Background
In data warehouse construction, Hive is usually used to process the raw data (at the petabyte level), perform time-consuming ETL jobs, and hand over the results (at the terabyte level) to a quasi-real-time computing engine such as Tencent Cloud TCHouse-P to connect BI tools and present reports in quasi real time.
This document describes how to import data from Hive on EMR to Tencent Cloud TCHouse-P via COS.
Directions
Note:
Tencent Cloud TCHouse-P supports only CSV and GZIP but not ORC and Parquet formats.
The efficiency of importing COS data to Tencent Cloud TCHouse-P depends on the number of files, which is recommended to be N times the number of compute nodes in Tencent Cloud TCHouse-P.
1. Enable EMR's capability to read and write COS data.
First, you need to ensure that EMR is able to read and write COS data. You can click Enable COS when creating an EMR instance.
2. Create a Hive local table and write data into it.
create table hive_local_table(c1 int, c2 string, c3 int, c4 string);insert into hive_local_table values(1001, 'c2', 99, 'c4'),(1002, 'c2', 100, 'c4'),(1003, 'c2', 101, 'c4'),(1004, 'c2', 100, 'c4'),(1005, 'c2', 101, 'c4')
3. Create a Hive COS external table.
create table hive_cos_table(c1 int, c2 string, c3 int, c4 string)row format delimited fields terminated by ','LINES TERMINATED BY '\\n'stored as textfile location 'cosn://{bucket_name}/{dir_name}';
4. Import the local data into COS.
insert into hive_cos_table select * from hive_local_table;
After successful write, you can see the file in the corresponding COS directory.
5. Create a COS external table in Tencent Cloud TCHouse-P.
CREATE READABLE EXTERNAL TABLE snova_cos_table (c1 int, c2 varchar(32), c3 int, c4 varchar(32))LOCATION('cos:// {BUCKET}-{APPID}.cos.{REGION}.myqcloud.com/{PREFIX} secretKey=**** secretId=***')FORMAT 'csv';
6. Create a local table in Tencent Cloud TCHouse-P and import data into it.
create table snova_local_table(c1 int, c2 text, c3 int, c4 text);insert into snova_local_table select * from snova_cos_table;
Yes
No
Was this page helpful?