On Friday 11 March 2011 4:38:41 AM Sergey Beryozkin wrote: > Hi > > On Thu, Mar 10, 2011 at 10:51 PM, Keith Hawes <[email protected]> wrote: > > While you are in there (if you still are gong to look at it) there is > > huge synchronous block: > > > > protected synchronized void updateDests(HttpServletRequest request) ... > > > > that limits the throughput of a CXF application. We have a XCF app that > > is backed by a Casandra cluster and the number of transactions per > > second is limited by this block (according to a co worker of mine) > > To be honest I'd like to remove this method altogether, or rather, have an > actual address added as an HTTP request parameter and then copied later on > as the message property.It is just a bit tricky at this stage given that > the WSDL and the services list creation does depend on the overriding > process...I'll see if the synchronization scope can be minimized - thanks > for the hint
Actually, on trunk, the wsdl might not need it. The wsdl generation is now done from an interceptor on the normal chain. I'm going to try and get the javascript generation also into an interceptor. At that point, the only thing left would be the services list. I'd LIKE the services list to actually be written as an actual Destination/Endpoint or similar that could be registered on any URL or context. Right now, the Jetty transport doesn't have the ability to have a services list as it's basically baked into the Servlet stuff. If it was an actual endpoint, it could be registered anywhere (and possibly not on /services) or maybe not registered at all. Dan > Sergey > > > On Thu, Mar 10, 2011 at 10:08 AM, Sergey Beryozkin <[email protected]>wrote: > >> I'm going to investigate this issue further - I think we can improve > >> ServletController it was written a long time ago and supporting encoded > >> URIs was not really on the map at a time. > >> The other thing I'll check if it is possible to pass matrix parameters > >> without values, just > >> > >> ;slash > >> > >> as opposed to ;slash=y > >> > >> this has to work - just need to add a test and confirm > >> > >> Sergey > >> > >> On Thu, Mar 10, 2011 at 5:59 PM, Keith Hawes <[email protected]> wrote: > >>> I was thinking along the lines of the matrix parameter as a work around > >>> as well. > >>> > >>> On Thu, Mar 10, 2011 at 1:25 AM, Sergey Beryozkin <[email protected]>wrote: > >>>> Hi > >>>> > >>>> On Thu, Mar 10, 2011 at 5:32 AM, Keith Hawes <[email protected]> wrote: > >>>>> Got the source and added breakpoints. > >>>>> > >>>>> in ServletController.invoke: > >>>>> String address = request.getPathInfo() == null ? "" : > >>>>> request.getPathInfo(); > >>>>> puts a decoded path info into address AND strips the repeated / > >>>>> > >>>>> I can call with "632%2F%2F04810" or "632//04810" and in both cases > >>>>> address is "/v1/device/632/04810" > >>>>> > >>>>> Later in ServletController.getBaseURL we have > >>>>> String reqPrefix = request.getRequestURL().toString(); > >>>>> and reqPrefix gets set to: > >>>>> http://localhost/myservice/v1/device/632//04810 > >>>>> > >>>>> It appears tomcat wants to strip the "extra" / on us. :( I cannot > >>>>> find a way to stop it for doing so. > >>>> > >>>> Thanks for this analysis...So may be you can do some workaround ? > >>>> Always use a single '/' but if you need two of them then indicate it > >>>> via an additional matrix parameter, say > >>>> > >>>> 632/04810;optionId=1;slash=y > >>>> > >>>> which can be seen as equivalent to > >>>> 632//04810;optionId=1 > >>>> > >>>> Will that work for you ? > >>>> > >>>> In meantime I'll need to investigate if ServletController can be > >>>> updated for request.getPathInfo() call be avoided... > >>>> > >>>> thanks, Sergey > >>>> > >>>> On Wed, Mar 9, 2011 at 8:21 PM, Keith Hawes <[email protected]> wrote: > >>>>>> Thanks. I had not tried the @Encoded annotation, I'll give that a > >>>>>> try and then > >>>>>> > >>>>>> as for configuring tomcat those two parameters need to be in system > >>>>>> properties, I just add them to the command line with -D > >>>>>> like this: > >>>>>> > >>>>>> -Dorg.apache.tomcat.util.buf.UDecoder.ALLOW_ENCODED_SLASH=true > >>>>>> -Dorg.apache.catalina.connector.CoyoteAdapter.ALLOW_BACKSLASH=true > >>>>>> > >>>>>> Yes I do have the matrix parameters (they are