Clemens Wyss wrote:

 

Unfortunately I haven't found an equally simple setup/installation example for ModeShape. For our needs ModeShape should make use of the jpa connector (preferably with a derby db/driver). The transient connector will not work out, because the evaluation/tests include many many nodes.

JcrConfiguration config = new JcrConfiguration().loadFrom("/path/to/configRepository.xml");

JcrEngine engine = config.build();

engine.start();

Repository repo = (Repository) engine.getRepository("repositoryNameFromConfigFile");

Session session = repository.login(  );  // Logs in with a guest session - works by default.

If you have any questions about how to set up the configRepository.xml, please feel free to post here, although there are several examples in the quickstarts and the documentation.

Unfortunately I haven't found an equally simple setup/installation example for ModeShape. For our needs ModeShape should make use of the jpa connector (preferably with a derby db/driver). The transient connector will not work out, because the evaluation/tests include many many nodes.

There's no scalability limit on our in-memory repository that I'm aware of.  In fact, if you're comparing ModeShape's performance to JR's transient repository, the in-memory repository is the nearest parallel.

Because we use Hibernate as our JPA implementation and Hibernate currently has some open defects around Derby support, you are better off using MySQL or PostgreSQL for an open source backing store for the JPA connector at this time.

General ModeShape question:

which ModeShape-setup/connector performs/scales best, if it is only used as a jcr container (i.e. replacement for jackrabbit), no federation etc (at least not yet)?

By default, the JCR container uses federation to share the jcr:system graph, so it's not an either/or decision.  I don't know that we've done any scalability tests.  If you do run some of these tests, could you share your results?

I've attached an example configuration file that we use for integration testing against JPA databases.  You would have to tweak this file for your own use, as described in http://docs.jboss.org/modeshape/latest/manuals/reference/html/jdbc-storage-connector.html.

We have built jars-with-dependencies for each of our connectors just like we have for our JCR implementation.  The one that you'd need for the JPA connector is at modeshape-2.3.0.Final-all-with-dependencies.jar.  Inside that JAR, you should find a JAR called modeshape-connector-store-jpa-2.3.0.Final-with-dependencies.jar.  Of course, you'd still need to add your JDBC driver JAR to your classpath.

Yeah, that configuration is used in some unit and/or integration tests. Here are some values when using MySQL:

dialect = org.hibernate.dialect.MySQL5InnoDBDialect

driverClassName = com.mysql.jdbc.Driver

url = jdbc:mysql://[host][,failoverhost...][:port]/[database][?propertyName1][=propertyValue1][&propertyName2][=propertyValue2]...

The rest of the properties are easily changed based upon your configuration and environment, but some general values might be:

maximumConnectionsInPool = 5

minimumConnectionsInPool = 1

numberOfConnectionsToAcquireAsNeeded = 1

maximumSizeOfStatementCache = 50

maximumConnectionIdleTimeInSeconds = 10

referentialIntegrityEnforced = true

largeValueSizeInBytes = 256

autoGenerateSchema = create

retryLimit = 3

compressData = true

Again, these are all explained in the first table in http://docs.jboss.org/modeshape/latest/manuals/reference/html/jdbc-storage-connector.html

This looks like there's a different (probably older) version of the Lucene Snowball JAR already on the classpath, perhaps provided by your application or another library (maybe Sling??). ModeShape currently uses 3.0.2 (the latest).