Re: [Hawkular-dev] Removal of embedded Cassandra

I agree with Mazz about the tests: tests should be predictable and fail only when there's a failure on the *code*, so, having to install a database, or having pre-existing data should *not* be something that would break tests. At least for JPA modules, it's common to use an embedded database for the tests, while the real module uses datasources configured on the application server. This problem could be avoided by having the integration tests to add the embedded Cassandra WAR to the nest, binding it to a different port (to avoid conflicting with possible Cassandra nodes running on the machine). For dev builds: I'm not sure I like the idea of having embedded Cassandra for dev builds but not for release builds. IMO, we should try to not deviate too much, specially because we would catch problems only when it's "too late". I'd rather require developers to install Cassandra locally, on this scenario. - Juca.