optional criteria > >>>>>> which narrow the results) > >>>>>> > >>>>>> On Wed, Mar 9, 2011 at 10:20 AM, Sergey Beryozkin < > >>>>>> > >>>>>> [email protected]> wrote: > >>>>>>> I've done a simple test (with Jetty) and it works fine. > >>>>>>> > >>>>>>> Client: > >>>>>>> > >>>>>>> @Test > >>>>>>> > >>>>>>> public void testBookWithComplexEncoding() throws Exception { > >>>>>>> > >>>>>>> BookStore store = JAXRSClientFactory.create(" > >>>>>>> > >>>>>>> http://localhost:"; + PORT, BookStore.class); > >>>>>>> > >>>>>>> String complex1 = "70%2F70%5C70.v1.0"; > >>>>>>> Book book = store.getBookComplexEncoded(complex1, 3L); > >>>>>>> assertEquals(3L, book.getId()); > >>>>>>> assertEquals(complex1, book.getName()); > >>>>>>> > >>>>>>> String complex2 = "70%2F%2F70%5C70.v1.0"; > >>>>>>> book = store.getBookComplexEncoded(complex2, 3L); > >>>>>>> assertEquals(3L, book.getId()); > >>>>>>> assertEquals(complex2, book.getName()); > >>>>>>> > >>>>>>> } > >>>>>>> > >>>>>>> Relevant server code: > >>>>>>> > >>>>>>> @GET > >>>>>>> > >>>>>>> @Path("/books/encoded/{complex}") > >>>>>>> public Book getBookComplexEncoded(@Encoded > >>>>>>> @PathParam("complex") > >>>>>>> > >>>>>>> String complex, > >>>>>>> > >>>>>>> @MatrixParam("optionId") Long > >>>>>>> > >>>>>>> id) { > >>>>>>> > >>>>>>> return new Book(complex, id); > >>>>>>> > >>>>>>> } > >>>>>>> > >>>>>>> I'm just using @Encoded to capture the exact sequence in the > >>>>>>> original form... > >>>>>>> > >>>>>>> Can you please explain how to configure those two properties in > >>>>>>> Tomcat ? > >>>>>>> > >>>>>>> org.apache.tomcat.util.buf.UDecoder.ALLOW_ENCODED_SLASH=true > >>>>>>> org.apache.catalina.connector.CoyoteAdapter.ALLOW_BACKSLASH=true > >>>>>>> > >>>>>>> What would really help of you could download CXF source and put a > >>>>>>> breakpoint in ServletController and JAXRSInInterceptor > >>>>>>> > >>>>>>> It is possible there is a flaw somewhere exposed on Tomcat only > >>>>>>> > >>>>>>> thanks, Sergey > >>>>>>> > >>>>>>> > >>>>>>> On Wed, Mar 9, 2011 at 10:45 AM, Sergey Beryozkin < > >>>>>>> > >>>>>>> [email protected]> wrote: > >>>>>>>> Hi > >>>>>>>> > >>>>>>>> Forwarding to the users list... > >>>>>>>> > >>>>>>>> I do suspect it's the ServletController's issue to do with the way > >>>>>>>> the handles matrix parameters > >>>>>>>> > >>>>>>>> Do you actual request URIs having matrix params like this : > >>>>>>>> > >>>>>>>> /service/v1/device/70%2F%2F70%5C70.v1.0;optionId=1 > >>>>>>>> > >>>>>>>> ? > >>>>>>>> > >>>>>>>> thanks, Sergey > >>>>>>>> > >>>>>>>> On Tue, Mar 8, 2011 at 6:42 PM, kh <[email protected]> wrote: > >>>>>>>>> I get this message when I have two encodes /s in the path %2F%2F > >>>>>>>>> WARNING: .No root resource matching request path > >>>>>>>>> /service/v1/device/70%2F%2F70%5C70.v1.0 is found. > >>>>>>>>> However: > >>>>>>>>> /service/v1/device/70%2F70%5C70.v1.0 > >>>>>>>>> works fine. > >>>>>>>>> > >>>>>>>>> I'm using cxf 2.2.3 > >>>>>>>>> > >>>>>>>>> I have > >>>>>>>>> > >>>>>>>>> org.apache.tomcat.util.buf.UDecoder.ALLOW_ENCODED_SLASH=true > >>>>>>>>> org.apache.catalina.connector.CoyoteAdapter.ALLOW_BACKSLASH=tru > >>>>>>>>> e > >>>>>>>>> > >>>>>>>>> set. > >>>>>>>>> > >>>>>>>>> My annotations are: > >>>>>>>>> > >>>>>>>>> > >>>>>>>>> @Path("/v1/device") > >>>>>>>>> public class Device extends com.netflix.customer.eds.Device { > >>>>>>>>> > >>>>>>>>> @GET > >>>>>>>>> @Path("/{esn}") > >>>>>>>>> @Consumes("application/xml") > >>>>>>>>> @Produces("application/xml") > >>>>>>>>> /* for error cases only */ > >>>>>>>>> @Transactional(rollbackFor = {Throwable.class}) > >>>>>>>>> public Response get(@Context HttpServletRequest request, > >>>>>>>>> > >>>>>>>>> @PathParam("esn") String esn, > >>>>>>>>> @MatrixParam("optionId") Long optionId) > >>>>>>>>> > >>>>>>>>> I gt the same response weather or not I include the matrix > >>>>>>>>> parameter. > >>>>>>>>> > >>>>>>>>> Am I missing something? > >>>> > >>>> -- > >>>> Sergey Beryozkin > >>>> > >>>> Application Integration Division of Talend <http://www.talend.com> > >>>> http://sberyozkin.blogspot.com > >> > >> -- > >> Sergey Beryozkin > >> > >> Application Integration Division of Talend <http://www.talend.com> > >> http://sberyozkin.blogspot.com -- Daniel Kulp [email protected] http://dankulp.com/blog Talend - http://www.talend.com
