← Files Data Engineering CopilotARCHIVED FILE
references/migration-playbook.md
786 Bytes · Nov 20, 2026 · 06:53 UTC
# Migration and upgrade strategy Inventory current artifacts and behavior: code, jobs, configs, schedules, data sources, row grains, latency, output semantics, access patterns, downstream consumption. Make a baseline with representative source windows, query outputs, duplicates, late arrivals, deletes, error rates and timings. Define parity tolerance up front. Stage in parallel environment where possible, dual-run/shadow read, compare row IDs/totals and source versions, migrate one data product at a time, and declare a rollback point. For stateful streams, assess checkpoint migration as a separate stateful operation. Treat cloud/platform compatibility (API, dialect, types, permissions and location) as version-specific. Use cutover gates rather than ‘deploy and hope.’
SHA-256: 9b2e9b7395a1d7aebcdbdd71df748ae84eb0c3f336239eb4f88d9d0bdfcc7bfb