One of my favorite experiences at conferences and events is talking to the open-source Airbyte community. Airbyte Core (open source) is beloved by the data community, and for good reason. But as with any self-hosted software, there are tradeoffs.
You chose OSS for the control and the zero licensing cost, and it delivered. But that means you also own the upgrades, the infrastructure, the scaling, the security scan findings, and the 2am pages when a sync fails. This is why many teams start exploring our paid plans. Airbyte Cloud and Airbyte Flex run the same connectors and the same sync engine as Airbyte Core. The difference is who runs the platform and where your data is processed. So from a technical perspective, the shift looks easy.
But a lot of teams stop when it actually comes to the migration, and we completely understand why.
Many of you have put hours of work into your pipelines and infrastructure on Airbyte, and until recently, migration has not been easy. Moving to a new instance has traditionally meant recreating everything by hand and re-syncing from scratch. That's costly in warehouse compute and in engineering hours.
We've built tooling that removes that cost. It migrates your workspace, connections, and sync state from a self-managed instance to Airbyte Cloud or Airbyte Flex, so your first sync after the move picks up where your last sync in OSS left off.
Cloud or Flex: decide where your data runs The migration works the same way for both. The choice comes down to where your data needs to live.
Airbyte Cloud is fully managed. We run the platform and the infrastructure, and you stop operating Airbyte altogether. If your data can be processed in Airbyte's managed environment, this is the shortest path off self-hosting.
Airbyte Flex keeps the data plane in your own infrastructure, whether that's a VPC in your cloud account or on-prem, while Airbyte manages the control plane. Record data is read and written inside your boundary and never passes through Airbyte's infrastructure. If you self-host today because data has to stay inside your boundary or in a specific region, Flex keeps that guarantee and moves you onto a managed control plane.
For a lot of OSS teams, Flex is the more natural move. You already run Airbyte in your own infrastructure, and the engineer who deployed OSS is usually the same person who sets up the Flex data plane. A pattern we see often: a team running OSS on Kubernetes in their own cloud account wants off because of security scan findings and maintenance load, but the data has to stay in region. Cloud doesn't fit that requirement. Flex does.
Moving to either one also gets you the governance features Airbyte Core doesn't include, like SSO, role-based access control, and audit logging, depending on your plan.
What the migration preserves Three things make a migration painful if you lose them:
Your resource configuration. Every source, destination, and connection, with its streams, sync modes, and schedules.Your sync state. The checkpoints that tell each connection where the last incremental sync ended.Your existing data. This is the data already stored in your warehouse, containing your entire business history and context.The migration tooling handles the first two directly, and because your destinations don't change, the third never moves at all. Your warehouse keeps all existing data, and your new Cloud or Flex workspace continues writing directly to it without missing a beat.
How it works The process runs in five stages, and a member of our Solutions team manages it end to end. You aren't handed a runbook and left alone with it. Your part is running the export, approving each stage, adding credentials, and choosing when to cut over.
1. Export and check your existing instance. You run our export tool against your OSS instance. It's read-only: it captures your resource configuration and connection state (sync checkpoints plus metadata) and changes nothing on your instance. You send the output back to us, and we turn it into migration-ready Terraform describing all of your resources, plus a plan summary that shows exactly what will be created. You don't need to know Terraform; it's the migration vehicle, and we drive it. Before anything is applied, we run the preflight checks described in the next section.
2. Stand up sources and destinations. We apply the generated Terraform to create your sources and destinations in your new Cloud or Flex workspace. Secrets never leave your instance, so this is your main hands-on step: you add credentials and test each connector in the UI. This is deliberate, since it confirms every source and destination is healthy before any connection exists.
3. Create the connections. A second apply creates your connections. They come up inactive and on a manual schedule, so nothing syncs before you're ready. Each new connection is mapped back to its original, which gives you a clear audit trail from old instance to new.
4. Transfer state and restore settings. We apply the saved sync checkpoints to the new connections, then restore the original names, schedules, and active status. At this point your new workspace mirrors your OSS instance, checkpoints included.
5. Validate. Together we run your syncs and confirm they resume from the correct checkpoint rather than re-syncing all your existing data. Incremental connections continue incrementally.
The connections stay inactive until step four, so you choose the cutover moment. There's no window where two instances are syncing the same connection into the same destination.
What we check before anything is created Most migration failures are predictable. We've hit enough of them in real migrations that we now check for them up front, using your exported configuration, before any Terraform is applied:
Duplicate stream names. Two streams with the same name in one connection can make an apply fail partway through.Missing streams. Streams in your configuration that the source no longer returns, which would otherwise surface as errors at apply time.Connector major-version mismatches. A connector pinned to a different major version than the one you'll run after the move. Major versions can include breaking changes to schemas or state, so we sort these out before any state moves.Destination drift. Differences between what your connections expect in the destination and what's actually there.You get a report with what we found, the evidence behind each finding, and the open items to resolve. We work through those with you before applying anything.
Why the checkpoint transfer matters We all know that re-syncing existing data is slow. For sources with strict rate limits, a full historical re-sync can take longer than the original backfill did, and some APIs can't serve old data at all.
There's also a cost element. On the destination side, reprocessing months of records means real warehouse spend and, depending on your sync modes, deduplication work to clean up afterward.
Transferring state sidesteps all of it. A connection that was syncing incrementally on your OSS instance on Monday syncs incrementally on your new workspace on Tuesday, from the same cursor.
We've done this with real workspaces This has been built the right way, alongside real customers managing production-grade data infrastructure.
We built this tooling by migrating actual customer instances, including large enterprise deployments, with the checkpoints, quirks, and connector variety that real production workspaces accumulate. Each migration surfaces something new, and those lessons become checks we run on the next one. Those teams are now running with their existing data intact, and the stages above are the same ones we ran with them.
Who should consider this If any of these sound familiar, the migration is worth a conversation:
You spend more time upgrading and debugging your Airbyte deployment than building with the data it moves. Your security team keeps flagging findings on your self-hosted deployment, and patching them falls on you. Your data has to stay in your own environment or region, and that's the main reason you still self-host. Look at Flex. You need SSO, RBAC, or audit logging, which Airbyte Core doesn't include. Your team is small enough that self-hosting is a tax on your actual roadmap. You want managed scaling, support SLAs, or predictable capacity-based pricing on Pro or Enterprise Flex, without rebuilding what you have. The value is real. Based on a Forrester TEI Study , teams moving off self-managed data pipelines to managed Airbyte see 239% ROI and payback in under six months.
This isn't a push to get you off open source. Airbyte Core stays free and self-managed, and plenty of teams should keep running it. But if the only thing keeping you on OSS is fear of the move, we're confident our migration tooling addresses that concern, whether you land on Cloud or Flex.
Get started Reach out to your account team or contact us , and we'll walk through your instance, help you choose between Cloud and Flex, and plan the cutover with you. Bring your gnarliest workspace… we're up for the challenge!