
//
// - When possible, measurements should be stored in "unsampled" form
// that is most useful to humans. There should be enough
// information present to determine the original sampled values.
//
// - On-disk, the serialized proto must be gzip-compressed.
//
// - The profile is represented as a set of samples, where each sample
// references a sequence of locations, and where each location belongs
// to a mapping.
// - There is a N->1 relationship from sample.location_id entries to
// locations. For every sample.location_id entry there must be a
// unique Location with that index.
// - There is an optional N->1 relationship from locations to
// mappings. For every nonzero Location.mapping_id there must be a
// unique Mapping with that index.
syntax = "proto3";
package opentelemetry.proto.profiles.v1.alternatives.pprofextended;
import "opentelemetry/proto/common/v1/common.proto";
option csharp_namespace = "OpenTelemetry.Proto.Profiles.V1.Alternatives.PprofExtended";
option go_package = "go.opentelemetry.io/proto/otlp/profiles/v1/alternatives/pprofextended";
// Represents a complete profile, including sample types, samples,
// mappings to binaries, locations, functions, string table, and additional metadata.
message Profile {
// A description of the samples associated with each Sample.value.
// For a cpu profile this might be:
// [["cpu","nanoseconds"]] or [["wall","seconds"]] or [["syscall","count"]]
// For a heap profile, this might be:
// [["allocations","count"], ["space","bytes"]],
// If one of the values represents the number of events represented
// by the sample, by convention it should be at index 0 and use
// sample_type.unit == "count".
repeated ValueType sample_type = 1;
// The set of samples recorded in this profile.
repeated Sample sample = 2;
// Mapping from address ranges to the image/binary/library mapped
// into that address range. mapping[0] will be the main binary.
repeated Mapping mapping = 3;
// Locations referenced by samples via location_indices.
repeated Location location = 4;
// Array of locations referenced by samples.
repeated int64 location_indices = 15;
// Functions referenced by locations.
repeated Function function = 5;
// Lookup table for attributes.
repeated opentelemetry.proto.common.v1.KeyValue attribute_table = 16;
// Represents a mapping between Attribute Keys and Units.
repeated AttributeUnit attribute_units = 17;
// Lookup table for links.
repeated Link link_table = 18;
// A common table for strings referenced by various messages.
// string_table[0] must always be "".
repeated string string_table = 6;
// frames with Function.function_name fully matching the following
// regexp will be dropped from the samples, along with their successors.
int64 drop_frames = 7; // Index into string table.
// frames with Function.function_name fully matching the following
// regexp will be kept, even if it matches drop_frames.
int64 keep_frames = 8; // Index into string table.
// The following fields are informational, do not affect
// interpretation of results.
// Time of collection (UTC) represented as nanoseconds past the epoch.
int64 time_nanos = 9;
// Duration of the profile, if a duration makes sense.
int64 duration_nanos = 10;
// The kind of events between sampled occurrences.
// e.g [ "cpu","cycles" ] or [ "heap","bytes" ]
ValueType period_type = 11;
// The number of events between sampled occurrences.
int64 period = 12;
// Free-form text associated with the profile. The text is displayed as is
// to the user by the tools that read profiles (e.g. by pprof). This field
// should not be used to store any machine-readable information, it is only
// for human-friendly content. The profile must stay functional if this field
// is cleaned.
repeated int64 comment = 13; // Indices into string table.
// Index into the string table of the type of the preferred sample
// value. If unset, clients should default to the last sample value.
int64 default_sample_type = 14;
}
// Represents a mapping between Attribute Keys and Units.
message AttributeUnit {
// Index into string table.
int64 attribute_key = 1;
// Index into string table.
int64 unit = 2;
}
// A pointer from a profile Sample to a trace Span.
// Connects a profile sample to a trace span, identified by unique trace and span IDs.
message Link {
// A unique identifier of a trace that this linked span is part of. The ID is a
// 16-byte array.
bytes trace_id = 1;
// A unique identifier for the linked span. The ID is an 8-byte array.
bytes span_id = 2;
}
// Specifies the method of aggregating metric values, either DELTA (change since last report)
// or CUMULATIVE (total since a fixed start time).
enum AggregationTemporality {
/* UNSPECIFIED is the default AggregationTemporality, it MUST not be used. */
AGGREGATION_TEMPORALITY_UNSPECIFIED = 0;
/** DELTA is an AggregationTemporality for a profiler which reports
changes since last report time. Successive metrics contain aggregation of
values from continuous and non-overlapping intervals.
