compute-management_skill

Overview

This skill is an expert guide for managing Oracle Cloud Infrastructure (OCI) compute instances. It helps with launching instances, resolving out-of-capacity or boot failures, optimizing compute costs, and handling instance lifecycle and security best practices. Use it when you need practical, OCI-specific operational guidance and decision flows.

How this skill works

The skill codifies anti-patterns, a capacity error decision tree, shape selection guidance, cost calculations, instance principal setup, and OCI-specific gotchas. It directs operators to check service limits first, choose appropriate shapes (Flex vs fixed), and use secure access patterns like instance principals and bastions. For CLI/API actions, it advises loading exact command references rather than guessing syntax.

When to use it

• When launching production or test instances and selecting shapes
• When you encounter “out of host capacity” or quota-related failures
• To optimize compute cost and right-size instances
• During boot failure or access troubleshooting (serial console use rules)
• When implementing instance-to-OCI authentication (instance principals)

Best practices

• Always check OCI service limits before attempting launches; many capacity errors are quota issues
• Prefer AD-agnostic templates and use fault domains instead of hardcoding Availability Domains
• Use flexible shapes (Flex) for cost/performance tuning; compare Flex pricing versus fixed shapes
• Never enable public IPs for production instances; use bastion or private endpoints
• Use instance principals for API access instead of embedding user credentials; employ dynamic groups and least-privilege policies

Example use cases

• Troubleshoot an out-of-capacity error: run the limits resource-availability check, try different ADs, then switch shape family or create a capacity reservation
• Right-size an application: start small (e.g., 2 OCPUs) and scale based on metrics; compute monthly cost via (OCPUs×$/OCPU + GB×$/GB)×730
• Recover a failed boot: use serial console only as a temporary troubleshooting tool and remove connections immediately after
• Migrate cost-sensitive workloads to ARM (A1.Flex) after compatibility testing to reduce OCPU costs
• Implement secure service calls from instances using dynamic groups and instance principal signers