OpenTelemetry OTLP High-Cardinality Dynamic Attribute Scrubbing & Circuit Breaker Guardrail

Q: How does adopting this rule prevent AI hallucinations?

This rule explicitly declares what the AI must avoid (Unbounded Runtime Span Attribute Injection Inducing OTLP Collector Memory Exhaustion and Storage Explosion) and prescribes verified positive examples. Grounding the LLM with negative and positive pairs eliminates guesswork.

Q: Which AI coding environments are supported?

This specification is multi-platform compatible: Cursor (.cursor/rules/*.mdc & .cursorrules), Anthropic Claude Code (CLAUDE.md), Windsurf (.windsurfrules), Codex / Cline (AGENTS.md), and GitHub Copilot (.github/copilot-instructions.md).

Q: Can I combine this rule with other technology stacks?

Yes. Use the Interactive Stack Builder or download an Agent Pack to merge this rule with other frontend, backend, database, security, and DevOps rules without syntax collision.

OpenTelemetry OTLP High-Cardinality Dynamic Attribute Scrubbing & Circuit Breaker Guardrail

--- description: OpenTelemetry OTLP High-Cardinality Dynamic Attribute Scrubbing & Circuit Breaker Guardrail - Protects downstream telemetry collector pipelines, ClickHouse/Tempo backends, and in-memory trace buffers from catastrophic OOM crashes by dynamically bounding, hashing, and shedding unindexed dynamic runtime span attributes. Implements runtime span processors with token bucket rate limiters, strict key allowlists, and deterministic truncated SHA-256 hashing for dynamic entity identifiers. globs: src/telemetry/**/*.{ts,js}, src/observability/**/*.{ts,js}, src/instrumentation/**/*.{ts,js}, src/tracing/**/*.{ts,js}, src/server/**/*.{ts,js}, src/app/**/*.{ts,js} alwaysApply: true verified_by: "agentrules.org" license: "CC-BY-4.0" source: "https://agentrules.org/rules/otel-otlp-cardinality-circuit-breaker-standard" --- # OpenTelemetry OTLP High-Cardinality Dynamic Attribute Scrubbing & Circuit Breaker Guardrail > **Severity**: CRITICAL | **Category**: devops > **Anti-Pattern Warning**: Unbounded Runtime Span Attribute Injection Inducing OTLP Collector Memory Exhaustion and Storage Explosion ## System Constraints You are a Principal Observability and Infrastructure Architect enforcing industrial-grade OpenTelemetry OTLP high-cardinality safeguards. You must strictly mandate that all span creation and attribute mutation passes through an inline sanitization and circuit-breaking layer. Unbounded runtime keys, raw UUIDs, high-entropy query payloads, and arbitrary user IDs injected directly into span attributes trigger catastrophic memory allocation and index degradation in downstream time-series storage (Prometheus, Mimir, Tempo, ClickHouse). You must reject any direct, unvalidated span attribute insertion, requiring bounded sets, deterministic hash-masking, maximum attribute count caps, and active circuit-breaking drop strategies. ## Strict Guidelines - Engineers MUST ALWAYS sanitize span attributes through a dedicated SpanProcessor or bounded telemetry wrapper before dispatching to BatchSpanProcessor. - Engineers MUST NOT inject raw dynamic identifiers (such as session tokens, raw user IDs, query strings, or memory addresses) directly into OTLP span attributes. - Dynamic entity identifiers (e.g., tenant-specific UUIDs, transaction identifiers) MUST be masked or deterministically hashed using truncated SHA-256 (e.g., 12-char hex) when low-cardinality aggregation is required. - Services MUST enforce hard caps on span attributes: MAX_ATTRIBUTES_PER_SPAN <= 32, MAX_ATTRIBUTE_KEY_LENGTH <= 64 bytes, and MAX_ATTRIBUTE_VALUE_LENGTH <= 256 bytes. - The telemetry layer MUST implement an in-memory circuit breaker that sheds non-critical span attributes or drops spans entirely when memory pressure exceeds 85% or attribute ingest rates exceed safe bucket capacities. - Span attribute keys MUST be constrained to a strictly defined, validated allowlist or hierarchical namespace (e.g., 'app.*', 'db.system', 'http.route'). - Dynamic keys constructed via string concatenation (e.g., `span.setAttribute(`user_${id}`, value)`) are STRICTLY FORBIDDEN. ## Bad vs Good Implementations ### Unbounded Attribute Injection vs. Bounded Cardinality Guardrail Processor #### Bad Antipattern (AI Common Hallucination) > **Why it fails**: Directly sets raw UUIDs, authentication tokens, and arbitrary user-supplied object keys onto spans. This creates millions of unique index entries in downstream telemetry collectors, rapidly exhausts collector heap memory, and triggers OOM crash loops. ```typescript import { trace } from '@opentelemetry/api'; const tracer = trace.getTracer('checkout-service'); export function processOrder(orderId: string, userToken: string, rawPayload: Record<string, unknown>) { return tracer.startActiveSpan('processOrder', (span) => { // FATAL: High-cardinality identifiers & arbitrary keys injected directly into trace buffer span.setAttribute('order.id', orderId); // High-cardinality UUID span.setAttribute('auth.token', userToken); // PII + Infinite Cardinality // FATAL: Dynamic, unconstrained object key generation causing index explosion in Tempo/ClickHouse for (const [key, value] of Object.entries(rawPayload)) { span.setAttribute(`payload.${key}`, String(value)); } span.end(); }); } ``` #### Production Standard (Correct) > **Standard**: Implements strict cardinality bounding: hashes dynamic UUIDs with truncated SHA-256, strictly limits key space against an allowlist, truncates attribute value length, and drops unindexed/unapproved attributes before they reach the export buffer. ```typescript import { createHash } from 'node:crypto'; import { trace, Span, SpanKind } from '@opentelemetry/api'; const tracer = trace.getTracer('checkout-service'); const MAX_ATTRIBUTES_PER_SPAN = 32; const MAX_VALUE_LENGTH = 128; const ALLOWED_KEYS = new Set(['order.hash', 'service.environment', 'http.status_code', 'checkout.tier']); class CardinalityGuard { private static hashIdentifier(raw: string): string { return createHash('sha256').update(raw).digest('hex').substring(0, 12); } public static safeSetAttribute(span: Span, key: string, value: unknown, currentCount: number): boolean { if (currentCount >= MAX_ATTRIBUTES_PER_SPAN || !ALLOWED_KEYS.has(key)) { return false; // Circuit-breaker shedding dynamic or excess attributes } const strValue = typeof value === 'string' ? value : JSON.stringify(value); const boundedValue = strValue.length > MAX_VALUE_LENGTH ? strValue.substring(0, MAX_VALUE_LENGTH) + '...[TRUNCATED]' : strValue; span.setAttribute(key, boundedValue); return true; } public static hashEntityId(id: string): string { return this.hashIdentifier(id); } } export function processOrder(orderId: string, userToken: string, tier: string) { return tracer.startActiveSpan('processOrder', { kind: SpanKind.INTERNAL }, (span) => { let attrCount = 0; // Invariant: Bounded cardinality through truncated hashing and key allowlisting if (CardinalityGuard.safeSetAttribute(span, 'order.hash', CardinalityGuard.hashEntityId(orderId), attrCount)) { attrCount++; } if (CardinalityGuard.safeSetAttribute(span, 'checkout.tier', tier, attrCount)) { attrCount++; } span.end(); }); } ```

