Skip to content

OpenTelemetry Orchestrator (OTO) Prompts

Six prompts for the OpenTelemetry Orchestrator — the protocol_engineering persona at Zachman cell ENGINEER/WHEN that owns FCC's per-lane TracerProviders, OTLP export configuration, span enrichment, and the observability contract shipped in v1.4.2.

OTO is how FCC converts persona events into OTel spans + metrics without losing the lane (PAOMBus vs UXBus) as a first-class dimension.

Context for Claude

As OTO you read the per-lane TracerProvider registry, span enrichment rules, and OTLP endpoint config. You produce observability contracts consumed by downstream SRE teams. You coordinate with LAR on lane identity and with SMC on scorecard ingestion.

Prompt 1: Span schema for a new event type

OTO, design the span schema for CRDT_SESSION_MERGED:
- Span name
- Required attributes (at least 6)
- Optional attributes
- Status mapping (OK, ERROR, UNSET)
- Link semantics (parent vs follows-from)

Expected outcome: span schema ready to wire into src/fcc/observability/tracing.py enrichers.

Prompt 2: OTLP endpoint rollout

SRE is rolling a new OTLP collector. Write the rollout checklist:
- TLS configuration
- Header auth
- Retries + timeouts
- Local buffer size
- Validation steps pre-switchover
- Rollback trigger

Name the single observability metric that must not regress during
rollout.

Expected outcome: rollout checklist + guardrail metric.

Prompt 3: Sampling policy

At scale FCC emits 20,000 spans/sec. Propose a sampling policy:
- Head sampling rate per lane
- Tail sampling rules (error, slow, rare paths)
- Persona whose spans are always kept (100%)
- Dashboards still readable at the sampled rate

Model the expected percentile accuracy loss for p50/p95/p99.

Expected outcome: policy + accuracy estimate.

Prompt 4: Metric to span audit

For each of the 7 pre-defined FccMetrics, identify whether a span
event or attribute complements it. Produce:
- Metric name
- Paired span attribute (or "none needed")
- 1-sentence rationale

Expected outcome: 7-row audit.

Prompt 5: Exporter comparison

Compare 3 OTLP exporter strategies for FCC:
- Direct push to collector (default)
- File-based exporter + sidecar tailer
- Logging exporter (dev mode)

Table: latency, durability, debug affinity, production-ready.
Recommend the default for v1.5.x.

Expected outcome: comparison + recommendation.

Prompt 6: OTel → PI scorecard

Coordinate with SMC: which OTel metrics should feed the PI
scorecard pipeline, and through what aggregation window?

Produce:
- 5 OTel metrics that feed SMC's scorecards
- Aggregation window per metric
- Exemplar span attachment strategy (so SMC can dig from scorecard
  to root-cause span)

Expected outcome: 5-metric feed plan.

  • Persona YAML: src/fcc/data/personas/opentelemetry_orchestrator.yaml
  • ADR: docs/decisions/ADR-006_dual_bus_events.md (v1.5.1)
  • Related prompts: persona-lar-prompts.md, operators-otel-collector-runbook.md