Can this proxy be used to route the request to available server

[ajayinnovative2]

Can this proxy be used, when I have multiple tcp clients trying to send request via proxy to multiple servers of same type associated with this proxy. Is it capable to route the request to the next available server. And in response path, can this proxy identify to whom I should give the response back?

[abhinavsingh]

Sorry, but it's unclear what exactly is that you are after. Can you expand on it? Best way to find out is to try the plugins and see what works for you https://github.com/abhinavsingh/proxy.py/tree/develop/proxy/plugin.

[abhinavsingh]

Hi @ajayinnovative2 , I am re-reading your comment today and may be I can get a sense what you are after. Here is what I understand:

1. Multiple Clients <-> proxy.py <-> Multiple Upstreams
2. Part that's unclear to me is, whether the clients are pure TCP clients or are they HTTP clients?
3. If they are HTTP clients, you can easily achieve your task with proxy.py defaults. You'll have to likely write a custom plugin which can manage your app state e.g. client IP to upstream proxy server IP mapping.
4. If they are TCP clients, you can still achieve your task with proxy.py core. I can guide you through if you are interested. It will simply involve write a custom core plugin.

I am assuming your clients are HTTP clients and hence can use proxy settings. On that assumption marking this as the answer. Re-open if my assumptions are wrong.

[ajayinnovative2]

They are tcp clients, such as telnet. I have written my own proxy (load balancer) which does exactly what I wanted initially.
My proxy is capable enough to figure out the available server associated and route the request/response accordingly.
My only concern it becomes platform specific, as I am creating thread for each tcp session on request basis. And when it reaches to maximum thread allowed, new request will be rejected.

Still I am thinking to extend it to support millions of connections with seamless user experience.

[abhinavsingh]

That's awesome. Indeed scaling will be another aspect for any codebase. With proxy.py you can achieve it automagically. Before you dig into it, here is a recap to understand whats going on:

1. Internally proxy.py is a TCP/UDP server. This core component is responsible for listening on one or several port (or sock files). It doesn't know about what functionality will run on-top of a listening server. It's job is to accept connections, delegate them to a worker class for handling the connection.

2. By default, --work-klass flag points to proxy.http.HttpProtocolHandler class. This class knows how to handle an incoming HTTP client connection (server or proxy is inferred at runtime and handled appropriately)

3. For your implementation, you need a custom --work-klass. Which is capable of handling and tunneling packets at TCP level. Luckily you have a baked in starter template under examples directory. Here is a little background on that :)

   a) We already have a BaseTcpTunnelHandler. This is exactly what you need to build upon. See

   proxy.py/proxy/core/base/tcp_tunnel.py

   Line 26 in ee424ac

   class BaseTcpTunnelHandler(BaseTcpServerHandler[TcpClientConnection]):

   
   b) HttpsConnectTunnelHandler example is actually build on top of BaseTcpTunnelHandler. Do also check

   proxy.py/examples/https_connect_tunnel.py

   Line 21 in 35e9c57

   class HttpsConnectTunnelHandler(BaseTcpTunnelHandler):

   for reference usage and how to kick start your server instance

That's it :). You can customize your implementation as necessary while using proxy.py core. All the flags related to core --num-workers, --num-acceptors, --recv-buf-size etc still works. You'll only loose HTTP proxy related flags. Do -h or --help to see what flags your app instance contains.

Lastly, do also check skeleton app structure https://github.com/abhinavsingh/proxy.py/tree/develop/skeleton. You likely want to manage all this within your own repo / directory structure with custom proxy.py based code. You can pin or upgrade proxy.py via requirements.txt as and when necessary, while keeping your application code isolated from proxy.py codebase.

Let me know how it goes :)