Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Latest commit


Git stats


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


CLI tool for programmatic generation of Cassandra SSTables

sstable generator Instaclustr

This tool simply generates SSTables programmatically. It uses Cassandra’s CQLSSTableWriter. After the generation of SSTables is finished, you can load them by sstableloader tool as usual.

The project consists of these modules:

  • api—impl is coded against this module

  • impl—the implementation of your population logic, depends on api

  • generator—the implementation of whole generator CLI application

  • cassandra-3—the implementation of SSTable generator using internals of Cassandra 3 artifact

  • cassandra- 4—the implementation of SSTable generator using internals of Cassandra 4 artifact


mvn clean install (or mvn clean install -DskipTests)


Let’s guide you through an example. We want to generate a SSTable by Cassandra 3 API so we can load it to Cassandra afterwards. The components you need to have on a class path are as follows:

  • generator jar

  • cassandra-3 module jar

  • jar with the implementation of your generation logic

java \
  -cp /path/to/impl-1.0.jar:/path/to/generator-1.0.jar:/path/to/cassandra-3.jar \
  com.instaclustr.sstable.generator.LoaderApplication \
  _command_ \

The concrete example of the invocation would be:

java \
    -cp impl/target/sstable-generator-impl-1.0:generator/target/sstable-generator-1.0.jar:cassandra-3/target/cassandra-3-1.0.jar \
    com.instaclustr.sstable.generator.LoaderApplication \
    fixed \
    --keyspace mykeyspace \
    --table mytable \
    --output-dir=/tmp/output \
    --schema cassandra-3/src/test/resources/cassandra/cql/table.cql \
    --threads 2

Please be aware that you need to have all libraries of Apache Cassandra on the classpath as well. For that reason, please use ./ script and modify it to suit your needs in order to generate SStables.

No command executes default command—help:

Usage: <main class> [-V] COMMAND
  -V, --version   print version information and exit
  csv     tool for bulk-loading of data from csv
  random  tool for bulk-loading of random data
  fixed   tool for bulk-loading of fixed data

random Command

By this command, you are expected to provide data which represents a row in random fashion.

csv Command

csv command has same arguments as random but --file is mandatory. There is supposed to be CSV file which represents rows. Each row will be parsed into a list of strings passed to RowMapper implementation where you have to map them to list of objects for a Cassandra INSERT statement as values.

fixed Command

By fixed command, we will generate a SSTable by using the exact list of "rows" with columns. This will be obvious from the documentation which follows.

Row Generation

In order to generate data for all three cases above you have to implement interface com.instaclustr.cassandra.bulkloader.RowMapper in api module. This implementation should be placed in impl (or its equivalent) and it should be on a class path.

RowMapper Interface

public interface RowMapper {

     * Maps list of strings from whatever input representing
     * a row to list of objects to insert into Cassandra.
     * This method e.g. called upon generation from CSV file.
     * @param row where values are consisting of list of strings
     * @return list of objects to put to insert statement
    List<Object> map(final List<String> row);

     * Used when we do not want to generate data randomly but we have exact list of what to insert.
     * @return list of rows to be created containing list of cells
    Stream<List<Object>> get();

     * Logically same as {@link #map(List)} but all data per row
     * needs to be generated inside of the method. The number
     * of items in the returned list has to match number of columns
     * in a row. Each such object represents value which will be
     * passed to Cassandra INSERT statement.
     * <p>
     * This method is called repeatedly. Number of calls
     * is equal to paramter `--numberOfRecords`.
     * This method is called upon "random" generation.
     * @return list of objects to put to insert statement
    List<Object> random();

     * @return string representation of INSERT INTO statement. Question marks in VALUES are not
     * meant to be replaced.
     * <p>
     * For example: 'INSERT INTO keyspace.table (field1, field2, field3) VALUES (?, ?, ?)'
    String insertStatement();

The implementation of RowMapper you are supposed to place on the class path would look like this:

public class RowMapper1 implements RowMapper {

    public static final String KEYSPACE = "mykeyspace";
    public static final String TABLE = "mytable";

    public static final UUID UUID_1 = UUID.randomUUID();
    public static final UUID UUID_2 = UUID.randomUUID();
    public static final UUID UUID_3 = UUID.randomUUID();

    public List<Object> map(final List<String> row) {
        return null;

    public Stream<List<Object>> get() {
        return Stream.of(
            new ArrayList<Object>() {{
            new ArrayList<Object>() {{
            new ArrayList<Object>() {{

    public List<Object> random() {
        return null;

    public String insertStatement() {
        return format("INSERT INTO %s.%s (id, name, surname) VALUES (?, ?, ?);", KEYSPACE, TABLE);

SPI Mechanism

There is a Java SPI mechanism for implementation discovery, so it means that besides implementing API you have to change the impl/src/main/resources/META-INF/services/com.instaclustr.sstable.generator.RowMapper file containing FQCN of your implemenation on one line.

Once the impl jar is placed on the class path, it will be automatically discovered by the generator module so you do not need to use any command-line arguments. Merely putting that JAR on the class path does the job.

The same mechanism works for cassandra-3/4 jar. In case you want to generate jars by CQLSSTableWriter for Cassandra 3, just put that jar on the class path. If you want to generate "Cassandra 4 SSTables", place the respective cassandra-4.jar on the class path as shown above.

In practice this means that you need to compile only an impl module which contains one class so the compilation and JAR building will take literally a few seconds (less than 1 sec here). The command line arguments for all will look the same.

Further Information