[
https://issues.apache.org/jira/browse/HIVE-11510?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14720434#comment-14720434
]
Sushanth Sowmyan commented on HIVE-11510:
-----------------------------------------
With the current patch, the metastore will do a LOG.debug for every single null
record, which can be a lot, and will also slow down that process a lot.
Would it be possible to simply update the UpdateMStorageDescriptorTblURIRetVal
class with a int numNullRecords initialized to zero and incremented each time
you get a null? Also, in that case, I would imagine that we shouldn't add that
location to badRecords, since that would bloat up the size of badRecords
unnecessarily. After we do that, we can then do a singular log in
HiveMetaTool.printTblURIUpdateSummary along with the other statistics,
mentioning how many null records we found, and that that is okay if the user
has that many indexes/views.
> Metatool updateLocation warning on views
> ----------------------------------------
>
> Key: HIVE-11510
> URL: https://issues.apache.org/jira/browse/HIVE-11510
> Project: Hive
> Issue Type: Bug
> Components: Database/Schema
> Affects Versions: 0.14.0
> Reporter: Eric Czech
> Assignee: Wei Zheng
> Attachments: HIVE-11510.1.patch
>
>
> If views are present in a hive database, issuing a 'hive metatool
> -updateLocation' command will result in an error like this:
> ...
> Warning: Found records with bad LOCATION in SDS table..
> bad location URI: null
> bad location URI: null
> bad location URI: null
> ....
> Based on the source code for Metatool, it looks like there would then be a
> "bad location URI: null" message for every view and it also appears this is
> happening simply because the 'sds' table in the hive schema has a column
> called location that is NULL only for views.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
