A zero-downtime route from Supabase Storage to Amazon S3

A Retail business working with Cloud Combinator on AWS. The client is anonymised at their request.

The migration had three focus areas, and none of them was the file copy itself.

Cloud Combinator structured the move so that every step is reversible until the final deletion. Access runs through one storage adapter, and behaviour is controlled by read and write target flags, so the team can advance or roll back by changing configuration rather than code.

This case study describes the migration plan Cloud Combinator prepared and readied for delivery with the client, so the figures below are the design targets of the engagement rather than measured production results. The plan gives the client a settled technical approach with the risky work identified, sequenced and made reversible.

• Access control preserved — Presigned S3 URLs and CloudFront Origin Access Control reproduce Supabase's ownership rules, so private media stays private and buckets stay locked.
• No broken links — Stored URLs are normalised to object keys and the key structure is kept identical across both stores, making cutover a configuration flip rather than a rewrite.
• Fully reversible — Read and write flags plus a read-through fallback mean any phase before the final deletion can be rolled back by changing configuration, with Supabase still holding every object.
• Predictable cost — Data transfer into S3 is free, the one-off Supabase egress for a 1 TB backfill is on the order of tens of dollars, and ongoing delivery moves to CloudFront with a CDN included.
• Handover — Cloud Combinator delivers the AWS side as a deployable CloudFormation and SAM template, plus the rclone backfill runbook and the storage adapter in the client's own stack language.

With the plan locked at kickoff, the client moves onto a S3 and CloudFront storage foundation that is no longer tied to a single provider, and is ready to support whatever the product takes on next.