separate jvm is the only mechanism to 'shutdown' in a test scenario right now....and its unlikely to change in the short term so designing around forking is your best bet
cheers, jesse -- jesse mcconnell jesse.mcconn...@gmail.com On Tue, Jul 20, 2010 at 05:47, Bjorn Borud <bbo...@gmail.com> wrote: > Jonathan Ellis <jbel...@gmail.com> writes: > >> there's some support for this in 0.7 (see >> http://issues.apache.org/jira/browse/CASSANDRA-1018) but fundamentally >> it's not really designed to be started and stopped multiple times >> within the same process. > > I am currently struggling with some of the same issues (on 0.6.3). > > I think there are very valid reasons for being able to start and stop > Cassandra within the same JVM. in particular for various testing-related > scenarios, but also when one wants to experiment with embedding > Cassandra. (Having a single-node, embedded Cassandra can be great for > testing Cassandra-based applications without forcing "casual users" to > go through the ordeal of installing Cassandra). > > Right now I have an ungodly mess that attempts to embed Cassandra by > wrapping the CassandraDaemon, generating a directory hierarchy for it > and populating the storage-conf.xml by way of a templating system and > then fires up Cassandra. > It *sort of* works, except that I haven't found a way to shut Cassandra > down reliably. the stop() method doesn't really do anything useful in > this respect. (I'd prefer some way of being able to ask it to shut down > and then have some way of blocking until it is down). I think some of > the errors I am seeing (NPE at at > org.apache.cassandra.db.commitlog.CommitLog.recover(CommitLog.java:166)) > are probably due to some race condition because of this, but I am > guessing here. > > I'll have another stare at the source. > > -Bjørn > > >
