[
https://issues.apache.org/jira/browse/HIVE-13560?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15260654#comment-15260654
]
Daniel Dai commented on HIVE-13560:
-----------------------------------
bq. In conf/tez/hive-site.xml did you mean to stomp the
hive.orc.splits.ms.footer.cache.enabled value?
Rollback this change. This is to debug the ORC writing conflict and shall not
be in the final patch.
bq. Why take out the option to test Tephra, since we haven't taken out the
Tephra connector?
We only have dummy code there. But anyway, the patch don't have to touch it.
bq. HBaseStore.java, line 451, why did you change the catch from IOException to
Exception? I can't see any other changes in the code that should require this.
Not necessary, rollback it back
bq. I don't understand why you removed the transactions from
getPartitionsByExpr.
I didn't remove it. My change to getPartitionsByExpr is to make it wrap well in
the transaction
bq. I think removing the transactions around the get/putFileMetadata is fine,
but we should explicitly comment that these operations are outside of the
transactions and why.
Added comments.
bq. It's not clear to me that you need OmidHBaseConnection.transaction to be a
thread local variable. HBaseReadWrite is already a thread local in HBaseStore,
so you should be guaranteed that there's an OmidHBaseConnection per thread.
Sure, didn't realize HBaseConnection already in thread local context, thanks!
> Adding Omid as connection manager for HBase Metastore
> -----------------------------------------------------
>
> Key: HIVE-13560
> URL: https://issues.apache.org/jira/browse/HIVE-13560
> Project: Hive
> Issue Type: Improvement
> Components: HBase Metastore
> Reporter: Daniel Dai
> Assignee: Daniel Dai
> Attachments: HIVE-13560.1.patch, HIVE-13560.2.patch,
> HIVE-13560.3.patch
>
>
> Adding Omid as a transaction manager to HBase Metastore.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
