My real application is still the only place I can reproduce the problem.
While watching the client and server both run within Eclipse I can add one
more observation about the order of events when the exceptions occur.

My client is POSTing a message to a resource.  When the problem occurs the
client's call to
{ new Client(Protocol.HTTP).post(url, representation) }
does not return for ~30 seconds.  The my client application does not throw
an exception until after the server has thrown it's exceptions.

My server resource is receiving an acceptRepresentation call but is throwing
an exception because the Representation entity it receives has a problem of
some sort.  It's attempting to create a new Form(entity).  When it throws
it's first exception.  The form is only partially parsed because of a
java.io.IOException: Broken pipe.    The resource then attempts to use the
incomplete form object, throws a few more exceptions that are caught.

Watching it happen I'd say that the server exceptions have all occured
before the client throws it's Premature EOF exception.

Jerome's initial suggestion to my problem was to try different http
connectors.   I initially didn't realize how easy that was to do until I saw
an old  posting from Thierry refering to these links.
http://www.restlet.org/documentation/1.1/faq#04 &
http://www.restlet.org/documentation/1.1/faq#05
Just change the class path.  I didn't realize it was that easy. :-)

My first change was to switch my client to the Apache httpclient.  I'm going
to run my client over night and see if I have any errrors.
--
Timothy Aanerud


On Sun, Jul 5, 2009 at 6:27 AM, Jerome Louvel <jerome.lou...@noelios.com>wrote:

>  Hi Timothy,
>
>
>
> Thanks for updating us. Let’s keep an eye on it and if you find a way to
> reproduce it, please let us know!
>
>
>
> Best regards,
> Jerome Louvel
> --
> Restlet ~ Founder and Lead developer ~ http://www.restlet.org
> Noelios Technologies ~ Co-founder ~ http://www.noelios.com
>
>
>
>
>
>
>
> *De :* Timothy Aanerud [mailto:taane...@aticonsulting.com]
> *Envoyé :* mercredi 1 juillet 2009 20:53
> *À :* discuss@restlet.tigris.org
>
> *Objet :* Re: Premature EOF / Broken Pipe
>
>
>
> I tried adding a second resource to my sample application.  But it doesn't
> fail.
>
> In my actual application I have a scheduled background thread doing a POST
> transaction, immediately followed by a GET transaction. The Premature EOF
> exception has always occurred in the POST transaction.  Since order does not
> matter between these two message exchanges, I switched them around.  After
> doing this and trying several different scheduled time intervals for the
> background thread, I have yet to see it fail using the
> org.restlet.Client.Client using org.restlet.data.Protocol.HTTP or
> Protocol.HTTPS.
>
> I can't say "problem solved" but this is interesting and confusing at the
> same time.
> --
> Timothy Aanerud
>
> On Mon, Jun 29, 2009 at 9:18 AM, Timothy Aanerud
> <tannerudATaticonsulting.com> wrote:
>
> I haven't tried switching HTTP connectors. :-(
>
> But, I did build a sanitized/stripped down example.  Unfortunately, my
> example does not fail like my application does.   One difference between my
> example is I'm only exchange data with one resource, In my actual
> application I'm sending two messages to two resources back-to-back before
> pausing.
> --
> Timothy Aanerud
> taanerudataticonsulting.com <taane...@aticonsulting.com>
>
>
>
>  On Mon, Jun 29, 2009 at 2:57 AM, Jerome Louvel <jerome.lou...@noelios.com>
> wrote:
>
> Hi Timothy,
>
>
>
> Were you able to make progress on this front?
>
>
>
> Best regards,
> Jerome Louvel
> --
> Restlet ~ Founder and Lead developer ~ http://www.restlet.org
> Noelios Technologies ~ Co-founder ~ http://www.noelios.com
>
>
>
>
>
> *De :* Timothy Aanerud 
> [mailto:taanerudATaticonsulting.com<taane...@aticonsulting.com>]
>
> *Envoyé :* vendredi 19 juin 2009 22:05
>
>
> *À :* discuss@restlet.tigris.org
>
> *Objet :* Re: Premature EOF / Broken Pipe
>
>
>
> No, I haven't tried switching HTTP connectors. I get the same failures for
> both HTTP and HTTPS.
> In another experiment, I changed the client message frequency to ~1 second
> intervals and at this rate on both
> Windows XP and Fedora10/Linux show now problems with the server running on
> Fedora.
>
> The various frequencies and failure rates:
> 1 second == no problems
> 1.5 seconds == ~25% failure rate
> 5 seconds == ~25% failure rate
> 10 seconds == ~3% failure rate
> 180 seconds == >0.5%, if any failures
>
> I'll switch the HTTP connectors out one at a time and see what happens.
> --
> Timothy Aanerud
>
> On Fri, Jun 19, 2009 at 2:02 PM, Jerome Louvel <jerome.lou...@noelios.com>
> wrote:
>
> Hi Timothy,
>
> This looks like a bug to me. Have you tried with different Restlet HTTP
> connectors (such as Jetty on the server-side and Apache HTTP client)?
>
> If you could send us a simple standalone test case, we could easily debug
> what's going bad.
>
> Best regards,
> Jerome Louvel
> --
> Restlet ~ Founder and Lead developer ~ http://www.restlet.org
> Noelios Technologies ~ Co-founder~ http://www.noelios.com
>
>
> -----Message d'origine-----
> De: Timothy Aanerud 
> [mailto:taanerudATaticonsulting.com<taane...@aticonsulting.com>
> ]
>
> Envoy頺 vendredi 19 juin 2009 18:18
>
>
> : discuss@restlet.tigris.org
> Objet: RE: Premature EOF / Broken Pipe
>
>
> As a test, I moved the client code to a Windows XP machine. With a five
> second update rate it fails regularly too, with the same exceptions and
> stack traces.
>
> ------------------------------------------------------
>
> http://restlet.tigris.org/ds/viewMessage.do?dsForumId=4447&dsMessageId=23635
> 62
>
> ------------------------------------------------------
>
>
> http://restlet.tigris.org/ds/viewMessage.do?dsForumId=4447&dsMessageId=2363638
>
>
>
>
>
>
>

------------------------------------------------------
http://restlet.tigris.org/ds/viewMessage.do?dsForumId=4447&dsMessageId=2371513

Reply via email to