Marc Giger created CXF-4733:
-------------------------------
Summary: AsyncHTTPConduit threading issue
Key: CXF-4733
URL: https://issues.apache.org/jira/browse/CXF-4733
Project: CXF
Issue Type: Bug
Components: Transports
Affects Versions: 2.7.1
Reporter: Marc Giger
Hi,
I get the following exception
Caused by: com.ctc.wstx.exc.WstxEOFException: Unexpected EOF in prolog
at [row,col {unknown-source}]: [1,0]
at
com.ctc.wstx.sr.StreamScanner.throwUnexpectedEOF(StreamScanner.java:677)
at
com.ctc.wstx.sr.BasicStreamReader.handleEOF(BasicStreamReader.java:2116)
at
com.ctc.wstx.sr.BasicStreamReader.nextFromProlog(BasicStreamReader.java:2022)
at com.ctc.wstx.sr.BasicStreamReader.next(BasicStreamReader.java:1114)
at
com.ctc.wstx.sr.BasicStreamReader.nextTag(BasicStreamReader.java:1137)
at
org.apache.cxf.binding.soap.saaj.SAAJInInterceptor$SAAJPreInInterceptor.handleMessage(SAAJInInterceptor.java:125)
... 31 more
under following conditions:
- httpClientPolicy.setConnection(ConnectionType.CLOSE);
- massive parallel invocations
- SSL/TLS transport (from my analysis it seems to be independent if transport
security is active or not but the issue can only be reproduced with SSL enabled
and also then occurs just occasionally.)
>From my understanding the folllowing happens:
The AsyncHTTPConduit leases a pooled connection, invokes the endpoint, consumes
the response and returns the connection to the pool. Because the client
requested that the connection must be closed by the server
(ConnectionType.CLOSE) the released connection is closed. A new
AsyncHTTPConduit leases the closed connection from the pool. On first use of
this (closed) connection an exception will be thrown that the connection is
closed and a new one will be created (handled internally by hc). The
AsyncHTTPConduit will be informed about the exception with the registered
callback:
FutureCallback<Boolean> callback = new FutureCallback<Boolean>() {
public void completed(Boolean result) {
}
public void failed(Exception ex) {
setException(ex);
inbuf.shutdown();
outbuf.shutdown();
}
public void cancelled() {
inbuf.shutdown();
outbuf.shutdown();
}
};
Here, if an exception happend, inbuf.shutdown() is called which itself sets the
this.shutdown = true; flag in the SharedInputBuffer.
During the same time another thread? continues with a fresh connection and hits
the
public int read() throws IOException {
if (this.shutdown) {
return -1;
}
in the same SharedInputBuffer instance which returns -1.
This seems to be the cause for the above exception.
Just for testing purposes I replaced the statement
this.shutdown = true;
with
this.shutdown = false;
This change seems to "fix" the issue I have. I don't know how a proper fix
should look like so I can't provide a patch. Sorry.
Thanks,
Marc
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
