[
https://issues.apache.org/jira/browse/TEZ-3331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15373977#comment-15373977
]
Hitesh Shah commented on TEZ-3331:
----------------------------------
The shims are straightforward in nature. The base HadoopShim class provides the
interface which can have different impls based on diff hadoop versions. At
runtime, based on the hadoop version in the classpath, the appropriate impl
will be picked ( via service loaders ).
The distributed cache handling could use the shim but that is a "maybe" given
that today, Tez expects users of the API to provide local resources that are
already on HDFS or any DFS and does not do any upload functionality. This means
that Hive/Pig, etc will need to make use of the YARN api for the distributed
cache lib if they do not want to manage local resources themselves.
> Add operation specific HDFS counters for Tez UI
> -----------------------------------------------
>
> Key: TEZ-3331
> URL: https://issues.apache.org/jira/browse/TEZ-3331
> Project: Apache Tez
> Issue Type: Bug
> Reporter: Jitendra Nath Pandey
> Attachments: TEZ-3331.wip.2.patch, TEZ-3331.wip.3.patch,
> TEZ-3331.wip.4.patch, TEZ-3331.wip.patch
>
>
> Hadoop has added several operation specific counters in the FileSystem
> statistics (HADOOP-13065). These counters are useful to track file system
> operations more granularly. It would be great to track these counters for Tez
> and expose them via UI as well.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
