CopyTable

Copies data from one MapR-DB table to another MapR-DB table.

Note: This utility is different from Apache HBase's CopyTable utility. When copying data to MapR-DB tables, it is recommended to use the MapR-DB version.

If the destination table does not exist, CopyTable creates the target table with the same metadata (column families and access control expressions) as the source table, and then copies data.

If the destination table exists, CopyTable copies data only.

If you manually set up replication to a MapR-DB table, you use this utility to perform an initial load of source data to the replica before table replication begins.

Required Permissions

The user that runs the CopyTable utility must have the following permissions:
  • The permission readAce on the volume where the source table is located, and the permission writeAce on the volume where the destination table is or will be located.
  • The permission adminperm on the source table.
  • The permission for column-family and column reads (readperm) on the data in the source table that you want to copy.
  • When bulkload = false, the permission for column writes (writeperm) on the destination table.
  • When bulkload = true (default), the permission to load the destination table with bulk loads (bulkloadperm).
  • If the destination table does not yet exist: createrenamefamily on the source table.
Note: The mapr user is not treated as a superuser. MapR-DB does not allow the mapr user to run this utility unless that user is given the relevant permission or permissions with access-control expressions.

If CopyTable is run between tables on different clusters, the user that runs the command must have the required permissions on each cluster.

For information about how to set permissions on volumes, see Setting/Modifying Whole Volume ACEs.

For information about how to set permissions on tables, see Enabling Table Authorizations with Access Control Expressions.

Syntax

hbase com.mapr.fs.hbase.tools.mapreduce.CopyTable
    -src <source table path> -dst <destination table path>
    [-columns cf1[:col1],...] [-maxversions <max number of versions to copy>]
    [-starttime <time>]
    [-endtime <time>]
    [-mapreduce <true|false> (default: true)]
    [-bulkload <true|false> (default: true)]
    [-numthreads <numThreads> (default:16, valid only when -mapreduce is false)]

Parameters

Parameters Description
src

The path to the source table that you want to replicate.

  • For a path on the local cluster, start the path at the volume mount point. For example, for a table named testsrc under a volume with a mount point at /volume1, specify the following path: /volume1/testsrc
  • For a path on another cluster, you must also specify the cluster name in the path. For example, for a table named customersrc under volume1 in the sanfrancisco cluster, specify the following path: /mapr/sanfrancisco/volume1/customersrc
dst

The path to the replica.

  • For a table on the local cluster, start the path at the volume mount point. For example, for a table named testdst under a volume with a mount point at /volume1, specify the following path: /volume1/testdst
  • For a table on another cluster, you must also specify the cluster name in the path. For example, for a table named customerdst under volume1 in the sanfrancisco cluster, specify the following path:/mapr/sanfrancisco/volume1/customerdst
columns By default, all columns in the source table are copied. If you do not want to copy all columns in the table, you can specify columns to copy. Provide a comma-separated list of column families or columns from a certain column family (column family:qualifier).

For example, use the following syntax to copy only the purchases column family and the stars column in the reviews column family: -columns purchases,reviews:stars

maxversions By default, all versions from the source table are copied. If you do not want to copy all versions, use this parameter to specify the number of versions to copy.
starttime By default, all table values regardless of their associated timestamp are copied. You can specify a timestamp to indicate the table cell version at which to start the copy. The timestamp is a long integer value that is either user-defined or system-defined (epoch in milliseconds). Values with timestamps lower than the specified timestamp will not be copied to the destination.
endtime By default, all table values regardless of their associated timestamp are copied. You can specify a timestamp to indicate the table cell version at which to end the copy. The timestamp is a long integer value that is either user-defined or system-defined (epoch in milliseconds). Values with timestamps greater than or equal to the specified timestamp will not be copied to the destination.
mapreduce

A Boolean value that specifies whether or not to use a MapReduce program to perform the copying operation. The default, preferred method is to use a MapReduce program (true).

When this parameter is set to false, a client process uses multiple threads to read rows of the source table and write rows to the destination table.

The MapReduce program runs as MapReduce v1 job or MapReduce v2 application based on the MapReduce mode that is configured on this node. For more information, see Managing the MapReduce Mode.

bulkload A Boolean value that specifies whether or not to perform a full bulk load of the table. The default is to use bulk loading (true). When you use bulk loading, the utility automatically unsets the bulk load mode on the table to restore normal client operations at the end of the table copy operation. For more information, see Bulk Loading and MapR-DB Tables.
numthreads When -mapreduce is false, this parameter specifies the number of threads allocated to perform the copying of data. The default is 16. If additional CPU resources are available, you might want to increase the number of threads to achieve better performance.

Monitoring the CopyTable Operation

Use one of the following methods to monitor the progress of the copying of table data:
  • If the copy table operation runs as a MapReduce v1 job, monitor the job using the JobTracker UI.
  • If the copy table operation runs as a MapReduce v2 application, monitor the application using the ResourceManager UI.
  • If the copy table operation runs as a client process, go to the Tables view of the destination table in the MapR Control System. Then, on the Region tab, monitor the pace at which the number of rows increases.

Example

Copies table data with timestamp greater than 1423226300000 (Fri, 06 Feb 2015 12:38:20 GMT) from one MapR-DB table to another MapR-DB table:
[user@hostname ~]$  hbase com.mapr.fs.hbase.tools.mapreduce.CopyTable -src /t1 -dst /t1_copy7  -starttime 1423226300000