使用Spring Data Cassandra动态创建键空间，表和生成表

Question

Using Cassandra, I want to create keyspace and tables dynamically using Spring Boot application. I am using Java based configuration.

I have an entity annotated with @Table whose schema I want to be created before application starts up since it has fixed fields that are known beforehand.

However depending on the logged in user, I also want to create additional tables for those user dynamically and be able to insert entries to those tables.

Can somebody guide me to some resources that I can make use of or point me in right direction in how to go about solving these issues. Thanks a lot for help!

Answer 1

The easiest thing to do would be to add the Spring Boot Starter Data Cassandra dependency to your Spring Boot application, like so...

<dependency>
  <groupId>org.springframework.boot</groupId>
  <artifactId>spring-boot-starter-data-cassandra</artifactId>
  <version>1.3.5.RELEASE</version>
</dependency>

In addition, this will add the Spring Data Cassandra dependency to your application.

With Spring Data Cassandra , you can configure your application's Keyspace(s) using the CassandraClusterFactoryBean (or more precisely, the subclass... CassandraCqlClusterFactoryBean ) by calling the setKeyspaceCreations(:Set) method.

The KeyspaceActionSpecification class is pretty self-explanatory. You can even create one with the KeyspaceActionSpecificationFactoryBean , add it to a Set and then pass that to the setKeyspaceCreations(..) method on the CassandraClusterFactoryBean .

For generating the application's Tables, you essentially just need to annotate your application domain object(s) (entities) using the SD Cassandra @Table annotation, and make sure your domain objects/entities can be found on the application's CLASSPATH.

Specifically, you can have your application @Configuration class extend the SD Cassandra AbstractClusterConfiguration class. There, you will find the getEntityBasePackages():String[] method that you can override to provide the package locations containing your application domain object/entity classes, which SD Cassandra will then use to scan for @Table domain object/entities.

With your application @Table domain object/entities properly identified, you set the SD Cassandra SchemaAction to CREATE using the CassandraSessionFactoryBean method, setSchemaAction(:SchemaAction) . This will create Tables in your Keyspace for all domain object/entities found during the scan, providing you identified the proper Keyspace on your CassandraSessionFactoryBean appropriately.

Obviously, if your application creates/uses multiple Keyspaces, you will need to create a separate CassandraSessionFactoryBean for each Keyspace, with the entityBasePackages configuration property set appropriately for the entities that belong to a particular Keyspace, so that the associated Tables are created in that Keyspace.

Now...

For the "additional" Tables per user, that is quite a bit more complicated and tricky.

You might be able to leverage Spring Profiles here, however, profiles are generally only applied on startup. If a different user logs into an already running application, you need a way to supply additional @Configuration classes to the Spring ApplicationContext at runtime.

Your Spring Boot application could inject a reference to a AnnotationConfigApplicationContext , and then use it on a login event to programmatically register additional @Configuration classes based on the user who logged into the application. You need to follow your register(Class...) call(s) with an ApplicationContext.refresh() .

You also need to appropriately handle the situation where the Tables already exist.

This is not currently supported in SD Cassandra, but see DATACASS-219 for further details.

Technically, it would be far simpler to create all the possible Tables needed by the application for all users at runtime and use Cassandra's security settings to restrict individual user access by role and assigned permissions.

Another option might be just to create temporary Keyspaces and/or Tables as needed when a user logs in into the application, drop them when the user logs out.

Clearly, there are a lot of different choices here, and it boils down more to architectural decisions, tradeoffs and considerations then it does technical feasibility, so be careful.

Hope this helps.

Cheers!

Answer 2

Following spring configuration class creates keyspace and tables if they dont exist.

@Configuration
public class CassandraConfig extends AbstractCassandraConfiguration {
    private static final String KEYSPACE = "my_keyspace";
    private static final String USERNAME = "cassandra";
    private static final String PASSWORD = "cassandra";
    private static final String NODES = "127.0.0.1"; // comma seperated nodes


    @Bean
    @Override
    public CassandraCqlClusterFactoryBean cluster() {
        CassandraCqlClusterFactoryBean bean = new CassandraCqlClusterFactoryBean();
        bean.setKeyspaceCreations(getKeyspaceCreations());
        bean.setContactPoints(NODES);
        bean.setUsername(USERNAME);
        bean.setPassword(PASSWORD);
        return bean;
    }

