hes-dimport is a elasticsearch plugin which generate shards files of index in mapreduce.
This plugin is applied to offline import index, each time will generate an index, so that you have to define the template to organize these indexes together.
Support Integer and Long and String and Double and Float and Short data types.
Compared to bulk's load API elasticsearch, it's advantages are below:
- In a node that improves the write performance of 30%, when there is a large scale of Hadoop clusters, the overall write throughput is linearly increased.
- The throughput is linear increase when scale-out hadoop clusters.
- The index is generated once, and can be pushed to shard primary and multiple shards replicas.
- Make the elasticsearch read and write separate,do not affect the performance of the read.
- When there is a large volume of data written, you can use the relatively inexpensive hardware to extend the cluster, when you need a higher query qps and low latency, you can use a better HW.
- Based on available of MapReduce job, the write is with HA.
wget https://github.com/kewn21/eagle-hes.git
cd eagle-hes
mvn package
./bin/plugin install eagle-hes-dimport-plugin-0.0.1.zip
hadoop jar HesDimportJob {-Dindex.name={index_name}} {-Dindex.type={index_type}} [-Dautogenerate.id={true|false}] [-Dfile={meta_file_path}] {input_file_path} [delimiter]
You can describe the data format in the original data, or define in the metafile
- Each column is separated by the delimiter, when property
autogenerate.idis set to false, the first column is for id. - Other columns be separated by
,, the column name and value and type be together with:.
- The definition of each column be separated by
,, the column name and type be together with:.