Thanks for the suggestion, but I've decided to go down the easy road and remove the Map from the response object.
Radek 2010/8/17 Daniel Kulp <[email protected]>: > > Is there any way this is reproducible with a small sample? Maybe tack an > "procuceMap" method onto hello world and hit it with a bunch of > threads/clients? I'm really not sure how to even start debugging such a > thing or what would be going on. > > If the map wasn't on the wire propery, I would have had a vague idea. > However, if it IS on the wire OK, not really a clue. :-( > > Dan > > > On Tuesday 17 August 2010 11:33:04 am Radosław Kozłowski wrote: >> Hello, >> >> we use a JAX-WS proxy (JaxWsProxyFactoryBean, CXF 2.2.3) to talk from >> app A (client) to app B (server) using SOAP. One of the methods called >> by A on B returns an object containing an object that contains, among >> other data members, a HashMap: >> >> HashMap<String, SubscriberServiceData> services; >> >> The server always initializes the services object and adds at least >> one service to it, so it's never null. However, sometimes services is >> null when received by the client. I discovered this while load testing >> the A app, and I can reproduce it most of the time, as it seems to >> fail (services == null) randomly. I sniffed the packets and enabled >> LoggingInInterceptor in the client to look at the raw XML, and the >> services map is always there (even if services becomes null in the >> client). So it looks as if the HashMap is not unmarshalled properly, >> but why only sometimes? >> >> I tested both apps with 2.2.9, but that didn't help. I'm out of ideas >> and was wondering if anyone could give any pointers. >> >> Thanks, >> Radek > > -- > Daniel Kulp > [email protected] > http://dankulp.com/blog >
