Return-Path: Delivered-To: apmail-cxf-issues-archive@www.apache.org Received: (qmail 3495 invoked from network); 26 Sep 2010 21:31:57 -0000 Received: from unknown (HELO mail.apache.org) (140.211.11.3) by 140.211.11.9 with SMTP; 26 Sep 2010 21:31:57 -0000 Received: (qmail 34486 invoked by uid 500); 26 Sep 2010 21:31:57 -0000 Delivered-To: apmail-cxf-issues-archive@cxf.apache.org Received: (qmail 34391 invoked by uid 500); 26 Sep 2010 21:31:56 -0000 Mailing-List: contact issues-help@cxf.apache.org; run by ezmlm Precedence: bulk List-Help: List-Unsubscribe: List-Post: List-Id: Reply-To: dev@cxf.apache.org Delivered-To: mailing list issues@cxf.apache.org Received: (qmail 34383 invoked by uid 99); 26 Sep 2010 21:31:56 -0000 Received: from athena.apache.org (HELO athena.apache.org) (140.211.11.136) by apache.org (qpsmtpd/0.29) with ESMTP; Sun, 26 Sep 2010 21:31:56 +0000 X-ASF-Spam-Status: No, hits=-2000.0 required=10.0 tests=ALL_TRUSTED,NORMAL_HTTP_TO_IP,WEIRD_PORT X-Spam-Check-By: apache.org Received: from [140.211.11.22] (HELO thor.apache.org) (140.211.11.22) by apache.org (qpsmtpd/0.29) with ESMTP; Sun, 26 Sep 2010 21:31:55 +0000 Received: from thor (localhost [127.0.0.1]) by thor.apache.org (8.13.8+Sun/8.13.8) with ESMTP id o8QLVZKX018581 for ; Sun, 26 Sep 2010 21:31:35 GMT Message-ID: <32747172.409931285536695126.JavaMail.jira@thor> Date: Sun, 26 Sep 2010 17:31:35 -0400 (EDT) From: "Sergey Beryozkin (JIRA)" To: issues@cxf.apache.org Subject: [jira] Commented: (CXF-2997) CXF JAX-RS not thread safe when accessing multiple destinations concurrently In-Reply-To: <9671129.231771284653158565.JavaMail.jira@thor> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit X-JIRA-FingerPrint: 30527f35849b9dde25b450d4833f0394 [ https://issues.apache.org/jira/browse/CXF-2997?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12915044#action_12915044 ] Sergey Beryozkin commented on CXF-2997: --------------------------------------- This is the initial attempt to fix it : http://svn.apache.org/viewvc?rev=1001520&view=rev If one has many servlets 'binding' to the same destination then one has to set either "disable-address-updates" servlet parameter on each servlet to true or call setDisableAddressUpdates on CXFServlet/CXFNonSpringServlet. Perhaps, another and better solution is to introduce a thread local storage and associate it with Destination and completely disable the address overwrites but I thought that with the CXF 2.3 release coming close this change would be a bit unsafe - we'd just not know what we'd break. Any other ideas and comments are welcome. Please verify if this commit fixes the problem. We can get a proper fix once we evaluate what side-effects are possible... > CXF JAX-RS not thread safe when accessing multiple destinations concurrently > ----------------------------------------------------------------------------- > > Key: CXF-2997 > URL: https://issues.apache.org/jira/browse/CXF-2997 > Project: CXF > Issue Type: Bug > Components: JAX-RS, Transports > Affects Versions: 2.2.10 > Reporter: Ate Douma > Priority: Blocker > Fix For: 2.3, 2.2.11 > > Attachments: cxf-rest-test.tar.gz > > > If a (single) JAX-RS service is invoked concurrently for different destinations, the CXF ServletController and ServletTransportFactory implementations override the current Destination state between these invocations. > I have created a simple test web application using Spring to expose this problem. > My example ServiceImpl.java looks like: > @Path("/myservice/") > @Produces("application/xml") > public class ServiceImpl { > volatile boolean locked; > @GET > @Path("/uris") > @Produces("text/plain") > public String getUris(@Context UriInfo uriInfo) { > StringBuilder uris = new StringBuilder("BaseUri on entry: "+uriInfo.getBaseUri().toString()).append("\n"); > try { > while (locked) { > Thread.sleep(1000); > } > } > catch (Exception x) {} > return uris.append("BaseUri on exit : " + uriInfo.getBaseUri().toString()).append("\n").toString(); > } > > @GET > @Path("/lock") > @Produces("text/plain") > public String lock() { > locked = true; > return "locked"; > } > @GET > @Path("/unlock") > @Produces("text/plain") > public String unlock() { > locked = false; > return "unlocked"; > } > And in my web.xml I defined two CXFServlet mappings as follows: > > CXFServlet > /one/* > > > CXFServlet > /two/* > > Finally, in Spring ApplicationContext.xml I setup the jaxrs server like this: > > > > > > > > > > As can be seen from the ServicesImpl.java, I used a trick to temporarily "lock" and "unlock" a call to /myservice/uris from another request using /myservice/lock and /myservice/unlock. > Without locking, a call to http://localhost:8085/one/myservice/uris produces the expected following result: > BaseUri on entry: http://localhost:8085/one/ > BaseUri on exit : http://localhost:8085/two/ > However, if I first call http://localhost:8085/one/myservice/lock, then http://localhost:8085/one/myservice/uris (blocked), and finally http://127.0.0.1:8085/two/myservice/unlock (note the different hostname 127.0.0.1 and servletPath /two) I get the following result: > BaseUri on entry: http://localhost:8085/one/ > BaseUri on exit : http://127.0.0.1:8085/two/ > Clearly, UriInfo.getBaseURI() isn't thread safe as is shown above. > After debugging a bit the current (CXF 2.2.10) implementation of the ServletController and ServletTransportFactory, it looks like the ServletController.updateDests(HttpServletRequest) method and the ServletTransportFactory as a whole don't maintain resolved Destination state isolated per request. > Our CXF (JAX-RS) application exposes the same REST service to be called from many different addresses (hosts) as well as upfront unknown servlet paths, this really is a blocking issue. > To be more concrete, we have to use a request wrapper to dynamically "mount" the same rest service for different urls (its a front end for a unlimited hierarchical resources repository). > Is there an easy way to protect JAX-RS service interactions from concurrent invocations, e.g. by using special configurations, or is this indeed a serious bug which will have to be fixed first? -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.