[
https://issues.apache.org/jira/browse/HIVE-16119?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15897755#comment-15897755
]
Vihang Karajgaonkar edited comment on HIVE-16119 at 3/6/17 6:18 PM:
--------------------------------------------------------------------
On a second thought, I wonder if we really need to increase the pool size
disregarding user configs. The reason it was added in the first place was due
to an issue with the recursive call. Since that is no more the case
(HIVE-15879), we should probably keep using the poolSize given by the config.
Disregarding the user given config also has a side effect mentioned in
HIVE-16014 where the mismatched pool size becomes a bottle neck, so increasing
the pool size here will not necessarily mean that query performance is improved
proportionally.
was (Author: vihangk1):
On a second thought, I wonder if we really need to increase the pool size
disregarding user configs. The reason it was added in the first place was due
to an issue with the recursive call. Since that is no more the case
(HIVE-15879), we should probably keep using the poolSize given by the user.
> HiveMetaStoreChecker - singleThread/parallel logic duplication
> --------------------------------------------------------------
>
> Key: HIVE-16119
> URL: https://issues.apache.org/jira/browse/HIVE-16119
> Project: Hive
> Issue Type: Improvement
> Reporter: Zoltan Haindrich
> Assignee: Zoltan Haindrich
> Priority: Minor
> Attachments: HIVE-16119.1.patch
>
>
> It looks to me that the main logic is duplicated, because of multithereading
> support:
> * {{HiveMetaStoreChecker#PathDepthInfoCallable#processPathDepthInfo}}
> * {{HiveMetaStoreChecker#checkPartitionDirsSingleThreaded}}
> It might be possible to remove the singleThreaded methods by using a special
> executor for single thread support: {{MoreExecutors.sameThreadExecutor()}}.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
