DolphinBeat

Other languages: 中文

This is a high available server that pulls MySQL binlog, parses binlog and pushs incremental update data into different sinks.

The types of sink supported currently and officially are Kafka and Stdout.

Features:

• Supports MySQL and MariaDB.
• Supports GTID and not GTID.
• Supports MySQL failover: if using GTID, dolphinbeat can work smoothly even if MySQL failover.
• Supports MySQL DDL: dolphinbeat can parse DDL statement and replay DDL upon it's own schema data in memory.
• Supports breakpoint resume: dolphinbeat has persistent metadata, it can resume to work after crash recover.
• Supports standalone and election mode: if election enabled, dolphinbeat follower will take over dead leader.
• Supports filter rules base on database and table for each sink.
• Supports HTTP API to inspect dolphinbeat.
• Supports metrics in Prometheus style.

The types of sink are scalable, you can implement your own sink if need, but I recommend you to use Kafka sink and let business consumes data from Kafka.

Quick start

Prepare your MySQL source, trun on binlog with ROW format, and type following commands and you will see JSON printed by dolphinbeat's Stdout sink.

docker run -e MYSQL_ADDR='8.8.8.8:3306' -e MYSQL_USER='root' -e MYSQL_PASSWORD='xxx' bytewatch/dolphinbeat
{
  "header": {
    "server_id": 66693,
    "type": "rotate",
    "timestamp": 0,
    "log_pos": 0
  },
  "next_log_name": "mysql-bin.000008",
  "next_log_pos": 4
}
...
...

The docker image above is for MySQL with GTID and only with Stdout sink enabled.

If your source database is not GTID, please add -e GTID_ENABLED='false' arg. If your source database is MariaDB, please add -e FLAVOR='mariadb' arg.

If you want to have a deep test, type following commands and you will get a shell:

docker run -e MYSQL_ADDR='8.8.8.8:3306' -e MYSQL_USER='root' -e MYSQL_PASSWORD='xxx' sh

In this shell, you can modify configurations in /data directory, and then start dolphinbeat manually.

Configuration description is presented in Wiki.

Compile from source

Type following commands and you will get builded binary distribution at build/dolphinbeat directory:

go get github.com/bytewatch/dolphinbeat
make 

Documents

• Configuration
• HTTP API

Sink

Kafka

This is a sink used for production. Dolphinbeat write data encoded with Protobuf into Kafka and business consumes data from Kafka.

Business need use client library to decode data in Kafka message, do stream processing on the binlog stream.

The Protobuf protocol is presented in protocol.proto .

Kafka sink has following features:

• Strong-ordered delivery: business will receive events in the same order with MySQL binlog.
• Exactly-once delivery: client library can dedup duplicated message with same sequence number which may caused by producer retry or Kafka failover.
• Unlimited event size: dolphinbeat use fragments algorithm like IPV4 if the binlog event is bigger than Kafka's max message size.

A small example using client library is presented in kafka-consumer.

kafka-consumer is a command tool to decode data in Kafka message and print out with JSON.

Stdout

This is a sink used for demonstration. Dolphinbeat write data encoded with JSON to Stdout.

Stdout sink doesn't support breakpoint resume.

Special thanks

Thank siddontang for his popular and powerful go-mysql library!

License

Apache License 2.0