Spark Binlog Library

A library for querying Binlog with Apache Spark structure streaming, for Spark SQL , DataFrames and MLSQL.

1. jianshu: How spark-binlog works
2. medium: How spark-binlog works

Requirements

This library requires Spark 2.4+ (tested). Some older versions of Spark may work too but they are not officially supported.

Linking

You can link against this library in your program at the following coordinates:

Scala 2.11

This is the latest stable versions.

MySQL Binlog:

groupId: tech.mlsql
artifactId: mysql-binlog_2.11
version: 1.0.4

HBase WAL:

groupId: tech.mlsql
artifactId: hbase-wal_2.11
version: 1.0.4

Limitation

1. mysql-binlog only support insert/update/delete events. The other events will ignore.
2. hbase-wal only support Put/Delete events. The other events will ignore.

MySQL Binlog Usage

The example should work with delta-plus

MLSQL Code:

set streamName="binlog";

load binlog.`` where 
host="127.0.0.1"
and port="3306"
and userName="xxxxx"
and password="xxxxx"
and databaseNamePattern="mlsql_console"
and tableNamePattern="script_file"
as table1;

save append table1  
as rate.`mysql_{db}.{table}` 
options mode="Append"
and idCols="id"
and duration="5"
and syncType="binlog"
and checkpointLocation="/tmp/cpl-binlog2";

DataFrame Code:

val spark = SparkSession.builder()
      .master("local[*]")
      .appName("Binlog2DeltaTest")
      .getOrCreate()

val df = spark.readStream.
  format("org.apache.spark.sql.mlsql.sources.MLSQLBinLogDataSource").
  option("host","127.0.0.1").
  option("port","3306").
  option("userName","root").
  option("password","123456").
  option("databaseNamePattern","test").
  option("tableNamePattern","mlsql_binlog").
  load()

val query = df.writeStream.
  format("org.apache.spark.sql.delta.sources.MLSQLDeltaDataSource").
  option("__path__","/tmp/datahouse/{db}/{table}").
  option("path","{db}/{table}").
  option("mode","Append").
  option("idCols","id").
  option("duration","3").
  option("syncType","binlog").
  option("checkpointLocation", "/tmp/cpl-binlog2").
  outputMode("append")
  .trigger(Trigger.ProcessingTime("3 seconds"))
  .start()

query.awaitTermination()

Before you run the streaming application, make sure you have fully sync the table

MLSQL Code:

connect jdbc where
 url="jdbc:mysql://127.0.0.1:3306/mlsql_console?characterEncoding=utf8&zeroDateTimeBehavior=convertToNull&tinyInt1isBit=false"
 and driver="com.mysql.jdbc.Driver"
 and user="xxxxx"
 and password="xxxx"
 as db_cool;
 
load jdbc.`db_cool.script_file`  as script_file;

run script_file as TableRepartition.`` where partitionNum="2" and partitionType="range" and partitionCols="id"
as rep_script_file;

save overwrite rep_script_file as delta.`mysql_mlsql_console.script_file` ;

load delta.`mysql_mlsql_console.script_file`  as output;

DataFrame Code:

import org.apache.spark.sql.SparkSession
val spark = SparkSession.builder()
  .master("local[*]")
  .appName("wow")
  .getOrCreate()

val mysqlConf = Map(
  "url" -> "jdbc:mysql://localhost:3306/mlsql_console?characterEncoding=utf8&zeroDateTimeBehavior=convertToNull&tinyInt1isBit=false",
  "driver" -> "com.mysql.jdbc.Driver",
  "user" -> "xxxxx",
  "password" -> "xxxx",
  "dbtable" -> "script_file"
)

import org.apache.spark.sql.functions.col
var df = spark.read.format("jdbc").options(mysqlConf).load()
df = df.repartitionByRange(2, col("id") )
df.write
  .format("org.apache.spark.sql.delta.sources.MLSQLDeltaDataSource").
  mode("overwrite").
  save("/tmp/datahouse/mlsql_console/script_file")
spark.close()

HBase WAL Usage

DataFrame code:

val spark = SparkSession.builder()
      .master("local[*]")
      .appName("HBase WAL Sync")
      .getOrCreate()

    val df = spark.readStream.
      format("org.apache.spark.sql.mlsql.sources.hbase.MLSQLHBaseWALDataSource").
      option("walLogPath", "/Users/allwefantasy/Softwares/hbase-2.1.8/WALs").
      option("oldWALLogPath", "/Users/allwefantasy/Softwares/hbase-2.1.8/oldWALs").
      option("startTime", "1").
      option("databaseNamePattern", "test").
      option("tableNamePattern", "mlsql_binlog").
      load()

    val query = df.writeStream.
      format("console").
      option("mode", "Append").
      option("truncate", "false").
      option("numRows", "100000").
      option("checkpointLocation", "/tmp/cpl-binlog25").
      outputMode("append")
      .trigger(Trigger.ProcessingTime("10 seconds"))
      .start()

    query.awaitTermination()

RoadMap

We hope we can support more DBs including traditional DB e.g Oracle and NoSQL e.g. HBase(WAL),ES,Cassandra in future.

