Altinity Sink Connector for ClickHouse

Sink connector is used to transfer data from Kafka to Clickhouse using the Kafka connect framework. The connector is tested with the following converters

• JsonConverter
• AvroConverter (Using Apicurio Schema Registry or Confluent Schema Registry)

Features

• Inserts, Updates and Deletes using ReplacingMergeTree - Updates/Deletes
• Auto create tables in ClickHouse
• Exactly once semantics
• Bulk insert to Clickhouse.
• Store Kafka metadata Kafka Metadata
• Kafka topic to ClickHouse table mapping, use case where MySQL table can be mapped to a different CH table name.
• Store raw data in JSON(For Auditing purposes)
• Monitoring(Using Grafana/Prometheus) Dashboard to monitor lag.
• Kafka Offset management in ClickHouse
• Increased Parallelism(Customize thread pool for JDBC connections)

Source Databases

• MySQL (Debezium) Note:GTID Enabled - Highly encouraged for Updates/Deletes Refer enabling Gtid in Replica for non-GTID sources - https://www.percona.com/blog/useful-gtid-feature-for-migrating-to-mysql-gtid-replication-assign_gtids_to_anonymous_transactions/
• PostgreSQL (Debezium)

Component	Version(Tested)
Redpanda	22.1.3, 22.3.9
Kafka-connect	1.9.5.Final
Debezium	2.1.0.Alpha1
MySQL	8.0
ClickHouse	22.9, 22.10
PostgreSQL	15

Quick Start (Docker-compose)

Docker image for Sink connector (Updated December 12, 2022) altinity/clickhouse-sink-connector:latest https://hub.docker.com/r/altinity/clickhouse-sink-connector

Recommended Memory limits.

Production Usage In docker-compose.yml file, its recommended to set Xmx to atleast 5G -Xmx5G when using in Production and if you encounter a Out of memory/Heap exception error. for both Debezium and Sink

- KAFKA_HEAP_OPTS=-Xms2G -Xmx5G

Kubernetes

Docker Image for Sink connector(with Strimzi) https://hub.docker.com/repository/docker/subkanthi/clickhouse-kafka-sink-connector-strimzi

Docker Image for Debezium MySQL connector(with Strimzi) https://hub.docker.com/repository/docker/subkanthi/debezium-mysql-source-connector

Recommended to atleast set 5Gi as memory limits to run on kubernetes using strimzi.

      limits:
        memory: 6Gi
      requests:
        memory: 6Gi

MySQL:

cd deploy/docker
./start-docker-compose.sh 

PostgreSQL:

export SINK_VERSION=latest
cd deploy/docker
docker-compose -f docker-compose.yaml -f docker-compose-postgresql.override.yaml up

For Detailed setup instructions - Setup

Development:

Requirements

• Java JDK 11 (https://openjdk.java.net/projects/jdk/11/)
• Maven (mvn) (https://maven.apache.org/download.cgi)
• Docker and Docker-compose

mvn install -DskipTests=true

Data Types

Note: Using float data types are highly discouraged, because of the behaviour in ClickHouse with handing precision.(Decimal is a better choice)

MySQL	Kafka Connect	ClickHouse
Bigint	INT64_SCHEMA	Int64
Bigint Unsigned	INT64_SCHEMA	UInt64
Blob		String + hex
Char	String	String / LowCardinality(String)
Date	Schema: INT64 Name: debezium.Date	Date(6)
DateTime(6)	Schema: INT64 Name: debezium.Timestamp	DateTime64(6)
Decimal(30,12)	Schema: Bytes Name: kafka.connect.data.Decimal	Decimal(30,12)
Double		Float64
Int	INT32	Int32
Int Unsigned	INT64	UInt32
Longblob		String + hex
Mediumblob		String + hex
Mediumint	INT32	Int32
Mediumint Unsigned	INT32	UInt32
Smallint	INT16	Int16
Smallint Unsigned	INT32	UInt16
Text	String	String
Time		String
Time(6)		String
Timestamp		DateTime64
Tinyint	INT16	Int8
Tinyint Unsigned	INT16	UInt8
varbinary(*)		String + hex
varchar(*)		String
JSON		String
BYTES	BYTES, io.debezium.bits	String
YEAR	INT32	INT32
GEOMETRY	Binary of WKB	String

Sink Connector Configuration

Property	Default	Description
tasks.max	No	SinkConnector task(essentially threads), ideally this needs to be the same as the Kafka partitions.
topics.regex	No	Regex of matching topics. Example: "SERVER5432.test.(.*)" matches SERVER5432.test.employees and SERVER5432.test.products
topics	No	The list of topics. topics or topics.regex has to be provided.
clickhouse.server.url		ClickHouse Server URL
clickhouse.server.user		ClickHouse Server username
clickhouse.server.pass		ClickHouse Server password
clickhouse.server.database		ClickHouse Database name
clickhouse.server.port	8123	ClickHouse Server port
clickhouse.topic2table.map	No	Map of Kafka topics to table names, <topic_name1>:<table_name1>,<topic_name2>:<table_name2> This variable will override the default mapping of topics to table names.
store.kafka.metadata	false	If set to true, kafka metadata columns will be added to Clickhouse
store.raw.data	false	If set to true, the entire row is converted to JSON and stored in the column defined by the store.raw.data.column field
store.raw.data.column	No	Clickhouse table column to store the raw data in JSON form(String Clickhouse DataType)
metrics.enable	true	Enable Prometheus scraping
metrics.port	8084	Metrics port
buffer.flush.time.ms	30	Buffer(Batch of records) flush time in milliseconds
thread.pool.size	10	Number of threads that is used to connect to ClickHouse
auto.create.tables	false	Sink connector will create tables in ClickHouse (If it does not exist)
snowflake.id	true	Uses SnowFlake ID(Timestamp + GTID) as the version column for ReplacingMergeTree
replacingmergetree.delete.column	"sign"	Column used as the sign column for ReplacingMergeTree.

ClickHouse Loader(Load Data from MySQL to CH for Initial Load)

Clickhouse Loader is a program that loads data dumped in MySQL into a CH database compatible the sink connector (ReplacingMergeTree with virtual columns _version and _sign)

Grafana Dashboard

Documentation

• Architecture
• Local Setup - Docker Compose
• Debezium Setup
• Kubernetes Setup
• Sink Configuration
• Testing
• Performance Benchmarking
• Confluent Schema Registry(REST API)

Blog articles

• ClickHouse as an analytic extension for MySQL
• Altinity Sink connector for ClickHouse