Skip to content

TiSpark with HDFS

shiyuhang0 edited this page Jun 20, 2022 · 1 revision

This article introduces how to use TiSpark with HDFS.

Set up

  1. create TiDB Table
CREATE TABLE `test`.`tidb` (
  `id` int(11) NOT NULL, 
  `name` varchar(255) NOT NULL, 
  PRIMARY KEY (`id`)
);
  1. create data.csv

take 0x01 delimiter and \n newline as example

2^Ashi
3^Ayu
  1. load the CSV to hdfs
hdfs dfs -put data.csv /
  1. start spark-shell
./bin/spark-shell --jars tispark-assembly-3.0-2.5.1.jar

Write to TiDB from HDFS

you can also use Spark JDBC DataSource to do it, see here

import org.apache.spark.sql.types._

// read from hdfs
val schema = new StructType().add("id",IntegerType).add("name",StringType)
val df = spark.read.format("csv").option("delimiter","\u0001").schema(schema).load("hdfs://${ip:port}/data.csv")

// write to tidb
val tidbOptions = Map(
  "tidb.addr" -> "${ip}",
  "tidb.password" -> "",
  "tidb.port" -> "4000",
  "tidb.user" -> "root"
)
df.write.format("tidb").options(tidbOptions).option("database", "test").option("table", "tidb").mode("append").save()

// if we want to replace
df.write.format("tidb").options(tidbOptions).option("database", "test").option("table", "tidb").option("replace","true").mode("append").save()

Write to HDFS from TiDB

you can also use Spark JDBC DataSource to do it, see here

// read from tidb
val df = spark.sql("select * from tidb_catalog.test.tidb")

// write to hdfs
df.write.format("csv").option("delimiter","\u0001").save("hdfs://${ip:port}/save") 

zip file

Spark support read from the zipped file in Hadoop directly, take gz file for example:

val schema = new StructType().add("id",IntegerType).add("name",StringType)
val df = spark.read.format("csv").option("delimiter","\u0001").schema(schema).load("hdfs://${ip:port}/data.csv.gz")
Clone this wiki locally