I have a Tornado server that listens on both an address/port, and on a socket. I create the server roughly like so (stripped down heavily):
from tornado import httpserver
from tornado.netutil import bind_unix_socket
server = httpserver.HTTPserver(
request_callback=some_callback,
io_loop=some_loop,
)
unix_socket = bind_unix_socket("mysock.sock")
server.add_socket(unix_socket)
server.listen(address="some_host", port=1234)
I'd like to be able to differentiate when a request is received via the socket, ie something like:
curl -XGET --unix-socket mysock.sock http:/ping
As opposed to:
curl http://some_address:1234/ping
I looked at the HTTPServerRequest that Tornado uses when a request is received, but I'm not sure what the best way is to tell the difference between the two. I can look at remote_ip
to see if it comes from localhost, but I don't think that's ideal.
localhost
is an internet-domain interface with a well-known IP address. It's not the same as a unix-domain socket, which has no IP address.
A brief look at the source suggests that the remote_ip
attribute will contain '0.0.0.0'
for a connection received on the unix socket.
(The remote_ip
would presumably be '127.0.0.1'
for a connection received via a localhost connection.)