Yi Jin updated HAWQ-991:
    Fix Version/s:     (was:

> "HAWQ register" could register tables according to .yml configuration file
> --------------------------------------------------------------------------
>                 Key: HAWQ-991
>                 URL: https://issues.apache.org/jira/browse/HAWQ-991
>             Project: Apache HAWQ
>          Issue Type: New Feature
>          Components: Command Line Tools
>    Affects Versions:
>            Reporter: hongwu
>            Assignee: hongwu
>             Fix For: backlog
> Scenario: 
> 1. For cluster Disaster Recovery. Two clusters co-exist, periodically import 
> data from Cluster A to Cluster B. Need Register data to Cluster B.
> 2. For the rollback of table. Do checkpoints somewhere, and need to rollback 
> to previous checkpoint. 
> Description:
> Register according to .yml configuration file. 
> hawq register [-h hostname] [-p port] [-U username] [-d databasename] [-c 
> config] [--force][--repair] <tablename> 
> Behaviors:
> 1. If table doesn't exist, will automatically create the table and register 
> the files in .yml configuration file. Will use the filesize specified in .yml 
> to update the catalog table. 
> 2. If table already exist, and neither --force nor --repair configured. Do 
> not create any table, and directly register the files specified in .yml file 
> to the table. Note that if the file is under table directory in HDFS, will 
> throw error, say, to-be-registered files should not under the table path.
> 3. If table already exist, and --force is specified. Will clear all the 
> catalog contents in pg_aoseg.pg_paqseg_$relid while keep the files on HDFS, 
> and then re-register all the files to the table.  This is for scenario 2.
> 4. If table already exist, and --repair is specified. Will change both file 
> folder and catalog table pg_aoseg.pg_paqseg_$relid to the state which .yml 
> file configures. Note may some new generated files since the checkpoint may 
> be deleted here. Also note the all the files in .yml file should all under 
> the table folder on HDFS. Limitation: Do not support cases for hash table 
> redistribution, table truncate and table drop. This is for scenario 3.
> Requirements:
> 1. To be registered file path has to colocate with HAWQ in the same HDFS 
> cluster.
> 2. If to be registered is a hash table, the registered file number should be 
> one or multiple times or hash table bucket number.

This message was sent by Atlassian JIRA

Reply via email to