Persistent MPDClient

[Qrtn]

I'm writing a web-based front end to MPD. I instantiate and connect MPDClient when my program starts. At first calling commands works fine, but after some time of inactivity,

  File "/home/ztang/music/music.py", line 19, in queue
    client.add(path)
  File "/home/ztang/music/venv/lib/python3.3/site-packages/mpd.py", line 588, in decorator
    return wrapper(self, name, args, bound_decorator(self, returnValue))
  File "/home/ztang/music/venv/lib/python3.3/site-packages/mpd.py", line 229, in _execute
    return retval()
  File "/home/ztang/music/venv/lib/python3.3/site-packages/mpd.py", line 583, in decorator
    return function(self, *args, **kwargs)
  File "/home/ztang/music/venv/lib/python3.3/site-packages/mpd.py", line 352, in _fetch_nothing
    line = self._read_line()
  File "/home/ztang/music/venv/lib/python3.3/site-packages/mpd.py", line 260, in _read_line
    raise ConnectionError("Connection lost while reading line")
mpd.ConnectionError: Connection lost while reading line

A further attempt to connect:

  File "/home/ztang/music/music.py", line 32, in queuedir
    client.add(path)
  File "/home/ztang/music/venv/lib/python3.3/site-packages/mpd.py", line 588, in decorator
    return wrapper(self, name, args, bound_decorator(self, returnValue))
  File "/home/ztang/music/venv/lib/python3.3/site-packages/mpd.py", line 227, in _execute
    self._write_command(command, args)
  File "/home/ztang/music/venv/lib/python3.3/site-packages/mpd.py", line 252, in _write_command
    self._write_line(" ".join(parts))
  File "/home/ztang/music/venv/lib/python3.3/site-packages/mpd.py", line 233, in _write_line
    self._wfile.write("%s\n" % line)
  File "/home/ztang/music/venv/lib/python3.3/site-packages/mpd.py", line 75, in _dummy
    raise ConnectionError("Not connected")
mpd.ConnectionError: Not connected

Is there a way to avoid the ConnectionError? Will I have to catch the exception and reconnect, or connect and disconnect every time I issue a command?

[msmucr]

I was thinking about similar issue in my script, which is using mpd module. It is matter of mdp daemon, which disconnect every client afrter some time of inactivity, default period there is one minute and could be configured in mpd.conf via connection_timeout variable. Mpd module throws that exception when trying to write to closed socket after that disconnection. Another invoke of connect method should be enough to issue next commands to mpd daemon.
I came to three ways to overcome this:

• set timeout to long period at server - not good solution.
• use ping method of MPDClient instance to keep connection alive (lets say spawn another thread in your Python application and run endless loop with ping and timeout shorter than default timeout) - also not so clean solution, because it doesn't cover other situations like restart of mpd daemon or other reasons for closed socket.
• handle that ConnectionError exception using try statement which wraps invoking of MPDClient methods

I've used last way and individually handled that exceptions in my python script, but thinking about future better solution for that.. create proxy object for MPDClient, which will reconnect automatically after throwing that exception.
Similarly like there: http://www.arngarden.com/2013/04/29/handling-mongodb-autoreconnect-exceptions-in-python-using-a-proxy/, but i haven't did that yet.

Generally i think, that including that auto reconnect functionality directly into python-mpd2 will be very useful.

[Qrtn]

On Tue, Dec 31, 2013 at 5:58 PM, msmucr notifications@github.com wrote:

I was thinking about similar issue in my script, which is using mpd
module. It is matter of mdp daemon, which disconnect every client afrter
some time of inactivity, default period there is one minute and could be
configured in mpd.conf via connection_timeout variable. Mpd module throws
that exception when trying to write to closed socket after that
disconnection. Another invoke of connect method should be enough to issue
next commands to mpd daemon.

Ah, so it's mpd that disconnects the client! I will probably handle the
ConnectionError as well. It is a little annoying having to try-except for
every call, so I agree, auto reconnect would be nice. Thanks msmucr.`

[Mic92]

In a different library I use/have committed to, the client tries to connect before every send, if no connection is available and returns an error in case of failure after this action. A different solution might be something like a ConnectionError handler, because library user might have different strategies how to recover from an error. Maybe it could even be a combination of both.

[msmucr]

Hello Jörg,

i've checked mpd module guts very briefly and thought little bit about its internal exception handling. Currently it is raised mostly at place, where module reads response from server and checking newline at its end. Simple reconnection in possible internal handler wouldn't be enough for recovery at that moment, it would be necessary to repeat last command after successful reconnection and parse response again. So it needed to store all connection details and last commands with arguments to some internal property of MPDClient instance. I haven't tried password protected server yet, but it will also maybe needs another send of password.
Variant, you've mentioned with preventive connection before any command will be without that command repeat, but maybe include some overhead.
And external wrapper (like proxy object) at level of application, which use library, wouldn't require any further internal modifications.
What do will be your preferred solution?

[dbrgn]

I'm having the same troubles here. We're moving to MPD with Orochi (see dbrgn/orochi#45) and handling this issue in the python-mpd2 library would be the better way in my opinion than to handle it all over our code with try-except blocks.

[Qrtn]

You could connect and disconnect every time you send a command.

from contextlib import contextmanager
import mpd

HOST, PORT = '127.0.0.1', 6600
client = mpd.MPDClient()

@contextmanager
def connection():
    try:
        client.connect(HOST, PORT)
        yield
    finally:
        client.close()
        client.disconnect()

def queue(path):
    with connection():
        try:
            client.add(path)
        except mpd.CommandError:
            return 'invalid path'

    return 'queued'

[dbrgn]

Ah, that would actually be a great workaround in my case. Still, support for persistent connections would be even better.

[Mic92]

As msmucr said, I could add a callback hander in _read_line(), which an application can overwrite. But just resending the last command can be dangerous, because it is unknown in which state the connection broke (it might be that the command was executed, but the connection broke during the response). This will not replace the try-catch statements around all mpd commands. But it will remove some boiler code, because the program does not need to recover from each mpd command. About the persistence connection issue: this could be solved by sending a ping command from time to time. The problem is, that this would be require some kind of event loop or timer, which is not available in every python script.

[ways]

I'm having the same problem using mopidy instead of mpd.

Traceback (most recent call last):
File "/root/RadiOS/RadiOS.py", line 442, in
while True:
File "/root/RadiOS/RadiOS.py", line 311, in Compare
str (nowPlaying) + " or ioChannel " + str (ioChannel[0]) )
File "/root/RadiOS/RadiOS.py", line 150, in StopMPD
c.clear ()
File "/usr/local/lib/python2.7/dist-packages/mpd.py", line 583, in decorator
return wrapper(self, name, args, bound_decorator(self, returnValue))
File "/usr/local/lib/python2.7/dist-packages/mpd.py", line 229, in _execute
return retval()
File "/usr/local/lib/python2.7/dist-packages/mpd.py", line 578, in decorator
return function(self, _args, *_kwargs)
File "/usr/local/lib/python2.7/dist-packages/mpd.py", line 352, in _fetch_nothing
line = self._read_line()
File "/usr/local/lib/python2.7/dist-packages/mpd.py", line 260, in _read_line
raise ConnectionError("Connection lost while reading line")
mpd.ConnectionError: Connection lost while reading line

[handsomegui]

So... it there any update & solution for this issue?

[ways]

@handsomegui My solution was to do a ping at a regular interval. See line 307 at https://github.com/ways/RadiOS/blob/master/RadiOS.py.

Works for me.