How to get the initial offset

You can mannually set binlog offset, For example:

bingLogNamePrefix="mysql-bin"
binlogIndex="4"
binlogFileOffset="4"

Try using command like following to get the offset you want:

mysql> show master status;
+------------------+----------+--------------+------------------+-------------------+
| File             | Position | Binlog_Do_DB | Binlog_Ignore_DB | Executed_Gtid_Set |
+------------------+----------+--------------+------------------+-------------------+
| mysql-bin.000014 | 34913156 |              |                  |                   |
+------------------+----------+--------------+------------------+-------------------+
1 row in set (0.04 sec)

In this example, we knows that:

bingLogNamePrefix      binlogFileOffset   binlogFileOffset
mysql-bin        .     000014             34913156

this means you should configure parameters like this:

bingLogNamePrefix="mysql-bin"
binlogIndex="14"
binlogFileOffset="34913156"

Or you can use mysqlbinlog command.

mysqlbinlog \ 
--start-datetime="2019-06-19 01:00:00" \ 
--stop-datetime="2019-06-20 23:00:00" \ 
--base64-output=decode-rows \
-vv  master-bin.000004

Questions

Q1

People may meet some log like following:

Trying to restore lost connectioin to .....
Connected to ....

Please check the server_id is configured in my.cnf of your MySQL Server.

Q2

When you have started your stream to consume the binlog, but it seem nothong happen or just print :

Batch: N
-------------------------------------------
+-----+
|value|
+-----+
+-----+

Please check spark log:

20/06/18 11:57:00 INFO MicroBatchExecution: Streaming query made progress: {
  "id" : "e999af90-8d0a-48e2-b9fc-fcf1e140f622",
  "runId" : "547ce891-468a-43c5-bb62-614b38f60c39",
  "name" : null,
  "timestamp" : "2020-06-18T03:57:00.002Z",
  "batchId" : 1,
  "numInputRows" : 1,
  "inputRowsPerSecond" : 0.4458314757021846,
  "processedRowsPerSecond" : 2.9673590504451037,
  "durationMs" : {
    "addBatch" : 207,
    "getBatch" : 3,
    "getOffset" : 15,
    "queryPlanning" : 10,
    "triggerExecution" : 337,
    "walCommit" : 63
  },
  "stateOperators" : [ ],
  "sources" : [ {
    "description" : "MLSQLBinLogSource(ExecutorBinlogServer(192.168.111.14,52612),....",
    "startOffset" : 160000000004104,
    "endOffset" : 170000000000154,
    "numInputRows" : 0,
    "inputRowsPerSecond" : 0,
    "processedRowsPerSecond" : 0
  } ],
  "sink" : {
    "description" : "org.apache.spark.sql.execution.streaming.ConsoleSinkProvider@4f82b82f"
  }
}

As we can see, the startOffset/f is changing but the numInputRows is not chagned. Please try a table with a simple schema to make sure the binlog connection works fine.

If the simple schema table works fine, this is may caused by some special sql type. Please address an issue and paste spark log and your target table schema.

You can use code like this to test in your local machine:

package tech.mlsql.test.binlogserver

import java.sql.Timestamp

import org.apache.spark.sql.SparkSession
import org.apache.spark.sql.streaming.Trigger
import org.scalatest.FunSuite


object Main{
  def main(args: Array[String]): Unit = {
    val spark = SparkSession.builder()
          .master("local[*]")
          .appName("MySQL B Sync")
          .getOrCreate()

        val df = spark.readStream.
          format("org.apache.spark.sql.mlsql.sources.MLSQLBinLogDataSource").
          option("host", "127.0.0.1").
          option("port", "3306").
          option("userName", "xxxx").
          option("password", "xxxx").
          option("databaseNamePattern", "wow").
          option("tableNamePattern", "users").
          option("bingLogNamePrefix", "mysql-bin").
          option("binlogIndex", "16").
          option("binlogFileOffset", "3869").
          option("binlog.field.decode.first_name", "UTF-8").
          load()

        // print the binlog(json format)
        val query = df.writeStream.
              format("console").
              option("mode", "Append").
              option("truncate", "false").
              option("numRows", "100000").
              option("checkpointLocation", "/tmp/cpl-mysql6").
              outputMode("append")
              .trigger(Trigger.ProcessingTime("10 seconds"))
              .start()

        query.awaitTermination()
  }
}