Are you still having this problem? I'd like to see some testcase to reproduce it; we can't rule out a deadlock problem in SI too soon.
Type: Posts; User: iwein; Keyword(s):
Are you still having this problem? I'd like to see some testcase to reproduce it; we can't rule out a deadlock problem in SI too soon.
The prevent-duplicates option uses AcceptOnceFileListFilter under the hood, which compares file names to what it has seen before (this is tough on memory if you process a lot of files). There are...
Yes, if you require caching during startup I'd suggest you use a queue channel. If you don't want a queue channel in the loop after startup, you can have a router that routes to a queue channel...
I was afraid you were going to challenge that :)
How about you just use an xml test context and in that context import the production context and create a bean like this:
<bean...
What I usually do in these white box cases is to use bean overriding to replace the jms adapter beans with stubs or mocks under control of the test case. This would give you a separate context...
That intentionally vague formulation in the javadoc is mine. I never did an exhaustive performance check for the RecursiveLeafOnlyDirectoryScanner, but if I say "vast number of files" I'm not talking...
It would be nice indeed to have a ZeroMQ template and maybe some syntactic sugar to do this (like we have for amqp). If you create the issue you have my vote :)
Good idea, can you create an improvement issue in our issue tracker for this and post a link back here? https://jira.springsource.org/browse/INT
The AcceptOnceFileListFilter was intentionally left simple to make simple things work, the scanning and filtering options were added to make complex things possible. From the view point of Spring...
The PassThroughMessageProcessor is related to the CorrelatingMessageBarrier which is a solution unrelated to your problem.
Maybe you could stub jms and http out of your config and write a testcase...
Maybe you can try the argument that if you can't run something from the command line you would not be able to start jboss either. The "firm standard" argument is less convincing than that one.
Yeah, I know I left the implementation as an exercise to the reader, it's a complex problem and I'm not solving it for you here, I assume you were not expecting that :D
You're right about needing...
I think that you're heading for some threading issues if you decide to shutdown the context at runtime, instead of preventing it from starting.
Why would auto-startup="#{conf.isActiveServer()}"...
Just remind me: what exactly is missing for cluster support? Can you provide some examples (preferably in the form of running code)?
Could you attach/link to your test project? I'm a bit lost figuring out exactly what you're doing.
There is also file locking support in SI 2.0. In general I don't recommend locking files, but it might be worth the experiment for your particular use case.
I've tried srinivas_vsk's option before. That works quite well. The mapping subelement is a little less work, but it means you'd still have to change a string in two places.
Interesting thought, but why not just make this an integration test and send a message through (part of) the system?
The chances of a bug in the framework code are rather small and if your logic...
DeliveryNinja,
It could be that you've been caught off guard by the rename of CompletionStrategy to ReleaseStrategy. Other than that aggregator should work fine with M7 (actually a lot better...
Possibly, but until you're sure I'd assume it's a bug. Try to reproduce the behavior from a test case and then log a bug against our issue tracker. Better safe than sorry.
At first I'd just write a java class that does the job and wire it with a service-activator in xml.
If you can distill that into a small test project, I'd like to see a bug report for that. I think you might be onto something.
Check out jmx for this. It works quite well in most cases.
No definitely not. It's a good idea to keep the number of threadpools small so you can keep track of the number of running threads. Sometimes you need to split the pool to be able avoid starvation,...
If you want errors to go to the errorChannel you need to do asynchronous hand-off somewhere up the chain.