Apache CarbonData Dev Mailing List archive › Apache CarbonData JIRA issues

[GitHub] carbondata pull request #1418: Support db location

Classic

List

Threaded

166 messages Options

1234 ... 9

qiuchenjian-2

[GitHub] carbondata pull request #1418: Support db location

GitHub user mohammadshahidkhan opened a pull request:

https://github.com/apache/carbondata/pull/1418

Support db location

**Problem:**
Currently carbon stores the database and tables at the fixed configured location âcarbon.storeLocationâ. So even though in case of federated cluster i.e. multiple independent Namespaces/ Namenodes only one Namespace/ Namenode can be used as storage for the database and tables.
The requirement is to support when create database can specify the LOCATION; the location can be viewfs path, name service path.
Using database location attribute configuration different name nodes can be used to store different databases. Having different name nodes for different databases will scale the read/write throughput.
**Solution:**
Setting the LOCATION attribute for a new database is a way to work with sets of files in an HDFS directory structure outside the fixed carbon store location.
In Create database command user can specify the location, the location where database and table under that database namespace could be stored.
1. The database at specified location could be created at defined location using create database command with location attribute.
Syntax:
CREATE DATABASE [IF NOT EXISTS] database_name[COMMENT 'database_comment'] [LOCATION hdfs_path];
The HDFS path can be viewfs path, nameservice path.
2. The database name and database location will be retrieved from the HiveSessionCatalog. The tables will be retrieved by scanning the table folders.
3. The read/write operation on old databases/ new databases created without specifying the LOCATION attribute should be performed the fixed carbon store location.
4. The table path formation will be same as the hive. The table path should be formed from database location or fixed Carbon store location and table name as given below.
The will be three possible scenarios:
I. Table path for the databases defined with location attribute.
tablePath = databaseLocation +â/â + tableName

II. Table path for the databases defined without location attribute.
tablePath = carbon.storeLocation + â/â + database_Name+â.dbâ +â/â + tableName

III. New table path for the default database.
tablePath = carbon.storeLocation +â/â + tableName

5. The carbon should support hybrid Nameservice path i.e. different operation could be performed in different name service. For example in load command the raw CSV files could be read from the nameservice ns1 i.e. âhdfs://ns1/user/data.csvâ and database location could point to viewfs path âviewfs://federation_cluster/dir1/carbonstoreâ.

Example:
CREATE DATABASE carbon LOCATION âviewfs://federation_cluster/dir1/carbonstoreâ;
CREATE TABLE carbon.carbon_table (c1 string, c2 int) stored by âcarbondataâ;
LOAD DATA INPATH âhdfs://ns1/user/data.csvâ into table carbon.carbon_table;

You can merge this pull request into a Git repository by running:

$ git pull https://github.com/mohammadshahidkhan/incubator-carbondata support_db_location

Alternatively you can review and apply these changes as the patch at:

https://github.com/apache/carbondata/pull/1418.patch

To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:

This closes #1418

----
commit b92be927761391f37018f5018521813f588825b8
Author: mohammadshahidkhan <[hidden email]>
Date: 2017-10-10T05:20:11Z

Support database location:
The database name and database location will be retrieved from the HiveSessionCatalog. The tables will be retrieved by scanning the table folders.
The read/write operation on old databases/ new databases created without specifying the LOCATION attribute should be performed the fixed carbon store location.

commit 2a88536a19f483b8aee833039ba364ec85a78ee4
Author: mohammadshahidkhan <[hidden email]>
Date: 2017-10-10T06:00:43Z

Support configurable path for the store modification info file (modifiedTime.mdt).

commit 7ab2f29b38d35e2590a40f5e2f0939b5033715dc
Author: mohammadshahidkhan <[hidden email]>
Date: 2017-10-10T10:17:00Z

Fixed drop database issue

commit 524f4532e8fcaf0c40a884c7c5b51fd0d887066e
Author: mohammadshahidkhan <[hidden email]>
Date: 2017-10-11T10:00:51Z

Corrected lock file location

commit dbb3160919a28c9b0eb93ab9ed8dfecfe93129b4
Author: mohammadshahidkhan <[hidden email]>
Date: 2017-10-16T08:58:07Z

Making carbon table path same as hive.

----

---

qiuchenjian-2