Windows socket messages stop arriving - ring any bells?

Hello list,

I've been developing a client/server application on Windows for the
last few years using LispWorks 5/6, and generally speaking it has all
gone swimmingly. However I have one issue that I wanted to run past
you all in the hope that someone might've seen it before.

I have written a basic publish & subscribe message bus where a server
process sends simple text messages to subscribers. The server is
written in Lisp, some clients are written in Lisp and some in Adobe
Flex (i.e. they run in the Flash Player). As I say it all works pretty
well, but after a (long) while the messages sent from the server are
no longer received by the clients (both Lisp and Flash). No errors are
seen at any point.

Since the problem can be made to go away most effectively by
restarting the server, that's where the finger of suspicion is
currently pointing. Because it is a fairly rare problem (particularly
on my development box) it's been very hard to nail down, but the end
of the project is approaching and something must be done.

My plan is to re-write the server (and possibly the client) socket
part in C++ using the Boost ASIO networking library, which I am
already using elsewhere in the project, to see whether the problem
goes away. If anyone can suggest any lines of enquiry I should pursue
first, I'd be very grateful.

Cheers,

John :^P
--
John Pallister
john@johnp.net
john@synchromesh.com

Re: Windows socket messages stop arriving - ring any bells?

John Pallister <john@synchromesh.com> writes:

> I have written a basic publish & subscribe message bus where a server
> process sends simple text messages to subscribers. The server is
> written in Lisp, some clients are written in Lisp and some in Adobe
> Flex (i.e. they run in the Flash Player). As I say it all works pretty
> well, but after a (long) while the messages sent from the server are
> no longer received by the clients (both Lisp and Flash). No errors are
> seen at any point.

Hello!

First: Are you 100% sure the messages are still being sent from the
server and that the server is well and alive?
Second: Are you using TCP connections for this? If so, could it be a
simple networking timeout problem?
-- 
  regards,
    Espen

Re: Windows socket messages stop arriving - ring any bells?

Unable to parse email body. Email id is 10864

Re: Windows socket messages stop arriving - ring any bells?

Hi all,

Thanks very much for your responses.

@Espen:

The pub/sub server process is logging messages that indicate it is
still running and should be sending the messages to the clients. The
testing with Wireshark on the client PC that we have done indicates
that the messages weren't arriving from the server.

We are using TCP connections for the pub/sub bus. Simultaneously there
is RTP/UDP traffic running between the clients and a different process
that is running on the same machine as the pub/sub server; the code
for this is C/C++ at both ends, and seems to work fine.

The two machines are on the same LAN (which is fairly lightly loaded,
AFAICT) so I wouldn't expect timeouts to be a problem.

@Martin:

I hadn't thought of running netstat (I didn't know it was available on
Windows), but I'll try this in future.

I had seen this problem occasionally on my development machine (when
that was a laptop running Windows; now I use VMware Fusion on a
Hackintosh, and I haven't seen it recur), so it's not just on the
production server, an IBM rackmount server (it's an intranet
application). I haven't checked whether the drivers are up-to-date, I
must admit; that's something else I could look at.

I'm not trying to suggest there's a problem with the LispWorks server
socket code on Windows, but I was hoping someone could suggest any
"gotchas" that I might be missing. The problem is that the
traditional(?) Lisp debugging technique of "adding more logging
statements to see what's going on" falls down when the problem is this
intermittent, so I'm at a loss as to how to debug it further.

Anyway, thanks again for the responses.

Cheers,

John :^P
--
John Pallister
john@johnp.net
john@synchromesh.com

On Fri, Feb 4, 2011 at 8:34 AM, Espen Vestre <ev@netfonds.no> wrote:
> John Pallister <john@synchromesh.com> writes:
>
>> I have written a basic publish & subscribe message bus where a server
>> process sends simple text messages to subscribers. The server is
>> written in Lisp, some clients are written in Lisp and some in Adobe
>> Flex (i.e. they run in the Flash Player). As I say it all works pretty
>> well, but after a (long) while the messages sent from the server are
>> no longer received by the clients (both Lisp and Flash). No errors are
>> seen at any point.
>
> Hello!
>
> First: Are you 100% sure the messages are still being sent from the
> server and that the server is well and alive?
> Second: Are you using TCP connections for this? If so, could it be a
> simple networking timeout problem?
> --
>  regards,
>    Espen
>
>