[
https://issues.apache.org/jira/browse/HIVE-18336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16308654#comment-16308654
]
Eugene Koifman commented on HIVE-18336:
---------------------------------------
It would be useful to add optional validation to acid processing.
For example, HIVE-18190 does a check whether the contents of a delta/ were
written by Acid code path, i.e. have ROW_IDs embedded. Perhaps in safe mode it
should check each file rather than choosing one.
Another example, on read of a delta/ check that it has a full compliment of
bucket files if running with MR.
That no "extra" files are present.
That all files have the same schema.
Acid tables are meant to be written by Hive but I've seen multiple times where
users forget this and try to move/copy files resulting strange issues.
Maybe "safe mode" is not the right name/approach. Perhaps this should be a
separate tool that can analyze an acid table and report potential issues.
> add Safe Mode
> -------------
>
> Key: HIVE-18336
> URL: https://issues.apache.org/jira/browse/HIVE-18336
> Project: Hive
> Issue Type: Bug
> Components: Transactions
> Reporter: Eugene Koifman
> Assignee: Eugene Koifman
>
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
