Re: Hessian failure with "Proxy-Connection: keep-alive"

On May 26, 2006, at 9:51 AM, Alon J Salant wrote:

> We have found an issue with Hessian running through some proxy
> servers. The issue appears to be with using keep-alives to send
> chunked data.

Do you know if the problem is on the server or client side?

The Proxy-Connection should be a red-herring, since it should only
affect following requests (and keep-alive is the default value anyway).

My guess would be something about the chunked encoding is wrong.
Either the client or the proxy is producing bad chunked encoding, or
Resin isn't decoding it properly. In other words, there might be
something about the particular chunked encoding bytes that the Resin
HTTP parser isn't decoding properly.

So, assuming it's the server that's choking, getting a dump of the
exact bytes would help.

-- Scott
>
>
> Here are Hessian POST headers for a working call:
>
>
>
> HTTP: 1: POST /tops-services/remote/ContentManagerRemote
> HTTP/1.1
>
> HTTP: 2: Authorization: Basic NTcyMDA2L2NsZWFjaDpjbGVhY2gx
>
> HTTP: 3: Content-Type: text/xml
>
> HTTP: 4: User-Agent: Mozilla/4.0 (Windows 2000 5.0) Java/
> 1.5.0_06
>
> HTTP: 5: Host: tops.taylorpub.com
>
> HTTP: 6: Accept: text/html, image/gif, image/jpeg, *; q=.2,
> */*; q=.2
>
> HTTP: 7: Connection: keep-alive
>
> HTTP: 8: Transfer-Encoding: chunked
>
> HTTP: 9: Cookie: JSESSIONID=aDLoOoVXfLj9c3TXRT
>
>
>
> Here are Hessian POST headers for the failing call:
>
>
>
> HTTP: 1: POST http://tops.taylorpub.com/tops-services/
> remote/ContentManagerRemote HTTP/1.1
>
> HTTP: 2: Authorization: Basic NTcyMDA2L2NsZWFjaDpjbGVhY2gx
>
> HTTP: 3: Content-Type: text/xml
>
> HTTP: 4: User-Agent: Mozilla/4.0 (Windows 2000 5.0) Java/
> 1.5.0_06
>
> HTTP: 5: Host: tops.taylorpub.com
>
> HTTP: 6: Accept: text/html, image/gif, image/jpeg, *; q=.2,
> */*; q=.2
>
> HTTP: 7: Proxy-Connection: keep-alive
>
> HTTP: 8: Transfer-Encoding: chunked
>
> HTTP: 9: Cookie: JSESSIONID=a6ODB1JO8wDcjDtYRT
>
>
>
> The only significant difference appears to be the ‘Connection: keep-
> alive’ v. ‘Proxy-Connection: keep-alive’ headers.
>
>
>
> We used a network sniffer to inspect the packets being sent. In the
> working case, the following packets include chunked data. In the
> failing case, the server responds with a 400 Bad Request and the
> client throws an exception:
>
>
>
> Caused by: java.io.IOException: Error writing request body to server
>
> at sun.net.www.protocol.http.HttpURLConnection
> $StreamingOutputStream.checkError(Unknown Source)
>
> at sun.net.www.protocol.http.HttpURLConnection
> $StreamingOutputStream.write(Unknown Source)
>
> at com.caucho.hessian.io.HessianOutput.writeBytes
> (HessianOutput.java:664)
>
>
>
>
>
> Hessian on the server reports an exception:
>
>
>
> com.caucho.hessian.io.HessianProtocolException: addContent:
> expected end of call ('z') at end of stream.
>
> at com.caucho.hessian.io.HessianInput.error
> (HessianInput.java:1640)
>
> at com.caucho.hessian.io.HessianInput.completeCall
> (HessianInput.java:259)
>
> at com.caucho.hessian.server.HessianSkeleton.invoke
> (HessianSkeleton.java:152)
>
> at
> org.springframework.remoting.caucho.HessianServiceExporter.handleReque
> st(HessianServiceExporter.java:87)
>
> ...
>
>
>
> In the failing case it appears that Resin/Hessian think that the
> call is finished after the first chunk of the request is sent
> instead of waiting for the coming content as it does in the
> succeeding case.
>
>
>
> My conclusion is that Resin/Hessian are not handling the ‘Proxy-
> Connection: keep-alive’ header correctly.
>
>
>
> This is a fairly significant issue for us right now. I’d be happy
> to take some time to dig a bit deeper into the code if that would
> be helpful. Pointers on where to look would be appreciated. A fix
> would be appreciated even more J
>
>
>
> Alon
>
>
>
>
Received on Fri 26 May 2006 13:15:37 -0700