[
https://issues.apache.org/jira/browse/HBASE-12219?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14167146#comment-14167146
]
Andrew Purtell commented on HBASE-12219:
----------------------------------------
bq. With the TableStateManager from HBASE-7767 this shouldn't be a problem,
That won't be a solution for 0.98.
I don't think a TTL based cache will be the right solution if it might serve
stale descriptors that have been changed on disk but haven't expired yet. What
about maintaining a cache that retains entries indefinitely and is also
updated in place by the other master functions that deal with schema changes?
> Use optionally a TTL based cache for FSTableDescriptors#getAll() and
> FSTableDescriptors#TableDescriptorAndModtime()
> -------------------------------------------------------------------------------------------------------------------
>
> Key: HBASE-12219
> URL: https://issues.apache.org/jira/browse/HBASE-12219
> Project: HBase
> Issue Type: Bug
> Components: master
> Affects Versions: 0.94.24, 0.99.1, 0.98.6.1
> Reporter: Esteban Gutierrez
> Assignee: Esteban Gutierrez
> Labels: scalability
>
> Currently table descriptors and tables are cached once they are accessed for
> the first time. Next calls to the master only require a trip to HDFS to
> lookup the modified time in order to reload the table descriptors if
> modified. However in clusters with a large number of tables or concurrent
> clients and this can be too aggressive to HDFS and the master causing
> contention to process other requests. A simple solution is to have a TTL
> based cached for FSTableDescriptors#getAll() and
> FSTableDescriptors#TableDescriptorAndModtime() that can allow the master to
> process those calls faster without causing contention without having to
> perform a trip to HDFS for every call. to listtables() or getTableDescriptor()
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
