[ 
https://issues.apache.org/jira/browse/CXF-8826?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Magnus Holm updated CXF-8826:
-----------------------------
    Description: 
I'm encountering issues with large SOAP response payloads using the http-hc5 
{{AsyncHTTPConduit}}. It seems the stream-reader encounters an unexpected EOF 
while processing the stream. The issue is not reproducible with the http-hc 
{{AsyncHTTPConduit}}. Test-cases are linked below. 

I've also monitored the existing {{AsyncHTTPConduitTest}} with wireshark in 
both modules http-hc and http-hc5. While http-hc behaves as you'd expect, 
http-hc5 produces RST,ACK (reset) instead of the expected FIN,ACK, even when 
processing is seemingly successful. Attached a couple of screenshots to 
illustrate. 

Unfortunately I've not been able to pinpoint what the cause of the issue is. 
This has happened using three different backends, jetty, undertow and some 
ancient SOAP production system that I'm sure are running neither of those, so 
it seems the issue is at least on the client-side.

Testcases: https://github.com/magnho/cxf/pull/1

Fails with http-hc5, succeeds with http-hc.

See:
{{org.apache.cxf.transport.http.asyncclient.hc5.AsyncHTTPConduitLargeResponseTest}}
{{org.apache.cxf.transport.http.asyncclient.AsyncHTTPConduitLargeResponseTest}}

Please let me know if you need me to provide any further details. 

  was:
Unexpected EOF while processing large response streams with hc5 async-client

I'm encountering issues with large SOAP response payloads using the http-hc5 
AsyncHTTPConduit. It seems the stream-reader encounters an unexpected EOF while 
processing the stream. The issue is not reproducible with the http-hc 
AsyncHTTPConduit. Test-cases are linked below. 

I've also monitored the existing AsyncHTTPConduitTest with wireshark in both 
modules http-hc and http-hc5. While http-hc behaves as you'd expect, http-hc5 
produces RST,ACK (reset) instead of the expected FIN,ACK, even when processing 
is seemingly successful. Attached a couple of screenshots to illustrate. 

Unfortunately I've not been able to pinpoint what the cause of the issue is. 
This has happened using three different backends, jetty, undertow and some 
ancient SOAP production system that I'm sure are running neither of those, so 
it seems the issue is at least on the client-side.

Testcases: https://github.com/magnho/cxf/pull/1

Fails with http-hc5, succeeds with http-hc.

See:
{{org.apache.cxf.transport.http.asyncclient.hc5.AsyncHTTPConduitLargeResponseTest}}
{{org.apache.cxf.transport.http.asyncclient.AsyncHTTPConduitLargeResponseTest}}

Please let me know if you need me to provide any further details. 


> AsyncHTTPConduit (hc5) - Unexpected EOF during response processing
> ------------------------------------------------------------------
>
>                 Key: CXF-8826
>                 URL: https://issues.apache.org/jira/browse/CXF-8826
>             Project: CXF
>          Issue Type: Bug
>          Components: Transports
>    Affects Versions: 3.5.5
>            Reporter: Magnus Holm
>            Priority: Major
>         Attachments: http-hc-wireshark.png, http-hc5-wireshark.png
>
>
> I'm encountering issues with large SOAP response payloads using the http-hc5 
> {{AsyncHTTPConduit}}. It seems the stream-reader encounters an unexpected EOF 
> while processing the stream. The issue is not reproducible with the http-hc 
> {{AsyncHTTPConduit}}. Test-cases are linked below. 
> I've also monitored the existing {{AsyncHTTPConduitTest}} with wireshark in 
> both modules http-hc and http-hc5. While http-hc behaves as you'd expect, 
> http-hc5 produces RST,ACK (reset) instead of the expected FIN,ACK, even when 
> processing is seemingly successful. Attached a couple of screenshots to 
> illustrate. 
> Unfortunately I've not been able to pinpoint what the cause of the issue is. 
> This has happened using three different backends, jetty, undertow and some 
> ancient SOAP production system that I'm sure are running neither of those, so 
> it seems the issue is at least on the client-side.
> Testcases: https://github.com/magnho/cxf/pull/1
> Fails with http-hc5, succeeds with http-hc.
> See:
> {{org.apache.cxf.transport.http.asyncclient.hc5.AsyncHTTPConduitLargeResponseTest}}
> {{org.apache.cxf.transport.http.asyncclient.AsyncHTTPConduitLargeResponseTest}}
> Please let me know if you need me to provide any further details. 



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to