The values for a DELTA metric are based only on the time interval
associated with one measurement cycle. There is no dependency on
previous measurements like is the case for CUMULATIVE metrics.
For example, consider a system measuring the number of requests that
it receives and reports the sum of these requests every second as a
DELTA metric:
1. The system starts receiving at time=t_0.
2. A request is received, the system measures 1 request.
3. A request is received, the system measures 1 request.
4. A request is received, the system measures 1 request.
5. The 1 second collection cycle ends. A metric is exported for the
number of requests received over the interval of time t_0 to
t_0+1 with a value of 3.
6. A request is received, the system measures 1 request.
7. A request is received, the system measures 1 request.
8. The 1 second collection cycle ends. A metric is exported for the
number of requests received over the interval of time t_0+1 to
t_0+2 with a value of 2. */
AGGREGATION_TEMPORALITY_DELTA = 1;
/** CUMULATIVE is an AggregationTemporality for a profiler which
reports changes since a fixed start time. This means that current values
of a CUMULATIVE metric depend on all previous measurements since the
start time. Because of this, the sender is required to retain this state
in some form. If this state is lost or invalidated, the CUMULATIVE metric
values MUST be reset and a new fixed start time following the last
reported measurement time sent MUST be used.
For example, consider a system measuring the number of requests that
it receives and reports the sum of these requests every second as a
CUMULATIVE metric:
1. The system starts receiving at time=t_0.
2. A request is received, the system measures 1 request.
3. A request is received, the system measures 1 request.
4. A request is received, the system measures 1 request.
5. The 1 second collection cycle ends. A metric is exported for the
number of requests received over the interval of time t_0 to
t_0+1 with a value of 3.
6. A request is received, the system measures 1 request.
7. A request is received, the system measures 1 request.
8. The 1 second collection cycle ends. A metric is exported for the
number of requests received over the interval of time t_0 to
t_0+2 with a value of 5.
9. The system experiences a fault and loses state.
10. The system recovers and resumes receiving at time=t_1.
11. A request is received, the system measures 1 request.
12. The 1 second collection cycle ends. A metric is exported for the
number of requests received over the interval of time t_1 to
t_0+1 with a value of 1.
Note: Even though, when reporting changes since last report time, using
CUMULATIVE is valid, it is not recommended. */
AGGREGATION_TEMPORALITY_CUMULATIVE = 2;
}
// ValueType describes the type and units of a value, with an optional aggregation temporality.
message ValueType {
int64 type = 1; // Index into string table.
int64 unit = 2; // Index into string table.
AggregationTemporality aggregation_temporality = 3;
}
// Each Sample records values encountered in some program
// context. The program context is typically a stack trace, perhaps
// augmented with auxiliary information like the thread-id, some
// indicator of a higher level request being handled etc.
message Sample {
// The indices recorded here correspond to locations in Profile.location.
// The leaf is at location_index[0]. [deprecated, superseded by locations_start_index / locations_length]
repeated uint64 location_index = 1;
// locations_start_index along with locations_length refers to to a slice of locations in Profile.location.
// Supersedes location_index.
uint64 locations_start_index = 7;
// locations_length along with locations_start_index refers to a slice of locations in Profile.location.
// Supersedes location_index.
uint64 locations_length = 8;
// A 128bit id that uniquely identifies this stacktrace, globally. Index into string table. [optional]
uint32 stacktrace_id_index = 9;
// The type and unit of each value is defined by the corresponding
// entry in Profile.sample_type. All samples must have the same
// number of values, the same as the length of Profile.sample_type.
// When aggregating multiple samples into a single sample, the
// result has a list of values that is the element-wise sum of the
// lists of the originals.
repeated int64 value = 2;
// label includes additional context for this sample. It can include
// things like a thread id, allocation size, etc.
//
// NOTE: While possible, having multiple values for the same label key is
// strongly discouraged and should never be used. Most tools (e.g. pprof) do
// not have good (or any) support for multi-value labels. And an even more
// discouraged case is having a string label and a numeric label of the same
// name on a sample. Again, possible to express, but should not be used.
