[
https://issues.apache.org/jira/browse/GRIFFIN-198?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16624970#comment-16624970
]
Kevin Yao commented on GRIFFIN-198:
-----------------------------------
hi,[~guoyp][~chemikadze]
First, I agree the fix of https://github.com/apache/incubator-griffin/pull/419
for nulls and empty lists.
Second, we can throw 500 exception as follows. Please customize thrown
exception in GriffinException class.
* hiveMetaStoreClient is null due to hive configuration
* cache is incomplete
* when we catch exception, such as client.getAllDatabases()
For evictHiveCache() method, we want to clear all caches. After that we call
getAllTable() or getAllDatabases() to cache all tables or databases, but for
other methods with parameters, we will not cache. I think this will give users
a better experience.
What do you think of the above views?
> HiveMetaStoreServiceImpl caches incorrect state of metastore
> ------------------------------------------------------------
>
> Key: GRIFFIN-198
> URL: https://issues.apache.org/jira/browse/GRIFFIN-198
> Project: Griffin (Incubating)
> Issue Type: Bug
> Reporter: Nikolay Sokolov
> Assignee: William Guo
> Priority: Major
>
> Currently, HiveMetaStoreServiceImpl is caching:
> * nulls and empty lists in some methods, when metastore request failed for
> some reason
> * empty list of tables as response to getAllTable(String) and getAllTable()
> calls
> Both problems result in confusing output on large amount of tables.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