Architectural Specification & Invariant Rationale

Architectural Invariant Design: OpenTelemetry OTLP High-Cardinality Dynamic Attribute Scrubbing & Circuit Breaker Guardrail

OpenTelemetry OTLP High-Cardinality Dynamic Attribute Scrubbing & Circuit Breaker Guardrail defines an industrial-grade behavioral standard for devops engineering. Protects downstream telemetry collector pipelines, ClickHouse/Tempo backends, and in-memory trace buffers from catastrophic OOM crashes by dynamically bounding, hashing, and shedding unindexed dynamic runtime span attributes. Implements runtime span processors with token bucket rate limiters, strict key allowlists, and deterministic truncated SHA-256 hashing for dynamic entity identifiers. Modern software architectures require strict adherence to deterministic patterns to guarantee high availability, maintainability, and zero security vulnerabilities.

Framework Paradigm Shift & Drift MechanicsLLMs are trained on millions of open-source repositories containing outdated code, legacy idioms, and unpatched antipatterns. When coding assistants generate implementations for this stack, they default to common statistical tokens: Unbounded Runtime Span Attribute Injection Inducing OTLP Collector Memory Exhaustion and Storage Explosion Without an explicit system instruction file, developers spend hours debugging silent failures.

Failure Modes & Production Blast Radius

Hazard Analysis: What Breaks Without This Rule

The following failure modes have been cataloged from hundreds of hours of production code reviews where AI assistants produced non-compliant implementations:

Failure Mode	AI Symptom (The Anti-Pattern)	Severity	Production Impact & Remedy
AI Hallucination & Deprecation Drift	Unbounded Runtime Span Attribute Injection Inducing OTLP Collector Memory Exhaustion and Storage Explosion	Critical	Violating this industrial standard introduces regression bugs, security risks, and unexpected runtime failures in production. Enforce explicit behavioral system directives across all AI coding assistants in your repository.
Silent Architecture Inconsistency	AI assistants produce mixed paradigms across files, creating fragmented codebase conventions.	Medium	Increases cognitive load during code reviews and leads to maintenance debt across engineering teams. Pin project conventions in .cursor/rules, CLAUDE.md, and AGENTS.md so all team agents follow identical guardrails.

Architectural Directives in Detail

Step-by-Step Production Directives

Each rule constraint below is injected into the AI agent's system prompt to establish deterministic boundaries:

Engineers MUST ALWAYS sanitize span attributes through a dedicated SpanProcessor or bounded telemetry wrapper before dispatching to BatchSpanProcessor.

Rationale:AI models frequently generate unconstrained code here because statistical language distributions lack project-specific architectural context. Mandating this directive ensures that generated code adheres to verified industrial production requirements.