// [deprecated, superseded by attributes]
repeated Label label = 3;
// References to attributes in Profile.attribute_table. [optional]
repeated uint64 attributes = 10;
// Reference to link in Profile.link_table. [optional]
uint64 link = 12;
// Timestamps associated with Sample represented in ms. These timestamps are expected
// to fall within the Profile's time range. [optional]
repeated uint64 timestamps = 13;
}
// Provides additional context for a sample,
// such as thread ID or allocation size, with optional units. [deprecated]
message Label {
int64 key = 1; // Index into string table
// At most one of the following must be present
int64 str = 2; // Index into string table
int64 num = 3;
// Should only be present when num is present.
// Specifies the units of num.
// Use arbitrary string (for example, "requests") as a custom count unit.
// If no unit is specified, consumer may apply heuristic to deduce the unit.
// Consumers may also interpret units like "bytes" and "kilobytes" as memory
// units and units like "seconds" and "nanoseconds" as time units,
// and apply appropriate unit conversions to these.
int64 num_unit = 4; // Index into string table
}
// Indicates the semantics of the build_id field.
enum BuildIdKind {
// Linker-generated build ID, stored in the ELF binary notes.
BUILD_ID_LINKER = 0;
// Build ID based on the content hash of the binary. Currently no particular
// hashing approach is standardized, so a given producer needs to define it
// themselves and thus unlike BUILD_ID_LINKER this kind of hash is producer-specific.
// We may choose to provide a standardized stable hash recommendation later.
BUILD_ID_BINARY_HASH = 1;
}
// Describes the mapping of a binary in memory, including its address range,
// file offset, and metadata like build ID
message Mapping {
// Unique nonzero id for the mapping. [deprecated]
uint64 id = 1;
// Address at which the binary (or DLL) is loaded into memory.
uint64 memory_start = 2;
// The limit of the address range occupied by this mapping.
uint64 memory_limit = 3;
// Offset in the binary that corresponds to the first mapped address.
uint64 file_offset = 4;
// The object this entry is loaded from. This can be a filename on
// disk for the main binary and shared libraries, or virtual
// abstractions like "[vdso]".
int64 filename = 5; // Index into string table
// A string that uniquely identifies a particular program version
// with high probability. E.g., for binaries generated by GNU tools,
// it could be the contents of the .note.gnu.build-id field.
int64 build_id = 6; // Index into string table
// Specifies the kind of build id. See BuildIdKind enum for more details [optional]
BuildIdKind build_id_kind = 11;
// References to attributes in Profile.attribute_table. [optional]
repeated uint64 attributes = 12;
// The following fields indicate the resolution of symbolic info.
bool has_functions = 7;
bool has_filenames = 8;
bool has_line_numbers = 9;
bool has_inline_frames = 10;
}
// Describes function and line table debug information.
message Location {
// Unique nonzero id for the location. A profile could use
// instruction addresses or any integer sequence as ids. [deprecated]
uint64 id = 1;
// The index of the corresponding profile.Mapping for this location.
// It can be unset if the mapping is unknown or not applicable for
// this profile type.
uint64 mapping_index = 2;
// The instruction address for this location, if available. It
// should be within [Mapping.memory_start...Mapping.memory_limit]
// for the corresponding mapping. A non-leaf address may be in the
// middle of a call instruction. It is up to display tools to find
// the beginning of the instruction if necessary.
uint64 address = 3;
// Multiple line indicates this location has inlined functions,
// where the last entry represents the caller into which the
// preceding entries were inlined.
//
// E.g., if memcpy() is inlined into printf:
// line[0].function_name == "memcpy"
// line[1].function_name == "printf"
repeated Line line = 4;
// Provides an indication that multiple symbols map to this location's
// address, for example due to identical code folding by the linker. In that
// case the line information above represents one of the multiple
// symbols. This field must be recomputed when the symbolization state of the
// profile changes.
bool is_folded = 5;
// Type of frame (e.g. kernel, native, python, hotspot, php). Index into string table.
uint32 type_index = 6;
// References to attributes in Profile.attribute_table. [optional]
repeated uint64 attributes = 7;
}
// Details a specific line in a source code, linked to a function.
message Line {
// The index of the corresponding profile.Function for this line.
uint64 function_index = 1;
// Line number in source code.
int64 line = 2;
// Column number in source code.
int64 column = 3;
}
// Describes a function, including its human-readable name, system name,
// source file, and starting line number in the source.
message Function {
// Unique nonzero id for the function. [deprecated]
uint64 id = 1;
// Name of the function, in human-readable form if available.
int64 name = 2; // Index into string table
// Name of the function, as identified by the system.
