observability

Diagnose distributed-system operation failures using Jaeger traces and OpenTelemetry instrumentation.

Updated Aug 23, 2026
One-click install
npx skills add https://github.com/kpiteira/ktrdr --skill observability-kpiteira
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: observability
Source: https://github.com/kpiteira/ktrdr/tree/main/.claude/skills/observability
Command: npx skills add https://github.com/kpiteira/ktrdr --skill observability-kpiteira

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill enables rapid diagnosis of distributed system issues by leveraging OpenTelemetry instrumentation and Jaeger traces, providing complete context for operation failures, slowness, or communication problems across all services.

Core Features & Use Cases

  • Observability-First Debugging: Prioritizes Jaeger traces over logs for first-response diagnosis, offering full visibility into distributed operations and their interactions.
  • Common Diagnostic Patterns: Provides specific Jaeger queries and analysis techniques for "stuck," "failed," or "slow" operations, and service communication failures.
  • Key Span Attributes Reference: Explains important attributes within traces for understanding operation status, worker selection, progress, error context, and performance metrics.
  • Use Case: If a training operation fails with an unclear error, use this skill to query Jaeger with the operation ID, identify the exact span where the error occurred, and extract the exception type and message, leading directly to the root cause.

Quick Start

A user reported a "stuck" training operation with ID op_training_12345. Use the observability skill to query Jaeger and diagnose the issue.

Frequently Asked Questions about observability

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I diagnose slow or failing operations in distributed systems?

Diagnose distributed-system failures by querying Jaeger traces with operation IDs to identify which span failed, then examine error tags and exception attributes to pinpoint the root cause across services.

What information should I look for in Jaeger trace spans?

Key span attributes include operation.id, error status, exception messages, http.* tags, and worker selection metadata—these reveal operation status, communication failures, and performance bottlenecks across your service mesh.

Do I need OpenTelemetry instrumentation to use Jaeger for debugging?

Yes, OpenTelemetry instrumentation is required to generate the traces Jaeger ingests; without it, no trace data exists to query or analyze for diagnosing distributed-system issues.

How do I identify inter-service communication problems with Jaeger?

Query Jaeger traces by operation ID, then examine spans across service boundaries for missing or delayed propagation, network errors, and http.* attributes that reveal where requests fail between services.

Can I use Jaeger traces to troubleshoot stuck operations?

Yes, query Jaeger with the operation ID to see which span is blocked, check for missing completion signals, worker selection issues, or timeouts—trace timings reveal exactly where operations hang.

What's the difference between using Jaeger traces versus logs for debugging?

Jaeger traces provide complete operation context across all services and timing relationships in one view, whereas logs are fragmented by service; traces enable faster root-cause diagnosis for distributed failures.