[
https://issues.apache.org/jira/browse/HIVE-16171?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15966335#comment-15966335
]
Sankar Hariappan commented on HIVE-16171:
-----------------------------------------
Hi [~ekoifman],
For this truncate table support with change management, there are few changes
done to update the table stats.
Could you please have a look at it?
I shall see couple of test failures due to this change where the result-logs
have differences. But, I believe, the new code is fine and the result file
should be updated for these 2 failed tests. Please give your opinion if I'm
wrong.
Brief description of my change:
Base:
1. Update Stats (RowNum=0, NumOfFiles=0 etc).
2. AlterTable with new Stats. -- This effectively regenerate the stats from the
table again and hence set to older values.
3. TruncateTable (Trash data files)
New Code:
1. TruncateTable (Trash data files)
2. Update Stats (RowNum=0, NumOfFiles=0 etc).
3. AlterTable with new Stats. This was tested and is working fine.
cc: [~thejas], [~sushanth]
> Support replication of truncate table
> -------------------------------------
>
> Key: HIVE-16171
> URL: https://issues.apache.org/jira/browse/HIVE-16171
> Project: Hive
> Issue Type: Sub-task
> Components: repl
> Affects Versions: 2.1.0
> Reporter: Sankar Hariappan
> Assignee: Sankar Hariappan
> Labels: DR
> Attachments: HIVE-16171.01.patch, HIVE-16171.02.patch,
> HIVE-16171.03.patch, HIVE-16171.04.patch, HIVE-16171.05.patch
>
>
> Need to support truncate table for replication. Key points to note.
> 1. For non-partitioned table, truncate table will remove all the rows from
> the table.
> 2. For partitioned tables, need to consider how truncate behaves if truncate
> a partition or the whole table.
> 3. Bootstrap load with truncate table must work as it is just
> loadTable/loadPartition with empty dataset.
> 4. It is suggested to re-use the alter table/alter partition events to handle
> truncate.
> 5. Need to consider the case where insert event happens before truncate table
> which needs to see their data files through change management. The data files
> should be recycled to the cmroot path before trashing it.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
