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/
- 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.
Environment Requirements of Build
- jdk 1.8+
-
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 .
-
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
-
Build the XSQL tar
xsql-[project.version]-[plugin|bin]-spark-[spark.version].tgz
following the above steps or Download from Release Pages. -
If you have installed spark in your machine, please use the
plugin
version which size is 30M+. Or you need to install thebin
version, which is about 300M + in size, which is far more than theplugin
version.For either
plugin
version orbin
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 tobin
version:tar xvf xsql-0.6.0-plugin-spark-2.4.3.tgz -C "/path/of/sparkhome"
-
XSQL needs to know the information (like url and authorization ) of each data source .You can configure them in
xsql.conf
underconf
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
-
If you are familiar with
spark-sql
, we provide an improved bash toolbin/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;
-
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")
Mail Lists: For developers [email protected]
, For users [email protected]
. Add yours by emailing it.
QQ Group for Chinese user : No.838910008