ThriftServer

Encapsulate bolierplate code and functionality for running Thrift servers in Ruby. Bundled functionality:

• Metrics for server & RPCs
• Logging for server events & RPCs
• Middleware & pub-sub based approach making it easy to extend
• Deep validation on outgoing protocol messages
• Thrift::ThreadPoolServer & Thrift::ThreadedServer support

Add this line to your application's Gemfile:

gem 'thrift_server'

And then execute:

$ bundle

Or install it yourself as:

$ gem install thrift_server

The library uses delegation to around a provided handler & Thrift processor to implement use production server behavior. There are two different ways to extend function: pub sub & middleware. Use pub sub for events & middleware when you want to modify the request & response before/after hitting the handler. Out of the box there is not extra behavior. Here's the bare bones useful server.

server = ThriftServer.threaded EchoService, Handler.new

server.log Logger.new($stdout)

server.start

The first arument is a module containing a generated Thrift::Processor subclass or the module itself. ThriftServer provides factories for building Thrift::ThreadedServer and Thrift::ThreadPoolServer instances.

server = ThriftServer.threaded EchoService, Handler.new

server.log Logger.new($stdout)

server.start

ThriftServer.threaded accepts the following options:

• port: - port to run server on. Defaults to 9090.

server = ThriftServer.thread_pool EchoService, Handler.new

server.log Logger.new($stdout)

server.start

ThriftServer.threaded accepts the following options:

• threads: - Pool size. Defaults to 25.
• port: - port to run server on. Defaults to 9090.

Subscriber objects may be attached. The following events are published:

• rpc_incoming - published everytime the server receives an RPC
• rpc_ok - Everything when according to plan
• rpc_exception - Handler raised an exception defined in the protocol
• rpc_error - Handler raised an unexpected error (useful for error tracking)
• server_start - Start started
• server_connection_opened - Client TCP connection
• server_connection_closed - Client TCP disconnect
• thread_pool_server_pool_change - Thread pool grow/shinks

The listener should implement a method. A listener will only receive notifications if the appropriate method is implemented. Here's an example:

class Counter
	def initialize
		@counter = 0
	end

	def rpc_incoming(rpc)
		@counter += 1
	end
end

server = ThriftServer.threaded EchoService, Handler.new
server.subscribe Counter.new

ThriftServer includes three subscribers: two for metrics and one for logging. The ThriftServer::LogSubscriber uses a standard library logger to print useful information when any of the previously mentioned events happen. Attaching the subscriber is so important that it has its own helper method.

server.log Logger.new($stdout)
# Same thing as above, just much longer
server.subscribe ThriftServer::LogSubscriber.new(Logger.new($stdout))

The remaining two middlware handle metrics. Each handles a different metric granularity. ThriftServer::ServerMetricsSubcriber does stats for all RPCs & server things. ThriftServer::RpcMetricsSubscriber gives metrics for each individual RPC. Naturally there are important subscribers and its highly recommend you add them. There is a shortcut method for adding them both. They require a statsd instance. You can customize statsd prefix & postfix on that instance.

server.metrics Statsd.new
# Same thing as above, just much longer
server.subscribe ThriftServer::ServerMetricsSubscriber.new(Statsd.new)
server.subscribe ThriftServer::RpcMetricsSubscriber.new(Statsd.new)

ThriftServer::ServerMetricsSubscriber instruments the following keys:

• rpc.latency - (timer)
• rpc.incoming - (counter)
• rpc.success - (counter) - Everything A-OK!
• rpc.exception - (counter) - Result was defined protocol exception
• rpc.error - (counter) - Uncaught errors
• server.pool.size - (guage) - Number of available threads
• server.connection.active - (guage) - Threads with active TCP connections

ThriftServer::RpcMetricsSubscriber produces the same metrics, but at an individual RPC level. Assume the RPC is named foo. keys:

• rpc.foo.latency (timer)
• rpc.foo.incoming (counter)
• rpc.foo.success (counter)
• rpc.foo.exception (counter)
• rpc.foo.exception.xxx (counter) - where xxx is listed in throws in the Thrift IDL.
• rpc.foo.error (counter)

The library uses a middleware approach to implement preprocessing of each RPC before handing it off the original handler. You can implement your own middleware easily. The middleware must respond to call and accept a ThriftServer::RPC object. ThriftServer::RPC is a simple struct with two members: name and args. Here's an example that dumps the args to stdout.

class ExampleMiddleware
	include Concord.new(:app)

	def call(rpc)
		puts rpc.args.inspect
		app.call rpc
	end
end

New middleware can be added at build time or afterwards.

ThriftServer.threaded processor, handler, options do |stack|
	stack.use ExampleMiddlware
end

Middleware can also be added after the server is built

server = ThriftServer.threaded processor, handler, options
server.use ExampleMiddleware

server.start # start it!

ThriftServer used metaprogramming and delegating to implement the funcationality. build used the processor class to determine which methods are required by the protocol. A middleware stack is created using the object passed in the options hash. A delegate class is created to wrap each method call to pass through the stack before calling the defined handler. A subclass of the processor is defined to wrap the handler provided in initialize in the delegate class.

$ vagrant up
$ vagrant ssh
$ cd /vagrant
$ make test-ci

1. Create your feature branch (git checkout -b my-new-feature)
2. Commit your changes (git commit -am 'Add some feature')
3. Push to the branch (git push origin my-new-feature)
4. Create a new Pull Request