[
https://issues.apache.org/jira/browse/HIVE-9582?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14381023#comment-14381023
]
Thiruvel Thirumoolan commented on HIVE-9582:
--------------------------------------------
Thanks [~sushanth] for your time and inputs. Much appreciated.
I was of the opinion getHiveClient() was internal and not supposed to be used
outside (yeah, we have lot of public methods). I asked Oozie team at Yahoo! to
use HCatClient going forward, I am not sure those changes are in trunk. I
didn't know falcon/sqoop might be using it.
I will change as you suggested and update another patch.
> HCatalog should use IMetaStoreClient interface
> ----------------------------------------------
>
> Key: HIVE-9582
> URL: https://issues.apache.org/jira/browse/HIVE-9582
> Project: Hive
> Issue Type: Sub-task
> Components: HCatalog, Metastore
> Affects Versions: 0.14.0, 0.13.1
> Reporter: Thiruvel Thirumoolan
> Assignee: Thiruvel Thirumoolan
> Labels: hcatalog, metastore, rolling_upgrade
> Attachments: HIVE-9582.1.patch, HIVE-9582.2.patch, HIVE-9582.3.patch,
> HIVE-9582.4.patch, HIVE-9582.5.patch, HIVE-9583.1.patch
>
>
> Hive uses IMetaStoreClient and it makes using RetryingMetaStoreClient easy.
> Hence during a failure, the client retries and possibly succeeds. But
> HCatalog has long been using HiveMetaStoreClient directly and hence failures
> are costly, especially if they are during the commit stage of a job. Its also
> not possible to do rolling upgrade of MetaStore Server.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
