Dynamic Request Routing with fly-replay

The fly-replay feature gives you fine-grained control over request routing in your Fly.io applications. It allows you to dynamically route requests between regions, specific Machines, or even different apps within your organization.

How fly-replay Works

When your app responds to a request with a fly-replay response, Fly Proxy will automatically replay the original request according to your specified routing rules. This enables advanced patterns like:

• Routing write operations to primary database regions
• Load balancing between specific Machines
• Cross-app request routing within your organization
• Implementing sticky sessions

Replay Header Format

Your app can add a fly-replay header to its response. The fly-replay header accepts the following fields:

Example Usage

Route to a specific region:

fly-replay: region=sjc

Route to one of the given regions, in order of preference:

fly-replay: region="iad,ord,us,na"

Route to a preferred region, or fallback to any available machine:

fly-replay: region="sjc,any"

Route to specific Machine:

fly-replay: instance=00bb33ff

Route to another app:

fly-replay: app=app-in-same-org

You can combine multiple fields:

fly-replay: region="sjc,any";app=app-in-same-org

Geographic groups and aliases

When replaying to a region, you can use geographic aliases like us, eu, or sa to target a broader area.

Replay JSON Format

Your app can set the response content-type to application/vnd.fly.replay+json and include replay instructions in the response body. For a complex replay the JSON body is easier to compose than the header format, and also supports more functionality.

JSON Structure

The application/vnd.fly.replay+json replay body accepts the following fields:

Example Usage

Route to another app, and modify the request:

{
  "app": "target-app",
  "region": "iad,us",
  "transform": {
    "path": "/new/path?param=value",
    "delete_headers": ["x-unwanted-header", "cookie"],
    "set_headers": [
      { "name": "x-custom-header", "value": "new-value" },
      { "name": "authorization", "value": "Bearer token123" }
    ]
  }
}

Replay Caching

In cases where the replay target does not change often, you can use the fly-replay-cache mechanism to relieve the original app of some load. Here’s how it works:

• Issue a fly-replay header as usual
• Set fly-replay-cache: /some/path/*
  • This needs to be a path-matching pattern ending with a wildcard /*. This is where the cached replay is applied.
  • The cached replay is unique to the request domain. The domain may also be set explicitly: example.com/some/path/*
  • The domain part should not include ports; use example.com, not example.com:80.
  • The pattern must also match the current request.
• Set fly-replay-cache-ttl-secs: number_of_seconds

If the replay target does eventually change, the replay target may proactively invalidate the cache by

• Issuing a fly-replay back to the origin
• Setting fly-replay-cache: invalidate

For apps using the JSON format, specify a cache object:

{
  "app": "target-app",
  "cache": {
    "prefix": "/some/path/*",
    "ttl": 60
  }
}

{
  "app": "origin-app",
  "cache": {
    "invalidate": true
  }
}

Bypassing the replay cache

Sometimes, you might want a client to bypass a cached replay without actually invalidating the cache. For example, you might have an API endpoint that mostly needs to be replayed to a “leader” instance. But depending on application logic, it could sometimes be handled by any instance.

The recommended approach is to clearly separate endpoints that require replay from those that don’t. However, when that is not possible, the replaying app can additionally set the header

fly-replay-cache-allow-bypass: yes

or set the JSON field "allow_bypass": true. Then, when your client makes a request, it can set the header:

fly-replay-cache-control: skip

to skip any cached replay that might be present.

Was my request served by the cache?

If your app needs to know whether a request was served from the replay cache, it can check the fly-replay-cache-status header. This header is sent to the replay target app or machine.

• Absent: the request was not replayed and was delivered directly to the app or machine.
• miss: the request didn’t use replay cache; it first hit another machine and got replayed here.
• hit: the request did use the replay cache! It never reached the source app or machine and was redirected straight to the target.
• bypass: even though the cache might have been set, it was explicitly skipped by setting the fly-replay-cache-control: skip header.

Implementation Details

Requirements and Limitations

• Your app must use the http handler
• Requests larger than 1MB cannot be replayed
• Field combinations must be logically valid (e.g., don’t specify both app and instance if instance isn’t in that app)

For large uploads that exceed the 1MB limit, consider:

• Using direct-to-storage uploads where possible
• Using the fly-prefer-region header instead

For fly-replay-cache, the following limitations apply:

• The state field cannot be set in the fly-replay intended to be cached
• Transformations for the headers or path cannot be defined.
• The TTL needs to be a minimum of 10 seconds
• Only one step of lookup is performed in the cache; as such, if the target app issues another fly-replay-cache, the caching behavior in this case is undefined
• The fly-replay-src header (described below) will not be set for requests replayed through the cache

The fly-replay-src Header

When a request is replayed, Fly Proxy adds a fly-replay-src header containing metadata about the original request:

This header is useful for tracking request paths and implementing consistency patterns. See the official Ruby client for an example of using these fields to prevent read-your-write inconsistency.

This header is not set when the request is replayed through a cached fly-replay entry (fly-replay-cache).

The fly-preferred-instance-unavailable Header

If you replay with prefer_instance set, Fly Proxy will attempt to route to this Machine. This may not happen for a number of reasons, for example the Machine may not be found, or found but at its configured hard_limit.

In these cases, the request will be delivered to a different Machine that matches the remaining fields in your replay. Along with the other Fly.io-specific headers, a fly-preferred-instance-unavailable header will be set containing the ID of the instance that could not be reached.

Web Socket Considerations

It is worth noting that an application returning fly-replay headers should not negotiate a web socket upgrade itself. Some frameworks automatically handle this process. Instead, the application or instance receiving the requests should handle the upgrade.

Alternative Routing Headers

For cases where fly-replay isn’t suitable, Fly.io provides two alternative request headers:

The fly-prefer-region Header

fly-prefer-region: ams

Attempts to route directly to specific region(s). You can specify multiple regions as comma-separated values for fallback preferences:

fly-prefer-region: iad,ord,us

Falls back to the next region in the list, or to the nearest region with healthy Machines if none of the specified regions are available. Useful for large uploads that can’t be replayed.

The fly-prefer-instance-id Header

fly-prefer-instance-id: 90801679a10038

Attempts to route to a specific Machine. Falls back to any matching Machine if the target Machine cannot be found, or is unable to accept requests.

If the request is delivered to a different Machine, the fly-preferred-instance-unavailable request header will be set.

The fly-force-instance-id Header

fly-force-instance-id: 90801679a10038

Forces routing to a specific Machine. The Fly Proxy will attempt to reach the Machine multiple times if the Machine is unhealthy. No fallback if Machine is ultimately unavailable.

Common Use Cases

Multi-Region Databases

When using global read replicas, use fly-replay to ensure write operations go to the primary region:

fly-replay: region=sjc

See our blueprint for Multi-region databases and fly-replay for a complete implementation guide.

Cross-App Routing

Use fly-replay to implement routing layers or FaaS-style architectures:

fly-replay: app=customer-function-app

This allows you to build router apps that can dynamically route requests to other apps in your organization.