Skip to content

Unified SQL Analytics Engine Based on SparkSQL

License

Apache-2.0, Apache-2.0 licenses found

Licenses found

Apache-2.0
LICENSE
Apache-2.0
LICENSE-binary
Notifications You must be signed in to change notification settings

zhangbinzaifendou/XSQL

 
 

XSQL-logo

English | 中文

XSQL is a multi-datasource query engine which designed easy to use, stable to run.1)First of all, XSQL provides a solution to read data from NoSQL database with standard SQL,so that big data engineer can concentrate on data but API with special data source . 2)XSQL takes some efforts of optimizing the execute plan of SQL execution as well as monitoring the running status of every SQL, which make user's job running healthier.

https://qihoo360.github.io/XSQL/

Features

  • XSQL supports eight built-in data source for now (e.g. Hive, Mysql, EleasticSearch, Mongo, Kafka, Hbase, Redis, Druid).
  • XSQL designs a 3-layer metadata architecture to organize data, which is datasource-database-table. So , we can provide a unified view of many data sources and no longer difficult to make a business analytical between off-line data and on-line data .
  • The main idea of XSQL are SQL Everything , SQL let program decoupling with concrete data source API , therefore DBA can upgrade data but need to taking into consideration how to migrating old tasks . More importantly, data analysts prefer SQL rather than special APIs.
  • XSQL only takes use of YARN cluster resources when necessary, this feature is useful for some usage scenario such as user treated spark-xsql as substitution of RDMS Client. We call this Pushdown Query, it makes XSQL get the ability to response DDL and Simple Query in ms delay level, as well as saving cluster resources as much as possible.
  • XSQL uses a different solution than routing , So it only parses SQL once.
  • XSQL caches metadata in runtime but don't manage metadata itself,in consideration of metadata synchronize may cause unnecessary trouble. This feature makes XSQL easy to deploy and ops.
  • XSQL provides a white-blacklist properties file to cover special usage scenario metadata should be carefully authorized.
  • XSQL can run on spark2.3 and spark2.4 for now. Jars of XSQL placed in isolated directory, which means XSQL won't take effect on your existed spark program unless you use our tool bin/spark-xsql. So, just try XSQL on your existed spark distribution, All things will work fine as normal.

Quick Start

Environment Requirements of Build

  • jdk 1.8+

Build XSQL:

  1. To get started with XSQL, you can build it by yourself. For example,

    git clone https://github.com/Qihoo360/XSQL
    

    You can also get pre-built XSQL from Release Pages .

  2. When you want to create a XSQL distribution of source code, which is similar to the release package in the Release Pages , use build-plugin.sh in the root directory of project. For example:

    XSQL/build-plugin.sh
    

    This will produce a .tgz file named like xsql-[project.version]-plugin-spark-[spark.version].tgz in the root directory of project.

    To create a XSQL distribution like natural Spark distribution, use build.sh in the root directory of project. For example:

    XSQL/build.sh
    

    This will produce a .tgz file named like xsql-[project.version]-bin-spark-[spark.version].tgz in the root directory of project.

Environment Requirements of Running

  • jdk 1.8+

  • hadoop 2.7.2+

  • spark 2.4.x

Installing XSQL:

  1. Build the XSQL tar xsql-[project.version]-[plugin|bin]-spark-[spark.version].tgz following the above steps or Download from Release Pages.

  2. If you have installed spark in your machine, please use the plugin version which size is 30M+. Or you need to install the bin version, which is about 300M + in size, which is far more than the plugin version.

    For either plugin version or bin version, both need to be extracted into your software directory at first. For example:

    tar xvf xsql-0.6.0-bin-spark-2.4.3.tgz -C "/path/of/software"

    The destination directory of plugin version is different to bin version:

    tar xvf xsql-0.6.0-plugin-spark-2.4.3.tgz -C "/path/of/sparkhome"
  3. XSQL needs to know the information (like url and authorization ) of each data source .You can configure them in xsql.conf under conf directory. We provided a template file to help user configuring XSQL. For example:

    mv conf/xsql.conf.template xsql.conf
    

    There is an example of MySQL configuration:

    spark.xsql.datasources                     default
    spark.xsql.default.database                real_database
    spark.xsql.datasource.default.type         mysql
    spark.xsql.datasource.default.url          jdbc:mysql://127.0.0.1:2336
    spark.xsql.datasource.default.user         real_username
    spark.xsql.datasource.default.password     real_password
    spark.xsql.datasource.default.version      5.6.19
    

Running XSQL:

  1. If you are familiar with spark-sql , we provide an improved bash tool bin/spark-xsql. XSQL can be started in Cli mode by following command:

    $SPARK_HOME/bin/spark-xsql

    Feel free to input any SQL/HiveSQL in the prompt line:

    spark-xsql> show datasources;
    
  2. If you are familiar with DataSet API, start from our scala api is a good choice. For example:

    var spark = SparkSession
      .builder()
      .enableXSQLSupport()
      .getOrCreate()
    spark.sql("show datasources")

FAQ

Connect to more datasource

Advanced Configuration

XSQL Specific Query Language

Contact Us

Mail Lists: For developers xsql_dev@groups.163.com, For users xsql_user@groups.163.com. Add yours by emailing it.

QQ Group for Chinese user : No.838910008

About

Unified SQL Analytics Engine Based on SparkSQL

Resources

License

Apache-2.0, Apache-2.0 licenses found

Licenses found

Apache-2.0
LICENSE
Apache-2.0
LICENSE-binary

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Scala 68.3%
  • Python 12.7%
  • HiveQL 8.4%
  • Java 6.3%
  • TSQL 1.6%
  • Shell 0.9%
  • Other 1.8%