Skip to content
Go to file


Failed to load latest commit information.
Latest commit message
Commit time

KarelDB - A Relational Database Backed by Apache Kafka

Build Status Maven Javadoc

KarelDB is a fully-functional relational database backed by Apache Kafka.


Releases of KarelDB are deployed to Maven Central.


Server Mode

To run KarelDB, download a release, unpack it, and then modify config/ to point to an existing Kafka broker. Then run the following:

$ bin/kareldb-start config/

At a separate terminal, enter the following command to start up sqlline, a command-line utility for accessing JDBC databases.

$ bin/sqlline
sqlline version 1.9.0

sqlline> !connect jdbc:avatica:remote:url=http://localhost:8765 admin admin

sqlline> create table books (id int, name varchar, author varchar);
No rows affected (0.114 seconds)

sqlline> insert into books values (1, 'The Trial', 'Franz Kafka');
1 row affected (0.576 seconds)

sqlline> select * from books;
| ID |   NAME    |   AUTHOR    |
| 1  | The Trial | Franz Kafka |
1 row selected (0.133 seconds)

To access a KarelDB server from a remote application, use an Avatica JDBC client. A list of Avatica JDBC clients can be found here.

If multiple KarelDB servers are configured with the same cluster group ID (see Configuration), then they will form a cluster and one of them will be elected as leader, while the others will become followers (replicas). If a follower receives a request, it will be forwarded to the leader. If the leader fails, one of the followers will be elected as the new leader.

Embedded Mode

KarelDB can also be used in embedded mode. Here is an example:

Properties properties = new Properties();
properties.put("schemaFactory", "io.kareldb.schema.SchemaFactory");
properties.put("parserFactory", "org.apache.calcite.sql.parser.parserextension.ExtensionSqlParserImpl#FACTORY");
properties.put("schema.kind", "io.kareldb.kafka.KafkaSchema");
properties.put("schema.kafkacache.bootstrap.servers", bootstrapServers);
properties.put("", "/tmp");

try (Connection conn = DriverManager.getConnection("jdbc:kareldb:", properties);
     Statement s = conn.createStatement()) {
        s.execute("create table books (id int, name varchar, author varchar)");
        s.executeUpdate("insert into books values(1, 'The Trial', 'Franz Kafka')");
        ResultSet rs = s.executeQuery("select * from books");

ANSI SQL Support

KarelDB supports ANSI SQL, using Calcite.

When creating a table, the primary key constraint should be specified after the columns, like so:

CREATE TABLE customers 
    (id int, name varchar, constraint pk primary key (id));

If no primary key constraint is specified, the first column in the table will be designated as the primary key.

KarelDB extends Calcite's SQL grammar by adding support for ALTER TABLE commands.

    ALTER TABLE tableName columnAction [ , columnAction ]*
    ( ADD tableElement ) | ( DROP columnName )

KarelDB supports the following SQL types:

  • boolean
  • integer
  • bigint
  • real
  • double
  • varbinary
  • varchar
  • decimal
  • date
  • time
  • timestamp

Basic Configuration

KarelDB has a number of configuration properties that can be specified. When using KarelDB as an embedded database, these properties should be prefixed with schema. before passing them to the JDBC driver.

  • listeners - List of listener URLs that include the scheme, host, and port. Defaults to
  • - The group ID to be used for leader election. Defaults to kareldb.
  • leader.eligibility - Whether this node can participate in leader election. Defaults to true.
  • kafkacache.backing.cache - The backing cache for KCache, one of memory (default), bdbje, lmdb, mapdb, or rocksdb.
  • - The root directory for backing cache storage. Defaults to /tmp.
  • kafkacache.bootstrap.servers - A list of host and port pairs to use for establishing the initial connection to Kafka.
  • - The group ID to use for the internal consumers, which needs to be unique for each node. Defaults to kareldb-1.
  • kafkacache.topic.replication.factor - The replication factor for the internal topics created by KarelDB. Defaults to 3.
  • - The timeout for initialization of the Kafka cache, including creation of internal topics. Defaults to 300 seconds.
  • - The timeout for an operation on the Kafka cache. Defaults to 60 seconds.



To use HTTPS, first configure the listeners with an https prefix, then specify the following properties with the appropriate values.


When using the Avatica JDBC client, the truststore and truststore_password can be passed in the JDBC URL as specified here.

HTTP Authentication

KarelDB supports both HTTP Basic Authentication and HTTP Digest Authentication, as shown below:

authentication.method=BASIC  # or DIGEST
authentication.realm=KarelDb-Props  # as specified in JAAS file

In the above example, the JAAS file might look like

KarelDb-Props {
  org.eclipse.jetty.jaas.spi.PropertyFileLoginModule required

The ProperyFileLoginModule can be replaced with other implementations, such as LdapLoginModule or JDBCLoginModule.

When starting KarelDB, the path to the JAAS file must be set as a system property.

$ export
$ bin/kareldb-start config/

When using the Avatica JDBC client, the avatica_user and avatica_password can be passed in the JDBC URL as specified here.

Kafka Authentication

Authentication to a secure Kafka cluster is described here.

Implementation Notes

KarelDB stores table data in topics of the form {tableName}_{generation}. A different generation ID is used whenever a table is dropped and re-created.

KarelDB uses three topics to hold metadata:

  • _tables - A topic that holds the schemas for tables.
  • _commits - A topic that holds the list of committed transactions.
  • _timestamps - A topic that stores the maximum timestamp that the transaction manager is allowed to return to clients.

Database by Components

KarelDB is an example of a database built mostly by assembling pre-existing components. In particular, KarelDB uses the following:

See this blog for more on the design of KarelDB.

Future Enhancements

Possible future enhancements include support for secondary indices.

You can’t perform that action at this time.