fix(otel): align MDC log-correlation keys with JS and Python SDKs

[SilanHe]

Issue

aws/aws-durable-execution-conformance-tests#23 (comment)

Summary

MdcSpanEnricher injected the MDC keys trace_id / span_id / traceSampled, but the JS (enrichLogContext) and Python (OtelContextLogFilter) plugins emit traceId / spanId / otelTraceSampled. Log↔trace correlation was therefore using two different field schemas across the SDKs (and Java's own naming was internally inconsistent — snake_case IDs but camelCase traceSampled).

This PR renames the three MDC keys to traceId / spanId / otelTraceSampled so all six plugins (JS / Python / Java × Execution / Invocation) share one field schema.

Both Java plugins reference these via the MdcSpanEnricher.MDC_* constants, so this reconciles both ExecutionOtelPlugin and InvocationOtelPlugin in a single change.

Field schema (after this PR)

	trace ID	span ID	sampled
JS (both)	traceId	spanId	otelTraceSampled (boolean)
Python (both)	traceId	spanId	otelTraceSampled (boolean)
Java (both)	traceId	spanId	otelTraceSampled (string)

Residual difference (inherent, documented): SLF4J MDC stores only strings, so Java's otelTraceSampled is the string "true"/"false" rather than a boolean. The key names are now unified across all three SDKs.

Compatibility

⚠️ This changes the MDC key names emitted into structured logs. Anyone parsing the old trace_id / span_id / traceSampled keys (log formatters, CloudWatch/Datadog field mappings) must update to the new names.

Tests

+1 test asserting the exact key strings (traceId/spanId/otelTraceSampled); existing MDC + both plugin suites pass unchanged (they reference the constants).

[github-actions]

Claude AI review

The rename is correct and complete at the code level. All production and test references go through the MdcSpanEnricher.MDC_TRACE_ID / MDC_SPAN_ID / MDC_TRACE_SAMPLED constants (MdcSpanEnricher.java:41-51, ExecutionOtelPlugin.java:233, InvocationOtelPlugin.java:275,561, and all tests), so changing the constant values (traceId / spanId / otelTraceSampled) propagates cleanly with no dangling literals. The Javadoc in all three files was updated in the diff, and the new test mdcKeyNames_matchJsAndPythonSchema pins the exact key strings. The MDC key rename is a deliberate, documented breaking change. I could not independently verify the JS/Python key names (those live in other repos) and am trusting the PR description on cross-SDK alignment.

Finding (documentation, medium): otel-plugin/README.md still documents the old MDC key names and is now stale/incorrect as a direct result of this PR:

• Line 12: Injects trace_id, span_id, and traceSampled ...
• Lines 200–202: table rows trace_id / span_id / traceSampled
• Line 204: The trace_id is also injected at invocation start ...
• Lines 254 and 278: enableMdc descriptions reference trace_id/span_id/traceSampled

Fix: update these to traceId / spanId / otelTraceSampled to match the renamed constants. (These are unchanged lines, so no inline comment was attachable; flagging here per instructions.)

Residual test risk: low. The change is a pure constant rename covered by the updated/added tests. No correctness, concurrency, replay, or serialization semantics are affected. The only downstream risk is external log-parsing/field-mapping configs (CloudWatch/Datadog), which the PR body already calls out as breaking.

Reviewed commit 5ab6b814f3c546d20ec825b17517d2c6b2a9fffe. Workflow run

[github-actions]

Codex AI review

• [P2] Update the published MDC documentation — otel-plugin/src/main/java/software/amazon/lambda/durable/otel/MdcSpanEnricher.java:36. The implementation now emits camel-case keys, but otel-plugin/README.md still documents trace_id, span_id, and traceSampled. Users following it will configure absent fields and lose log correlation. Update all README, example, and test-comment references to the new schema.

Reviewed commit 5ab6b814f3c546d20ec825b17517d2c6b2a9fffe. Workflow run