[GH-ISSUE #8827] [Feature] Import OpenTelemetry OTLP traces as read-only distributed-trace Evidence #7765

New issue

Open

opened 2026-11-03 17:35:03 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-11-03 17:35:03 +00:00

Owner

Originally created by @jayykernel on GitHub (Nov 3, 2026).
Original GitHub issue: https://github.com/morluto/rea/issues/8827

Problem

Reverse-engineering a distributed application often requires understanding how a request crosses service boundaries. REA can inspect individual artifacts and selected network captures, but retained OpenTelemetry traces are a useful offline source of parent/child span relationships, service boundaries, timing, status and instrumentation metadata. There is no focused general-purpose OTLP trace import request; the existing LLM inference-trace proposal (#3680) addresses a narrower domain-specific interpretation.

Proposed outcome

Add a read-only importer for caller-supplied OTLP trace exports (JSON and supported protobuf encodings).

Parse resource, scope and span records, including trace/span/parent IDs, timestamps, status, events, links and instrumentation scope.
Preserve service/resource attributes and span attributes with configurable allowlist/redaction; never assume arbitrary attributes are safe to expose.
Reconstruct parent-child relationships while explicitly representing missing parents, duplicate IDs, incomplete batches and out-of-order spans.
Derive duration and critical-path candidates only when timestamps and relationships support them; preserve clock-skew and missing-time uncertainty.
Correlate spans to caller-provided HTTP/network Evidence only when stable identifiers or explicit mapping evidence exists.
Preserve original artifact digest, encoding, source provenance and parser coverage.

Safety and limits

Offline only: do not connect to collectors, query observability backends or fetch linked resources. Bound input size, span count, attribute count and nesting. Redact common credentials, tokens, authorization headers and high-risk attributes by default. Never treat a trace as proof that an unrecorded operation did not occur.

Acceptance criteria

Fixtures cover JSON/protobuf exports, parent-child trees, links, events, missing parents, duplicate IDs, malformed records and partial exports.
Time calculations preserve raw timestamps and identify invalid or suspect intervals.
Redaction tests prove sensitive attribute values are not retained in default output.
CLI/MCP results retain input identity, provenance and explicit partial-coverage diagnostics.
Deterministic tests and docs cover supported OTLP versions/encodings and unsupported signal types.

Scope

General recorded-trace ingestion and structural analysis. Metrics and logs ingestion, live collector integration, and domain-specific LLM inference interpretation are out of scope.

Originally created by @jayykernel on GitHub (Nov 3, 2026). Original GitHub issue: https://github.com/morluto/rea/issues/8827 ## Problem Reverse-engineering a distributed application often requires understanding how a request crosses service boundaries. REA can inspect individual artifacts and selected network captures, but retained OpenTelemetry traces are a useful offline source of parent/child span relationships, service boundaries, timing, status and instrumentation metadata. There is no focused general-purpose OTLP trace import request; the existing LLM inference-trace proposal (#3680) addresses a narrower domain-specific interpretation. ## Proposed outcome Add a read-only importer for caller-supplied OTLP trace exports (JSON and supported protobuf encodings). - Parse resource, scope and span records, including trace/span/parent IDs, timestamps, status, events, links and instrumentation scope. - Preserve service/resource attributes and span attributes with configurable allowlist/redaction; never assume arbitrary attributes are safe to expose. - Reconstruct parent-child relationships while explicitly representing missing parents, duplicate IDs, incomplete batches and out-of-order spans. - Derive duration and critical-path candidates only when timestamps and relationships support them; preserve clock-skew and missing-time uncertainty. - Correlate spans to caller-provided HTTP/network Evidence only when stable identifiers or explicit mapping evidence exists. - Preserve original artifact digest, encoding, source provenance and parser coverage. ## Safety and limits Offline only: do not connect to collectors, query observability backends or fetch linked resources. Bound input size, span count, attribute count and nesting. Redact common credentials, tokens, authorization headers and high-risk attributes by default. Never treat a trace as proof that an unrecorded operation did not occur. ## Acceptance criteria - [ ] Fixtures cover JSON/protobuf exports, parent-child trees, links, events, missing parents, duplicate IDs, malformed records and partial exports. - [ ] Time calculations preserve raw timestamps and identify invalid or suspect intervals. - [ ] Redaction tests prove sensitive attribute values are not retained in default output. - [ ] CLI/MCP results retain input identity, provenance and explicit partial-coverage diagnostics. - [ ] Deterministic tests and docs cover supported OTLP versions/encodings and unsupported signal types. ## Scope General recorded-trace ingestion and structural analysis. Metrics and logs ingestion, live collector integration, and domain-specific LLM inference interpretation are out of scope.

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

morluto/rea#7765

No description provided.

Rows
Columns