socket.timeout: timed out

[imperio-wxm]

Hi,how can I set timeout in thriftpy?

I whrite like:

with client_context(dd_thrift.DingService, '127.0.0.1', 9090, proto_factory=dd_factory,
    socket_timeout=5000,connect_timeout=5000) as c:

But error:

UserWarning: `timeout` deprecated, use `socket_timeout` and `connect_timeout` instead.
warnings.warn("`timeout` deprecated, use `socket_timeout` and "
..........
..........
File "/usr/local/lib/python2.7/dist-packages/thriftpy/transport/socket.py", line 108, in read
buff = self.sock.recv(sz)
socket.timeout: timed out

Thanks!

[MrKiven]

Of course you can.
It seems like your service down..

[imperio-wxm]

@MrKiven
I am not sure set timeout in my server or client or both?

[MrKiven]

client

[imperio-wxm]

@MrKiven

My server is fine,but client set socket_timeout=5000 or 5000+ does not work, it is also default 3000.

When I count 3s, time out error come out.

[MrKiven]

server side default timeout is 3s, @lxyu

[imperio-wxm]

@MrKiven
When I use param like “timeout”,works!, but I notice warning:

UserWarning: `timeout` deprecated, use `socket_timeout` and `connect_timeout` instead.

timeout really deprecated?

[MrKiven]

yes, will use socket_timeout and connect_timeout, but if you set timeout, it also works

what's your thriftpy version?

[imperio-wxm]

@MrKiven
thriftpy 0.3.8

[AllanDaemon]

I having the same problem. I ported my implementation of the Apache Python module. IT was working ok. I made a shell for it (stills using the Apache Python Module), that connects and send requests as client to the server and prints the response. But with Thriftpy after 3 seconds the client connects the connection crashes. I can create a new working connection though. Before the time out, it works fine.

ERROR:thriftpy.server:timed out
Traceback (most recent call last):
  File "/usr/local/lib/python2.7/dist-packages/thriftpy/server.py", line 95, in handle
    self.processor.process(iprot, oprot)
  File "/usr/local/lib/python2.7/dist-packages/thriftpy/thrift.py", line 257, in process
    api, seqid, result, call = self.process_in(iprot)
  File "/usr/local/lib/python2.7/dist-packages/thriftpy/thrift.py", line 212, in process_in
    api, type, seqid = iprot.read_message_begin()
  File "thriftpy/protocol/cybin/cybin.pyx", line 429, in cybin.TCyBinaryProtocol.read_message_begin (thriftpy/protocol/cybin/cybin.c:6312)
  File "thriftpy/protocol/cybin/cybin.pyx", line 60, in cybin.read_i32 (thriftpy/protocol/cybin/cybin.c:1531)
  File "thriftpy/transport/buffered/cybuffered.pyx", line 65, in thriftpy.transport.buffered.cybuffered.TCyBufferedTransport.c_read (thriftpy/transport/buffered/cybuffered.c:1872)
  File "thriftpy/transport/buffered/cybuffered.pyx", line 69, in thriftpy.transport.buffered.cybuffered.TCyBufferedTransport.read_trans (thriftpy/transport/buffered/cybuffered.c:1939)
  File "thriftpy/transport/cybase.pyx", line 61, in thriftpy.transport.cybase.TCyBuffer.read_trans (thriftpy/transport/cybase.c:1463)
  File "/usr/local/lib/python2.7/dist-packages/thriftpy/transport/socket.py", line 111, in read
    buff = self.sock.recv(sz)
timeout: timed out

I believe this is a bug. A there is no reason to a connection close after 3 seconds. Sometimes even network latency is longer than it. And for my example, the shell I made is supposed to receive commands from the user, at any time. Also, the canonical Apache implementation doesn't behave this way.

I installed Thriftpy last night with pip:
Successfully installed cython-0.24 ply-3.8 thriftpy-0.3.8

[AllanDaemon]

I changed the default client_timeout in thriftpy/transport/socket.py from 3000 to None and now my server is working as expected, just like Apache implementation.

from:

class TServerSocket(object):
    """Socket implementation for server side."""

    def __init__(self, host=None, port=None, unix_socket=None,
                 socket_family=socket.AF_INET, client_timeout=3000,
                 backlog=128):

to:

class TServerSocket(object):
    """Socket implementation for server side."""

    def __init__(self, host=None, port=None, unix_socket=None,
                 socket_family=socket.AF_INET, client_timeout=None,
                 backlog=128):

I see I have the option to pass the timeout when creating a service manually. But the function make_server in thriftpy/rpc.py uses the default parameter for this. I didn't read all the source to figure out where to fix this. One option is passing explicitly client_timeout=None when creating the TServerSocket in make_server function. But It seems to me that the correct approach is setting the default timeout in the TServerSocket class initialization.

By now, the simpler way to use make_server with canonical behavior is this:

from thriftpy.rpc import make_server
s = make_server(Service, ServiceHandler(), host, port)
s.trans.client_timeout = None
(...)

[AllanDaemon]

These 2 commits fixed it to me.
AllanDaemon@6a4b118
AllanDaemon@0944247

I running tests and I will write some tests for it latter before I make pull request. Does this solutions sounds ok?

[lxyu]

Sorry for the late reply, the issue is because the client_timeout default set to 3000ms and the setting was not exposed in make_server helper function.

You may now pass the client_timeout=None to force the server side never timeout a client socket.

If it doesn't resolve your issue, feel free to reopen this thread.

[ghost]

Hello. I have thriftpy version 0.3.9 and collided with bug like @AllanDaemon report. As advised above, I replaced the value of client_timeout to None. And it worked. In this regard, I have a question. Maybe it makes sense to initially set the value to None? Thank you for attention.