How to clean up socket connection to printer

Chris Angelico rosuav at gmail.com
Thu Jul 11 06:13:42 EDT 2013


On Thu, Jul 11, 2013 at 7:28 PM, loial <jldunn2000 at gmail.com> wrote:
> Replies to questions :
>
> 1. Does the printer accept connections again after some time?
>
> Yes, bit seems to vary how long that takes
>
> 2. Does the printer accept connections if you close and re-open the
> Python interpreter?
>
> Not after a Connection reset error. The script exits after trapping the "Connection reset by peer" error and it is only when a new instance of the script is kicked off that the "Connection refused" issue is encountered.
>

What's the network between you and the printer like? Are you, for
instance, going through an el cheapo NAT router? I've had some routers
that have a tendency to just randomly dump their NAT mappings (or so
it seems), which results in all manner of fun.

The "connection reset" error most likely means that you thought you
had a connection but the printer (or a router along the way, shouldn't
happen but see above) thought it didn't. You won't actually get the
error until you try to send some more data [1], which can result in
the variance in time. In fact, if the printer is indeed rebooting (as
per Ulrich's suggestion), you could get immediate success (reset,
reconnect, succeed), or you could get a delay of anything up to the
time it takes to reboot (if you "got in" straight after it went down).
What's the longest you've ever seen it take from conn reset to able to
connect again?

It's almost [2] certainly not that you're failing to properly close
the connection, though. I'd be inclined to simply code in a retry loop
(NOT instant, chances are you'll get some fast failures and you don't
want to spin; a progressive back-off delay is usually appropriate
here) and treat the failures as uninteresting.

ChrisA

[1] Unless you're using TCP keepalive, which you probably aren't.
[2] Insane stuff can be done, but hardware is presumed sane until
proven otherwise.



More information about the Python-list mailing list