github.com/getlantern/http-proxy

HTTP/S Proxy with extensions for Lantern

Note - this project versions its dependencies using dep.

Just run dep ensure to download the vendored packages.

These are Lantern-specific middleware components for the HTTP Proxy in Go:

• A filter for access tokens

• A filter for devices, based on their IDs

• A filter for Pro users

• A connection preprocessor to intercept bad requests and send custom responses

• Custom responses for mimicking Apache in certain cases

Usage

Build it with dep ensure && go build or with make build.

To get list of the command line options, please run http-proxy-lantern -help.

config.ini.default also has the list of options, make a copy (say, config.ini) and tweak it as you wish, then run the proxy with

http-proxy-lantern -config config.ini

To regenerate config.ini.default just run http-proxy-lantern -dumpflags.

Testing with Lantern extensions and configuration

Run tests

go test

Use this for verbose output:

TRACE=1 go test

Manual testing

Keep in mind that cURL doesn't support tunneling through an HTTPS proxy, so if you use the -https option you have to use other tools for testing.

Run the server with:

cd http-proxy
go install && http-proxy -https -token <your-token> -enablereports -stackdriver-project-id http-proxy-lantern -stackdriver-creds /Users/afisk/lantern_aws/salt/http_proxy/lantern-stackdriver.json

Run a Lantern client accordingly, as in:

./lantern -force-proxy-addr localhost:8080 -force-auth-token <your-token>

You have two options to test it: the Lantern client or checkfallbacks.

Keep in mind that they will need to send some headers in order to avoid receiving 404 messages (the chained server response if you aren't providing them).

Currently, the only header you need to add is X-Lantern-Device-Id.

If you are using checkfallbacks, make sure that both the certificate and the token are correct. A 404 will be the reply otherwise. Running the server with -debug may help you troubleshooting those scenarios.

Handle requests config server specially

To prevent spoofers from fetching Lantern config with fake client IP, we need to attach auth tokens to such requests. Both below options should be supplied. Once http-proxy-lantern receives GET request to one of the cfgsvrdomains, it sets X-Lantern-Config-Auth-Token header with supplied cfgsvrauthtoken, and X-Lantern-Config-Client-IP header with the IP address it sees.

  -cfgsvrauthtoken string
        Token attached to config-server requests, not attaching if empty
  -cfgsvrdomains string
        Config-server domains on which to attach auth token, separated by comma

When something bad happens

With option -pprofAddr=localhost:6060, you can always access lots of debug information from http://localhost:6060/debug/pprof. Ref https://golang.org/pkg/net/http/pprof/.

Be sure to only listen on localhost or private addresses for security reason.

Building for distribution and deploying

When building for distribution make sure you're creating a linux/amd64 binary and that the resulting binary is compressed with upx.

You can use the following command to do all this automatically:

make dist

Once you've build a binary for distribution, you can deploy it to all live proxies with:

make deploy

Temporarily Deploying a Preview Binary to a Single Server

Sometimes it's useful to deploy a preview binary to a single server. This can be done using either deployTo.bash or onlyDeployTo.bash. They do the same thing but deployTo.bash first runs make dist whereas onlyDeployTo.bash copies the existing binary at dist/http-proxy.

Deploy Preview

./onlyDeployTo.bash <ip address>

Revert to Production Binary

Once you're done checking out the preview, revert back to the production binary with:

./revertToProductionBinary.bash <ip addres>