Standardize how to record DroppedAttributeCount in protocols that have no equivalent #1656
Assignees
Labels
area:semantic-conventions
Related to semantic conventions
spec:protocol
Related to the specification/protocol directory
Comments
tigrannajaryan
added
the
spec:protocol
tigrannajaryan
added a commit
to tigrannajaryan/opentelemetry-specification
that referenced
this issue
Apr 29, 2021
Fixes open-telemetry#1656 The documents that describe transformations to Jaeger and Zipkin contained duplicate information about how to record InstrumentationLibrary. They also contained incomplete and incorrect information about how to record dropped attribute count. This change extracts common transformation rules into a separate document (non-otlp.md) and references this document where appropriate. We also define otel.event.dropped_attributes_count as the only way to record dropped attribute count, regardless of the entity for which it is recorded. Previously in Jaeger format we specified that for Events we must use otel.event.dropped_attributes_count and did not specify at all what to use for Spans. I believe using the same uniform key for Spans and Events is correct. I consider this a specification bug and do not believe it should be treated as a breaking change. Also delete the Zipkin mapping for LocalChildSpanCount since there is no such OpenTelemetry field anymore.
tigrannajaryan
added a commit
to tigrannajaryan/opentelemetry-specification
that referenced
this issue
Apr 29, 2021
Fixes open-telemetry#1656 The documents that describe transformations to Jaeger and Zipkin contained duplicate information about how to record InstrumentationLibrary. They also contained incomplete and incorrect information about how to record dropped attribute count. This change extracts common transformation rules into a separate document (non-otlp.md) and references this document where appropriate. We also define otel.event.dropped_attributes_count as the only way to record dropped attribute count, regardless of the entity for which it is recorded. Previously in Jaeger format we specified that for Events we must use otel.event.dropped_attributes_count and did not specify at all what to use for Spans. I believe using the same uniform key for Spans and Events is correct. I consider this a specification bug and do not believe it should be treated as a breaking change. Also delete the Zipkin mapping for LocalChildSpanCount since there is no such OpenTelemetry field anymore.
tigrannajaryan
added a commit
to tigrannajaryan/opentelemetry-specification
that referenced
this issue
Apr 29, 2021
Fixes open-telemetry#1656 The documents that describe transformations to Jaeger and Zipkin contained duplicate information about how to record InstrumentationLibrary. They also contained incomplete and incorrect information about how to record dropped attribute count. This change extracts common transformation rules into a separate document (non-otlp.md) and references this document where appropriate. We also define otel.event.dropped_attributes_count as the only way to record dropped attribute count, regardless of the entity for which it is recorded. Previously in Jaeger format we specified that for Events we must use otel.event.dropped_attributes_count and did not specify at all what to use for Spans. I believe using the same uniform key for Spans and Events is correct. I consider this a specification bug and do not believe it should be treated as a breaking change. Also delete the Zipkin mapping for LocalChildSpanCount since there is no such OpenTelemetry field anymore.
tigrannajaryan
added a commit
to tigrannajaryan/opentelemetry-specification
that referenced
this issue
Apr 29, 2021
Fixes open-telemetry#1656 The documents that describe transformations to Jaeger and Zipkin contained duplicate information about how to record InstrumentationLibrary. They also contained incomplete and incorrect information about how to record dropped attribute count. This change extracts common transformation rules into a separate document (non-otlp.md) and references this document where appropriate. We also define otel.event.dropped_attributes_count as the only way to record dropped attribute count, regardless of the entity for which it is recorded. Previously in Jaeger format we specified that for Events we must use otel.event.dropped_attributes_count and did not specify at all what to use for Spans. I believe using the same uniform key for Spans and Events is correct. I consider this a specification bug and do not believe it should be treated as a breaking change. Also delete the Zipkin mapping for LocalChildSpanCount since there is no such OpenTelemetry field anymore.
tigrannajaryan
added a commit
to tigrannajaryan/opentelemetry-specification
that referenced
this issue
Apr 30, 2021
Fixes open-telemetry#1656 The documents that describe transformations to Jaeger and Zipkin contained duplicate information about how to record InstrumentationLibrary. They also contained incomplete and incorrect information about how to record dropped attribute count. This change extracts common transformation rules into a separate document (non-otlp.md) and references this document where appropriate. We also define otel.event.dropped_attributes_count as the only way to record dropped attribute count, regardless of the entity for which it is recorded. Previously in Jaeger format we specified that for Events we must use otel.event.dropped_attributes_count and did not specify at all what to use for Spans. I believe using the same uniform key for Spans and Events is correct. I consider this a specification bug and do not believe it should be treated as a breaking change. Also delete the Zipkin mapping for LocalChildSpanCount since there is no such OpenTelemetry field anymore.
tigrannajaryan
added a commit
to tigrannajaryan/opentelemetry-specification
that referenced
this issue
Apr 30, 2021
Fixes open-telemetry#1656 The documents that describe transformations to Jaeger and Zipkin contained duplicate information about how to record InstrumentationLibrary. They also contained incomplete and incorrect information about how to record dropped attribute count. This change extracts common transformation rules into a separate document (non-otlp.md) and references this document where appropriate. We also define otel.event.dropped_attributes_count as the only way to record dropped attribute count, regardless of the entity for which it is recorded. Previously in Jaeger format we specified that for Events we must use otel.event.dropped_attributes_count and did not specify at all what to use for Spans. I believe using the same uniform key for Spans and Events is correct. I consider this a specification bug and do not believe it should be treated as a breaking change. Also delete the Zipkin mapping for LocalChildSpanCount since there is no such OpenTelemetry field anymore.
|
@reyang to follow up with @carlosalberto and create milestones for tracking. |
tigrannajaryan
added a commit
that referenced
this issue
May 4, 2021
…1662) Fixes #1656 The documents that describe transformations to Jaeger and Zipkin contained duplicate information about how to record InstrumentationLibrary. They also contained incomplete and incorrect information about how to record dropped attribute count. This change extracts common transformation rules into a separate document (non-otlp.md) and references this document where appropriate. We also define otel.event.dropped_attributes_count as the only way to record dropped attribute count, regardless of the entity for which it is recorded. Previously in Jaeger format we specified that for Events we must use otel.event.dropped_attributes_count and did not specify at all what to use for Spans. I believe using the same uniform key for Spans and Events is correct. I consider this a specification bug and do not believe it should be treated as a breaking change. Also delete the Zipkin mapping for LocalChildSpanCount since there is no such OpenTelemetry field anymore.
|
@tigrannajaryan it's unclear to me what generates these dropped attributes/events in the first place. Is that defined somewhere in the specification? To be clear, I'm asking what would be causing this value to be non-zero
|
I believe they should be populated by SDKs when the limits are hit: https://github.com/open-telemetry/opentelemetry-specification/blob/5a19b53d71e967659517c02a69b801381d29bf1e/specification/trace/sdk.md#span-limits It is not defined explicitly in the spec, worth adding. |
|
Thank you for clarifying! I can add this to the spec. |
This was referenced May 13, 2021
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
For Jaeger we suggest to use
otel.event.dropped_attributes_counttag for Events (Jaeger Logs). We say nothing about how to record the value of DroppedAttributeCount for Spans.For Zipkin this is not defined (the spec says TBD).
I believe we need one standard way to record DroppedAttributeCount in protocols that have no equivalent and do it the same way for Spans, SpanEvents, Logs, Records.
I suggest that we always use
otel.dropped_attributes_countattribute name to represent DroppedAttributeCount field of Spans, SpanEvents, Logs, Records when the equivalent does not exist in the protocol.The text was updated successfully, but these errors were encountered: