render-oss/skills

Overview

This skill debugs failed Render deployments by analyzing logs, metrics, and database state to identify root causes and recommend fixes. It surfaces common errors like missing environment variables, port binding issues, OOM, build failures, and database connection problems. Use it to triage failing deploys, crashing services, and production errors quickly.

How this skill works

The skill queries Render via MCP tools or the CLI to list services, retrieve build/runtime logs, and fetch metrics. It matches log patterns to a catalog of known error classes, inspects CPU/memory and latency metrics, and runs targeted database checks when needed. For identified issues it recommends concrete fixes (update env vars, change port binding, add dependencies, increase plan) and provides the commands or steps to apply and verify the fix.

When to use it

• A deployment fails or never finishes
• A service won’t start or keeps crashing
• Users report errors, 5xx responses, or timeouts
• Health checks are timing out or failing
• Performance degradations or high latency are observed
• Database connections are failing or saturating connections

Best practices

• Prefer MCP tools (list_services, list_logs, get_metrics) for structured data; fall back to the CLI when MCP is not configured
• Collect build and runtime logs before making changes to reproduce the failure pattern
• Match errors to known patterns (missing env, port binding, OOM, dependency errors) and apply grouped fixes if the same error repeats
• Verify fixes by redeploying and checking recent deploy status, error logs, and key metrics
• When changing resource limits or plans, confirm memory and cpu usage trends to avoid repeated OOMs

Example use cases

• Deploy failed with ‘Command failed’ during build: fetch build logs, identify missing dependency, update package manifest, redeploy
• App crashes with exit 137: pull memory metrics for the service, increase memory or optimize allocation, redeploy and monitor memory_usage
• Users see 502/503 errors: search error logs for ECONNREFUSED, check DATABASE_URL and DB instance status, inspect DB active connections and fix credentials or scale DB
• Health check timeout on startup: verify app binds to 0.0.0.0:$PORT and expose /health endpoint, then update start command and redeploy