Camel Cassandra Component

Available as of Camel 2.15

Apache Cassandra is an open source NoSQL database designed to handle large amounts on commodity hardware. Like Amazon's DynamoDB, Cassandra has a peer-to-peer and master-less architecture to avoid single point of failure and garanty high availability. Like Google's BigTable, Cassandra data is structured using column families which can be accessed through the Thrift RPC API or a SQL-like API called CQL.

This component aims at integrating Cassandra 2.0+ using the CQL3 API (not the Thrift API). It's based on Cassandra Java Driver provided by DataStax.

Maven users will need to add the following dependency to their pom.xml:

<dependency>
    <groupId>org.apache.camel</groupId>
    <artifactId>camel-cassandraql</artifactId>
    <version>x.y.z</version>
    <!-- use the same version as your Camel core version -->
</dependency>

URI format

The endpoint can initiate the Cassandra connection or use an existing one.

To fine tune the Cassandra connection (SSL options, pooling options, load balancing policy, retry policy, reconnection policy...), create your own Cluster instance and give it to the Camel endpoint.

Endpoint Options

Messages

Incoming Message

The Camel Cassandra endpoint expects a bunch of simple objects (Object or Object[] or Collection<Object>) which will be bound to the CQL statement as query parameters. If message body is null or empty, then  CQL query will be executed without binding parameters.

Headers:

• CamelCqlQuery (optional, String or RegularStatement): CQL query either as a plain String or built using the QueryBuilder.

Outgoing Message

The Camel Cassandra endpoint produces one or many a Cassandra Row objects depending on the resultSetConversionStrategy:

• List<Row> if resultSetConversionStrategy is ALL or LIMIT_[0-9]+
• Single Row if resultSetConversionStrategy is ONE
• Anything else, if resultSetConversionStrategy is a custom implementation of the ResultSetConversionStrategy

Repositories

Cassandra can be used to store message keys or messages for the idempotent and aggregation EIP.

Cassandra might not be the best tool for queuing use cases yet, read Cassandra anti-patterns queues and queue like datasets. It's advised to use LeveledCompaction and a small GC grace setting for these tables to allow tombstoned rows to be removed quickly.

Idempotent repository

The NamedCassandraIdempotentRepository stores messages keys in a Cassandra table like this:

CREATE TABLE CAMEL_IDEMPOTENT (
  NAME varchar,   -- Repository name
  KEY varchar,    -- Message key
  PRIMARY KEY (NAME, KEY)
) WITH compaction = {'class':'LeveledCompactionStrategy'}
  AND gc_grace_seconds = 86400;

This repository implementation uses lightweight transactions (also known as Compare and Set) and requires Cassandra 2.0.7+.

Alternatively, the CassandraIdempotentRepository does not have a NAME column and can be extended to use a different data model.

Aggregation repository

The NamedCassandraAggregationRepository stores exchanges by correlation key in a Cassandra table like this:

CREATE TABLE CAMEL_AGGREGATION (
  NAME varchar,        -- Repository name
  KEY varchar,         -- Correlation id
  EXCHANGE_ID varchar, -- Exchange id
  EXCHANGE blob,       -- Serialized exchange
  PRIMARY KEY (NAME, KEY)
) WITH compaction = {'class':'LeveledCompactionStrategy'}
  AND gc_grace_seconds = 86400;

Alternatively, the CassandraAggregationRepository does not have a NAME column and can be extended to use a different data model.