// For instance, it can be a C++ mangled name.
int64 system_name = 3; // Index into string table
// Source file containing the function.
int64 filename = 4; // Index into string table
// Line number in source file.
int64 start_line = 5;
}
```
<!-- proto2 -->
### Message Descriptions
These are detailed descriptions of protobuf messages that are used to represent profiling data.
<!-- messages -->
#### Message `ProfilesData`
ProfilesData represents the profiles data that can be stored in persistent storage,
OR can be embedded by other protocols that transfer OTLP profiles data but do not
implement the OTLP protocol.
The main difference between this message and collector protocol is that
in this message there will not be any "control" or "metadata" specific to
OTLP protocol.
When new fields are added into this message, the OTLP request MUST be updated
as well.
#### Message `ResourceProfiles`
A collection of ScopeProfiles from a Resource.
<details>
<summary>Field Descriptions</summary>
##### Field `resource`
The resource for the profiles in this message.
If this field is not set then no resource info is known.
##### Field `scope_profiles`
A list of ScopeProfiles that originate from a resource.
##### Field `schema_url`
This schema_url applies to the data in the "resource" field. It does not apply
to the data in the "scope_profiles" field which have their own schema_url field.
</details>
#### Message `ScopeProfiles`
A collection of Profiles produced by an InstrumentationScope.
<details>
<summary>Field Descriptions</summary>
##### Field `scope`
The instrumentation scope information for the profiles in this message.
Semantically when InstrumentationScope isn't set, it is equivalent with
an empty instrumentation scope name (unknown).
##### Field `profiles`
A list of ProfileContainers that originate from an instrumentation scope.
##### Field `schema_url`
This schema_url applies to all profiles and profile events in the "profiles" field.
</details>
#### Message `ProfileContainer`
A ProfileContainer represents a single profile. It wraps pprof profile with OpenTelemetry specific metadata.
<details>
<summary>Field Descriptions</summary>
##### Field `profile_id`
A globally unique identifier for a profile. The ID is a 16-byte array. An ID with
all zeroes is considered invalid.
This field is required.
##### Field `start_time_unix_nano`
start_time_unix_nano is the start time of the profile.
Value is UNIX Epoch time in nanoseconds since 00:00:00 UTC on 1 January 1970.
This field is semantically required and it is expected that end_time >= start_time.
##### Field `end_time_unix_nano`
end_time_unix_nano is the end time of the profile.
Value is UNIX Epoch time in nanoseconds since 00:00:00 UTC on 1 January 1970.
This field is semantically required and it is expected that end_time >= start_time.
##### Field `attributes`
attributes is a collection of key/value pairs. Note, global attributes
like server name can be set using the resource API.
##### Field `dropped_attributes_count`
dropped_attributes_count is the number of attributes that were discarded. Attributes
can be discarded because their keys are too long or because there are too many
attributes. If this value is 0, then no attributes were dropped.
##### Field `original_payload_format`
Specifies format of the original payload. Common values are defined in semantic conventions. [required if original_payload is present]
##### Field `original_payload`
Original payload can be stored in this field. This can be useful for users who want to get the original payload.
Formats such as JFR are highly extensible and can contain more information than what is defined in this spec.
Inclusion of original payload should be configurable by the user. Default behavior should be to not include the original payload.
If the original payload is in pprof format, it SHOULD not be included in this field.
The field is optional, however if it is present `profile` MUST be present and contain the same profiling information.
##### Field `profile`
This is a reference to a pprof profile. Required, even when original_payload is present.
</details>
#### Message `Profile`
Profile is a common stacktrace profile format.
Measurements represented with this format should follow the
following conventions:
- Consumers should treat unset optional fields as if they had been
set with their default value.
- When possible, measurements should be stored in "unsampled" form
that is most useful to humans. There should be enough
information present to determine the original sampled values.
- On-disk, the serialized proto must be gzip-compressed.
- The profile is represented as a set of samples, where each sample
references a sequence of locations, and where each location belongs
to a mapping.
- There is a N->1 relationship from sample.location_id entries to
locations. For every sample.location_id entry there must be a
unique Location with that index.
- There is an optional N->1 relationship from locations to
mappings. For every nonzero Location.mapping_id there must be a
unique Mapping with that index.
