[
https://issues.apache.org/jira/browse/HBASE-16504?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sai Teja Ranuva resolved HBASE-16504.
-------------------------------------
Resolution: Not A Problem
Assignee: Sai Teja Ranuva
The timestamp of the mutation is not useful as the cells already have their
timestamp set at the time of replication.
In detail:
Normally, if the cells do not have their timestamp set, put/delete timestamp is
used instead, when it is LATEST_TIMESTAMP, server time is used.
This flow doesn't happen in case of replication, as we need not look outside of
the cell to get timestamp.
> Replication is not preserving the timestamp of the cell
> -------------------------------------------------------
>
> Key: HBASE-16504
> URL: https://issues.apache.org/jira/browse/HBASE-16504
> Project: HBase
> Issue Type: Bug
> Components: Replication
> Reporter: Sai Teja Ranuva
> Assignee: Sai Teja Ranuva
> Priority: Minor
> Labels: replication, replication-testing
>
> The Replication procedure in ReplicationSink.replicateEntries() method is not
> preserving the timestamp of the cell.
> Pointer to the code:
> m = CellUtil.isDelete(cell) ? new Delete(cell.getRowArray(),
> cell.getRowOffset(),cell.getRowLength()) : new Put(cell.getRowArray(),
> cell.getRowOffset(),cell.getRowLength());
> The Put and Delete constructors called here assign the timestamp with
> HConstants.LATEST_TIMESTAMP. Instead we need to keep the timestamp of the
> cell here.
> Also there doesn't seem to be a test which checks if replication is
> preserving the timestamp.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
