HTTP Request Timeout

问题:

In Python 2.6, a new "timeout" parameter was added to the httplib.HTTPConnection class: http://docs.python.org/library/httplib.html#httplib.HTTPConnection

However, this is only a timeout for the connection to a server. I am looking to set a timeout value for the request, not the connection. This doesn't seem to be supported by httplib.

Is there any way to emulate this behavior?

回答1:

You can set a global socket timeout ^(*):

import socket

timeout = 10
socket.setdefaulttimeout(timeout)

---

^(*) EDIT: As people in the comments correctly point out: This is technically true, but it only has predictable results for tasks that involve a single socket operation. An HTTP request consists of multiple socket operations (e.g. DNS requests or other things that might be abstracted away from an HTTP client). The timeout of the overall operation becomes unpredictable because of that.

回答2:

No, there isn't.

It's because the HTTP spec does not provide anything for the client to specify time-to-live information with a HTTP request. You can do this only on TCP level, as you mentioned.

On the other hand, the server may inform the client about timeout situations with HTTP status codes 408 Request Timeout resp. 504 Gateway Timeout.

回答3:

You could also use settimeout on the socket of the connection (works with Python 2.5):

connection = HTTPConnection('slow.service.com')
connection.request(...)
connection.sock.settimeout(5.0)
response = connection.getresponse()
response.read()
connection.close()

If the server cannot send the response within 5 seconds, a socket.error will be raised.

回答4:

When you initiate the connection, it is posible to give as third argument the timout value.

connection = HTTPConnection(<IP of URL>, <port or None>, <timeout>)
...

is a float in seconds.

connection = HTTPConnection('slow.service.com', None, 20.0)