Local LLM / Local deployment and operations
Local LLM: Operational runbook
Document startup, model inventory, common failures and rollback commands for the deployed stack.
Environment and inputs: Self-hosted inference and retrieval services; gateway, OS and runtime settings require deployment-specific configuration.
When to use it
Treat the model server as a production dependency with an owner and a recovery path. Verify the actual network boundary and running configuration, then retain enough sanitized evidence to explain failures without collecting unnecessary prompt content.
Practical example
Example operator checklist. Fill in deployment-specific service names, approved commands and recovery owners before use.
1. Identify the affected service and model revision.
2. Check readiness, queue depth and recent error classes.
3. Preserve sanitized evidence and operation identifiers.
4. Apply the documented recovery or rollback procedure.
5. Repeat the critical task fixture and record the outcome.Check the result
Have another operator recover a staging failure using only the runbook.
Save and compare results
Save the test input, configuration and result. Fix the failing boundary, repeat the same test and compare it with the saved working configuration.
Scope and limits
These exercises have not been tested on your systems. Check feature support and configuration names against the installed release and linked documentation.