A simple proxy for API versioning
Diversion is a versioning HTTP proxy based on bouncy and semver. It uses semver to parse version numbers (or ranges) and then bounces the request to the highest-versioned backend that satisfies the requested using the maxSatisfying algorithm. This means you have all the power of npm's version ranges (including ~0.2.1 style matching) for your proxied services.
npm install diversion
diversion config.json # See example_config.json to get started
By default the proxy will look for a version/range in the 'X-Version' header, but it can also check the first path component of the requested URL for a version and use that instead if you set "useURL" to true in your config.
Also included is an optional management API that listens on a different port and allows on-line updating of the proxy configuration. To enable this service include a 'management' port in the 'ports' object in your config. Note that there is no authentication on this service, so don't expose the port to anybody you don't want to be able to modify the proxy.
The management service also serves a monitoring page on /status where you can see the state of registered backends, add and remove backends, and watch requests get forwarded.
If you'd rather not run the management service, you will need to write your state file manually and make sure your config.json has the correct path to it. Here is an example state.json file to get you started:
"backends":"0.1.1":"127.0.0.1:3000":"alive": true"healthCheckPath": "/""127.0.0.1:3001":"alive": true"healthCheckPath": "/""0.2.1":"127.0.0.1:3300":"alive": true"healthCheckPath": "/""127.0.0.1:3301":"alive": true"healthCheckPath": "/"
If 'pollFrequency' is set to a non-zero value in the config, diversion will poll any backends that specified a healthCheckPath and update their live/dead status according to whether they respond with 200 status codes.
When an error occurs, and health checks are being performed, diversion will disable the backend that caused the error and continue to poll it until it come back online.
If 'maxRetries' is set to a non-zero value in the config, diversion will try the next live backend up to 'maxRetries' times or it runs out of live backends.