bigquery-troubleshooting

Diagnose root causes of BigQuery performance, compute, and storage cost incidents.

23.7k|2.0k|Updated Mar 31, 2026
Ready to run
npx skills add https://github.com/google/skills --skill bigquery-troubleshooting-google
💬 Prompt for ChatGPT / Claude Chat▼
Please help me install this Agent Skill.
Skill: bigquery-troubleshooting
Source: https://github.com/google/skills/tree/main/skills/cloud/bigquery-troubleshooting
Command: npx skills add https://github.com/google/skills --skill bigquery-troubleshooting-google

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? BigQuery incidents—slow queries, cost spikes, and execution failures—require rapid root-cause diagnosis across performance bottlenecks, reservation saturation, and billing models. Manual telemetry collection and multi-stage reasoning consume hours. This Skill automates the diagnostic funnel to isolate the root cause and recommend remediation in minutes. ## Core Features & Use Cases - Performance Troubleshooting: Diagnose single-job stage bottlenecks (join explosions, shuffle spills, slot contention) and compare against historical baselines to isolate slowdowns. - Compute Cost Diagnosis: Pinpoint whether cost spikes stem from the "Bully Query" (unpartitioned scans), BQML training rates, RLS redaction, On-Demand volume surges, capacity commitment expirations, or autoscaler thrashing. - Storage Cost Investigation: Identify historical partition 90-day timer resets (the DML trap), unpartitioned table active traps, Time Travel and Fail-Safe churn on physical billing, and dropped table drain periods. - Configuration Audits: Track reservation slot capacity changes, commitment timelines, baseline reductions, and assignment reconfigurations correlated with performance regression windows. - Use Case: Your BigQuery bill jumped 40% overnight. Use this Skill to scan the last 7 days against the prior week, isolate whether the spike stems from On-Demand query volume, capacity commitment expiration, or storage tier promotion, then recommend concrete remediation (partition enforcement, commitment renewal, or Time Travel reduction). ## Quick Start Analyze why BigQuery queries have been running slower over the past two days by comparing today's performance metrics against the same timeframe from last week.

Frequently Asked Questions about bigquery-troubleshooting

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

FAQPage Schema
How do I find which query caused a BigQuery cost spike?▼

Query INFORMATION_SCHEMA.JOBS grouped by query_info.query_hashes.normalized_literals to identify top-billing queries by total_bytes_billed. Compare the current period against a baseline week using absolute TiB volume and percentage growth to isolate cost anomalies by user or query template.

What causes BigQuery queries to suddenly run slow?▼

Slowness stems from slot contention (reservation saturation or noisy neighbors), data volume growth (unpartitioned scans or missing pruning), stage bottlenecks (join explosions or shuffle spills), or configuration changes (baseline reduction or commitment expiration). Use execution graph analysis and comparative baseline queries to isolate the root cause.

Why did my BigQuery storage bill increase even though I deleted tables?▼

Under physical storage billing, deleted tables continue incurring charges during the Time Travel window (2-7 days) plus an additional 7-day Fail-Safe retention period. Deleting long-term storage data promotes it to active physical storage at higher rates, causing temporary cost spikes before the drain period expires.

How do I prevent historical partitions from resetting the 90-day Long-Term discount timer?▼

Add partition pruning predicates (WHERE _PARTITIONDATE >= DATE_SUB(CURRENT_DATE(), INTERVAL 7 DAY)) to all ETL pipelines. Unscoped UPDATE, DELETE, or MERGE queries touching even one row in old partitions reset the last_modified_time, triggering immediate reversion from 50% discounted Long-Term pricing to full Active rates.

Does BigQuery troubleshooting require different tools for On-Demand versus Editions compute models?▼

Yes. On-Demand diagnoses focus on total_bytes_billed and per-query scans via INFORMATION_SCHEMA.JOBS. Editions troubleshooting examines slot utilization, commitment coverage, and autoscaling headroom via INFORMATION_SCHEMA.RESERVATIONS_TIMELINE and INFORMATION_SCHEMA.CAPACITY_COMMITMENT_CHANGES_BY_PROJECT.

What happens if my capacity commitment expires without auto-renewal?▼

Reserved slot capacity instantly reverts to uncovered Pay-As-You-Go (PAYG) baseline billing at 24/7 variable hourly rates, typically doubling compute costs for steady workloads. Verify commitment status in INFORMATION_SCHEMA.CAPACITY_COMMITMENT_CHANGES_BY_PROJECT and enable auto-renewal or purchase new commitments to avoid sudden cost spikes.