On Thursday 17 August 2006 23:03, Davanum Srinivas wrote: > Are u using CommonsHTTPSender? Is it possible for you to try that? >
No we are not - any reason why using it may solve the issue? Are there any known problems with the default http implementation used by Axis? Thanks Manuel > -- dims > > On 8/17/06, Manuel Mall <[EMAIL PROTECTED]> wrote: > > I am supporting a multi-threaded server application which uses Axis > > as a web services client. We observe at what appear random and rare > > occasions (< less than once a week or less than 1 in 100000 > > requests) that thread 1 sends a request which is actually a > > duplicate of a request send by a thread 2 and is not the request it > > should actually send. All debugging / logging seems to indicate > > that Axis is the culprit but given that it is not reproducible in > > our test environments it is impossible to be certain about this. To > > be a bit more precise in describing what is happening: > > > > Assume the service we are calling is about retrieving an imaged > > document page by page, i.e. the service is something like > > getImage(docid, pageid). So we have thread 1 dealing with document > > A and thread 2 with document B. > > > > So thread 1 calls 'getImage(docid A, pageid X)' and thread 2 > > simultaneously calls 'getImage(docid B, pageid Y)'. However, both > > thread 1 and thread 2 send the same SOAP message requesting the > > same docid, pageid. Logging indicates that we call the Axis client > > stub with the correct parameters and logging also shows the wrong > > outgoing SOAP messages. > > > > We never observed such a mixup when different services are invoked > > simultaneously, i.e. thread 1 calls 'getImage(...)' and thread 2 > > calls 'getDocumentDetails(...)'. However, we did observe the same > > problem with simultaneous calls to services other than getImage. > > > > As the documents we are dealing with are A) business critical > > (related to property dealings like mortgages, titles, ...) and B) > > the customer pays for them returning the wrong data is actually > > very bad. > > > > It is also worthwhile mentioning that these problems only appeared > > after a deployment of new faster multi CPU hardware. The problem > > doesn't appear to be JRE specific, that is is has been observed > > under Java 1.4 and 1.5. > > > > We are close to throwing Axis out (as this is, if my recollections > > are correct, the 3rd multi-threading issue we have come across in > > Axis) but I am wondering if anyone, especially any experienced Axis > > committers, may have an idea where in the Axis codebase the problem > > could be? > > > > Thanks > > > > Manuel > > > > ------------------------------------------------------------------- > >-- To unsubscribe, e-mail: [EMAIL PROTECTED] For > > additional commands, e-mail: [EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
