cloud-run-alert-configuration

Configures SRE-driven alerting policies for Cloud Run services, jobs, and worker pools using PromQL and Terraform.

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

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) and scripts (resource) components.

What problem does it solve? Production Cloud Run workloads require rigorous observability to detect failures, scaling exhaustion, and cost anomalies before they impact users. Manual alerting configuration is error-prone and misses critical signals. This Skill automates best-practice alert policy generation grounded in SRE practices, eliminating guesswork and ensuring comprehensive coverage. ## Core Features & Use Cases - Service Availability SLOs: Fast/slow burn rate alerts for 99.9% availability targets with configurable 5xx error thresholds. - Instance Saturation Monitoring: Warn at 80% and alert critically at 95% of max_instances ceiling to prevent scaling failures. - Resource Protection: OOM prevention (P95 memory ≥90%), CPU throttling detection (P95 CPU >85%), and billable instance time surge alerts (>3x baseline). - Job Execution Failure Detection: Immediate alerts (duration=0s) on any failed Cloud Run job execution. - Worker Pool Queue Health: Task success SLOs, estimated queue drain time (ETD), and oldest message age monitoring. - Use Case: Deploy a production video-processing service. Use this Skill to generate Terraform alert policies monitoring 5xx errors, instance saturation, container memory/CPU, latency SLOs, and billable instance time surges—then apply them to centralized Cloud Monitoring metrics scope. ## Quick Start Provide the Cloud Run service name, maximum instance count, and target notification channels, and I will generate complete Terraform alert policy configuration with all default SLO targets and thresholds applied.

Frequently Asked Questions about cloud-run-alert-configuration

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

FAQPage Schema
How do I set up Cloud Run service availability SLO alerts in Terraform?▼

This Skill generates fast-burn (1h/5m windows at 14.4× threshold) and slow-burn (3d/6h windows at 1.0× threshold) alert policies using PromQL against `run.googleapis.com:request_count` with `response_code_class="5xx"` filters. Set your SLO target (default 0.999) and notification channels; Terraform applies them to your scoping project or metrics scope.

What is the difference between fast burn and slow burn alerts?▼

Fast burn (1h/5m windows, 14.4× rate multiplier) triggers immediately when error rates spike, causing high-urgency pages. Slow burn (3d/6h windows, 1.0× multiplier) fires warning tickets when error budgets deplete over days, enabling planned remediation before SLA breach.

Why does the skill skip discovery when I name the Cloud Run service?▼

Named services bypass `gcloud` CLI and file scanning, enabling immediate parameterized Terraform generation. This fast path is critical when working with services in large monorepos or when gcloud authentication is unavailable—the skill uses explicit HCL variables with documented defaults.

Does this Skill support Cloud Run Jobs and Worker Pools?▼

Yes. Jobs route to immediate failure detection (duration=0s on any `result="failed"` execution). Worker Pools apply task success SLOs via Pub/Sub subscription metrics, queue backlog drain time (ETD), and oldest unacked message age alerts. Each workload type has dedicated reference guides.

What does the mandatory `created-with-google-skill` user label do?▼

This label identifies all generated alert policies as skill-originated for resource governance, cost attribution, and audit compliance. It is required in all `google_monitoring_alert_policy` resources and cannot be removed without re-running the skill.

Why are long lookback windows (>25h) omitting the `duration` parameter?▼

Cloud Monitoring's PromQL engine rejects `duration` settings on lookbacks exceeding 25 hours with INVALID_ARGUMENT errors. For 3d/7d slow-burn queries, the skill omits `duration` entirely to comply with API validation constraints.