Represents a complete profile, including sample types, samples,
mappings to binaries, locations, functions, string table, and additional metadata.
<details>
<summary>Field Descriptions</summary>
##### Field `sample_type`
A description of the samples associated with each Sample.value.
For a cpu profile this might be:
[["cpu","nanoseconds"]] or [["wall","seconds"]] or [["syscall","count"]]
For a heap profile, this might be:
[["allocations","count"], ["space","bytes"]],
If one of the values represents the number of events represented
by the sample, by convention it should be at index 0 and use
sample_type.unit == "count".
##### Field `sample`
The set of samples recorded in this profile.
##### Field `mapping`
Mapping from address ranges to the image/binary/library mapped
into that address range. mapping[0] will be the main binary.
##### Field `location`
Locations referenced by samples via location_indices.
##### Field `location_indices`
Array of locations referenced by samples.
##### Field `function`
Functions referenced by locations.
##### Field `attribute_table`
Lookup table for attributes.
##### Field `attribute_units`
Represents a mapping between Attribute Keys and Units.
##### Field `link_table`
Lookup table for links.
##### Field `string_table`
A common table for strings referenced by various messages.
string_table[0] must always be "".
##### Field `drop_frames`
frames with Function.function_name fully matching the following
regexp will be dropped from the samples, along with their successors.
##### Field `keep_frames`
Index into string table.
frames with Function.function_name fully matching the following
regexp will be kept, even if it matches drop_frames.
##### Field `time_nanos`
Index into string table.
The following fields are informational, do not affect
interpretation of results.
Time of collection (UTC) represented as nanoseconds past the epoch.
##### Field `duration_nanos`
Duration of the profile, if a duration makes sense.
##### Field `period_type`
The kind of events between sampled occurrences.
e.g [ "cpu","cycles" ] or [ "heap","bytes" ]
##### Field `period`
The number of events between sampled occurrences.
##### Field `comment`
Free-form text associated with the profile. The text is displayed as is
to the user by the tools that read profiles (e.g. by pprof). This field
should not be used to store any machine-readable information, it is only
for human-friendly content. The profile must stay functional if this field
is cleaned.
##### Field `default_sample_type`
Indices into string table.
Index into the string table of the type of the preferred sample
value. If unset, clients should default to the last sample value.
</details>
#### Message `ValueType`
ValueType describes the type and units of a value, with an optional aggregation temporality.
<details>
<summary>Field Descriptions</summary>
##### Field `unit`
Index into string table.
##### Field `aggregation_temporality`
Index into string table.
</details>
#### Message `Sample`
Each Sample records values encountered in some program
context. The program context is typically a stack trace, perhaps
augmented with auxiliary information like the thread-id, some
indicator of a higher level request being handled etc.
<details>
<summary>Field Descriptions</summary>
##### Field `location_index`
The indices recorded here correspond to locations in Profile.location.
The leaf is at location_index[0]. [deprecated, superseded by locations_start_index / locations_length]
##### Field `locations_start_index`
locations_start_index along with locations_length refers to to a slice of locations in Profile.location.
Supersedes location_index.
##### Field `locations_length`
locations_length along with locations_start_index refers to a slice of locations in Profile.location.
Supersedes location_index.
##### Field `stacktrace_id_index`
A 128bit id that uniquely identifies this stacktrace, globally. Index into string table. [optional]
##### Field `value`
The type and unit of each value is defined by the corresponding
entry in Profile.sample_type. All samples must have the same
number of values, the same as the length of Profile.sample_type.
When aggregating multiple samples into a single sample, the
result has a list of values that is the element-wise sum of the
lists of the originals.
##### Field `label`
label includes additional context for this sample. It can include
things like a thread id, allocation size, etc.
NOTE: While possible, having multiple values for the same label key is
strongly discouraged and should never be used. Most tools (e.g. pprof) do
not have good (or any) support for multi-value labels. And an even more
discouraged case is having a string label and a numeric label of the same
name on a sample. Again, possible to express, but should not be used.
[deprecated, superseded by attributes]
##### Field `attributes`
References to attributes in Profile.attribute_table. [optional]
##### Field `link`
Reference to link in Profile.link_table. [optional]
##### Field `timestamps`
Timestamps associated with Sample represented in ms. These timestamps are expected
to fall within the Profile's time range. [optional]
</details>
#### Message `AttributeUnit`