[
https://issues.apache.org/jira/browse/HIVE-1841?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12969440#action_12969440
]
Edward Capriolo commented on HIVE-1841:
---------------------------------------
Ning,
I agree with
{noformat}
<property>
<name>datanucleus.autoCreateSchema</name>
<value>true</value>
</property>
{noformat}
But datanucleus.fixedDatastore is not specified and it defaults to true. This
causes and auto-upgrade which could be dangerous. I believe we should set
datanucleus.fixedDatastore to true, or at very least document it in
hive-default.xml.
I say this because my first assumption was if i build a hive from trunk and
point it at a production metastore, run some select queries for QA, it could
"do no harm". However since datanucleus.fixedDatastore is true as soon as a
client touches the metastore it will make any changes it feels are appropriate.
In the future imagine and old hive install laying around, what if an old hive
0.4.0 instances points at an hive 0.9.0 metastore if could try to add something
that has since been removed, or worse.
> datanucleus.fixedDatastore should be true in hive-default.xml
> --------------------------------------------------------------
>
> Key: HIVE-1841
> URL: https://issues.apache.org/jira/browse/HIVE-1841
> Project: Hive
> Issue Type: Improvement
> Components: Configuration
> Affects Versions: 0.6.0
> Reporter: Edward Capriolo
> Fix For: 0.7.0
>
>
> Two datanucleus variables:
> {noformat}
> <property>
> <name>datanucleus.autoCreateSchema</name>
> <value>false</value>
> </property>
> <property>
> <name>datanucleus.fixedDatastore</name>
> <value>true</value>
> </property>
> {noformat}
> are dangerous. We do want the schema to auto-create itself, but we do not
> want the schema to auto update itself.
> Someone might accidentally point a trunk at the wrong meta-store and
> unknowingly update. I believe we should set this to false and possibly trap
> exceptions stemming from hive wanting to do any update. This way someone has
> to actively acknowledge the update, by setting this to true and then starting
> up hive, or leaving it false, removing schema modifies for the user that hive
> usages, and doing all the time and doing the updates by hand.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
