Runbook Design Principles

A runbook is an important document for cloud migration. It outlines the process and steps for service cutover, guiding team collaboration. A runbook has two parts: a checklist and an operation procedure. This section explains how to design the cutover runbook in several ways.

Runbook design principles:

• A runbook is created for a specific cutover.
• A runbook should provide details of the cutover procedure, operators, confirmers, and estimated start time, end time, and execution time.
• Each step should be clear for one operator to do and one confirmer to check. This avoids having multiple people confirm the same step.
• A runbook should list specific commands. Use scripts or tools for the runbook so operators can run it directly without changes, reducing the risk of errors.
• The runbook can have both parallel and serial operations. Clearly mark the order of these operations to avoid errors that could affect the execution time and results.
• The runbook procedure might fail. Predict possible issues and decide in advance whether to roll back or continue the cutover. This prevents delays in decision-making.

Design principles of rollback conditions:

• Each cutover phase has a set completion time. If it is not done by then, decide if you need to roll back.
• If the core table data comparison shows inconsistencies, decide if you need to roll back.
• If the core P0 test fails, decide if you need to roll back.
• If the performance does not meet the expectation, decide if you need to roll back.