Question

I have written an OSS plugin to start/stop Derby during a Maven build process.
The plugin works fine in plain old single modules. However, if I have an aggregator of several modules and more than one of them has database-related tests, I seem to be hitting some weird problem.

I am invoking the plugin’s start and stop goals (respectively during the process-resources and test phases), as shown below:

        <plugin>
            <groupId>org.carlspring.maven</groupId>
            <artifactId>derby-maven-plugin</artifactId>
            <version>1.4</version>

            <configuration>
                <failIfAlreadyRunning>false</failIfAlreadyRunning>
            </configuration>

            <executions>
                <execution>
                    <id>start-derby</id>
                    <phase>process-resources</phase>
                    <goals>
                        <goal>start</goal>
                    </goals>
                </execution>
                <execution>
                    <id>stop-derby</id>
                    <phase>test</phase>
                    <goals>
                        <goal>stop</goal>
                    </goals>
                </execution>
            </executions>
        </plugin>

The problem consists in the fact that upon the second attempt to start the Derby (in-memory) server, Derby appears to still be running, or to have loaded with the database contents from the first module. I can tell this because the first module in the aggregator creates and populates some data in a table. My expectation is that once I’ve shutdown Derby and started it all over again from the other module, that it would be a fresh database without any existing contents.

Here is my code inside the plugin that deals with shutting down Derby:

try
{
    try
    {
        server.ping();
    }
    catch (Exception e)
    {
        if (failIfNotRunning)
        {
            throw new MojoExecutionException("Failed to stop the Derby server, no server running!", e);
        }

        getLog().error("Derby server was already stopped.");
        return;
    }

    server.shutdown();

    while (true)
    {
        Thread.sleep(1000);
        try
        {
            server.ping();
        }
        catch (Exception e)
        {
            getLog().info("Derby has stopped!");
            return;
        }
    }
}
catch (Exception e)
{
    throw new MojoExecutionException(e.getMessage(), e);
}

The full source of this rather simple plugin can be checked out or viewed in GitHub here.

Answer 1

I ended up doing the following:

try
{
    try
    {
        server.ping();
    }
    catch (Exception e)
    {
        if (failIfNotRunning)
        {
            throw new MojoExecutionException("Failed to stop the Derby server, no server running!", e);
        }

        getLog().error("Derby server was already stopped.");
        return;
    }

    try
    {
        DriverManager.getConnection(getConnectionURL());
        DriverManager.getConnection("jdbc:derby:;shutdown=true");
    }
    catch (SQLException e)
    {
        // Apparently this prints out a bunch of stuff we're not currently interested in,
        // we just want it to shutdown properly.
        // Perhaps further handling might be required at a future point in time.
    }

    server.shutdown();

    while (true)
    {
        Thread.sleep(1000);
        try
        {
            server.ping();
        }
        catch (Exception e)
        {
            getLog().info("Derby has stopped!");
            break;
        }
    }

    System.getProperties().remove("derby.system.home");
}
catch (Exception e)
{
    throw new MojoExecutionException(e.getMessage(), e);
}

This wasn’t the way I expected it had to be done.