    @Override
    public SchemaAction getSchemaAction() {
        return SchemaAction.CREATE_IF_NOT_EXISTS;
    }

    @Override
    protected String getKeyspaceName() {
        return KEYSPACE;
    }

    @Override
    public String[] getEntityBasePackages() {
        return new String[]{"com.panda"};
    }


    protected List<CreateKeyspaceSpecification> getKeyspaceCreations() {
        List<CreateKeyspaceSpecification> createKeyspaceSpecifications = new ArrayList<>();
        createKeyspaceSpecifications.add(getKeySpaceSpecification());
        return createKeyspaceSpecifications;
    }

    // Below method creates "my_keyspace" if it doesnt exist.
    private CreateKeyspaceSpecification getKeySpaceSpecification() {
        CreateKeyspaceSpecification pandaCoopKeyspace = new CreateKeyspaceSpecification();
        DataCenterReplication dcr = new DataCenterReplication("dc1", 3L);
        pandaCoopKeyspace.name(KEYSPACE);
        pandaCoopKeyspace.ifNotExists(true).createKeyspace().withNetworkReplication(dcr);
        return pandaCoopKeyspace;
    }

}

Answer 3

Using @Enes Altınkaya answer:

@Value("${cassandra.keyspace}")
private String keySpace;

@Override
protected List<CreateKeyspaceSpecification> getKeyspaceCreations() {
    return Arrays.asList(
            CreateKeyspaceSpecification.createKeyspace()
                    .name(keySpace)
                    .ifNotExists()
                    .withNetworkReplication(new DataCenterReplication("dc1", 3L)));
}

To define your varaibles use an application.properties or application.yml file:

cassandra:
  keyspace: yout_keyspace_name

Using config files instead of hardcoded strings you can publish your code on for example GitHub without publishing your passwords and entrypoints ( .gitignore files) which may be a security risk.

Answer 4

The following cassandra configuration will create a keyspace when it does not exist and also run the start-up script specified

@Configuration
@PropertySource(value = {"classpath:cassandra.properties"})
@EnableCassandraRepositories
public class CassandraConfig extends AbstractCassandraConfiguration {

  @Value("${cassandra.keyspace}")
  private String cassandraKeyspace;

  @Override
  protected List<CreateKeyspaceSpecification> getKeyspaceCreations() {
    return Collections.singletonList(CreateKeyspaceSpecification.createKeyspace(cassandraKeyspace)
                .ifNotExists()
                .with(KeyspaceOption.DURABLE_WRITES, true)
                .withSimpleReplication());
  }

  @Override
  protected List<String> getStartupScripts() {
    return Collections.singletonList("CREATE TABLE IF NOT EXISTS "+cassandraKeyspace+".test(id UUID PRIMARY KEY, greeting text, occurrence timestamp) WITH default_time_to_live = 600;");
  }

}

Answer 5

This answer is inspired by Viswanath's answer.

My cassandra.yml looks as follows:

spring:
  data:
    cassandra:
      cluster-name: Test Cluster
      keyspace-name: keyspace
      port: 9042
      contact-points:
        - 127.0.0.1

@Configuration
@PropertySource(value = { "classpath:cassandra.yml" })
@ConfigurationProperties("spring.data.cassandra")
@EnableCassandraRepositories(basePackages = "info.vishrantgupta.repository")
public class CassandraConfig extends AbstractCassandraConfiguration {

    @Value("${keyspacename}")
    protected String keyspaceName;

    @Override
    protected String getKeyspaceName() {
        return this.keyspaceName;
    }

    @Override
    protected List getKeyspaceCreations() {
        return Collections.singletonList(CreateKeyspaceSpecification
                .createKeyspace(keyspaceName).ifNotExists()
                .with(KeyspaceOption.DURABLE_WRITES, true)
                .withSimpleReplication());
    }

    @Override
    protected List getStartupScripts() {
        return Collections.singletonList("CREATE KEYSPACE IF NOT EXISTS "
                + keyspaceName + " WITH replication = {"
                + " 'class': 'SimpleStrategy', "
                + " 'replication_factor': '3' " + "};");

    }
}

You might have to customize @ConfigurationProperties("spring.data.cassandra") , if your configuration starts with cassandra in cassandra.yml file then use @ConfigurationProperties("cassandra")

Answer 6

对于表的创建，您可以在application.properties文件中使用它

spring.data.cassandra.schema-action=CREATE_IF_NOT_EXISTS