hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ted Yu <yuzhih...@gmail.com>
Subject Re: Hbase Snapshot Export Data storage
Date Thu, 21 Dec 2017 04:41:47 GMT
For #1, major compaction would produce data files under /hbase/data,
releasing archived data.

For #2, /hbase/data

For #3, you can access your data before major compaction is performed. You
should follow best practice for major compaction on restored table.

On Wed, Dec 20, 2017 at 8:36 PM, Subash K <subash.k@ericsson.com> wrote:

> We have a use case to transfer data from one cluster to another cluster.
> As of now we are using CopyTable, but it is having impact on region server
> and it is taking lot of time to complete data transfer from one to another.
> So we are exploring on HBase Export Snapshot feature and we have planned
> to go ahead with the below steps.
>   1.  Take snapshot of a table in Source
>   2.  Execute ExportSnapshot job and send the snapshot to the destination
>   3.  Restore the snapshot sent from source.
>   4.  Now we are able to access the data.
> We want to understand how the data is handled in destination after
> restoring the snapshot. Because we can still see the data under
> /hbase/archive/data directory in HDFS and only reference data is being
> maintained in /hbase/data/
> Can someone help us to understand
>   1.  When the data under /hbase/archive/data will be removed?
>   2.  When new data is inserted into the table, where the data will be
> stored either in /hbase/archive/data or /hbase/data?
>   3.  I tried to delete the snapshot and run major_compaction for the
> table, the data got moved from /hbase/archive/data to /hbase /data. So, is
> major_compaction required always after restoring snapshot to move the data
> to its respective data location?
>   4.  I'm able to see that data is being stored in archive even if there
> is no snapshot. Under what other scenario data will be stored in
> /hbase/archive/data/ ?
> Regards,
> Subash Kunjupillai

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message