You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@tez.apache.org by "Hitesh Shah (JIRA)" <ji...@apache.org> on 2014/09/04 22:27:24 UTC
[jira] [Commented] (TEZ-1384) Move recovery related code into inner
class
[ https://issues.apache.org/jira/browse/TEZ-1384?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14121937#comment-14121937 ]
Hitesh Shah commented on TEZ-1384:
----------------------------------
Comments:
- class name DAGRecover, TaskRecover, etc do not really signify what functionality the class provides.
- would be good to have a better name if possible.
- also, it would be good to add docs.
- at this point, it is not clear what the class provides and when new functions are added, whether the functions belong in this class or the parent class.
- also, the member field name should be changed to be a bit more clear than just "recover".
- Is all the code just re-factored to move into an inner class or there were other changes also done? Also, the original objective of the re-factoring was to keep the stored and restored state in sync - however, the inner class seems to be updated data in the parent class - is that intentional?
- there are functions like logJob* in all classes. Maybe they should be renamed for clarity.
- VertexRecover has recover() and startRecover() - what is the difference? Would be good to add docs and rename functions for clarity.
> Move recovery related code into inner class
> -------------------------------------------
>
> Key: TEZ-1384
> URL: https://issues.apache.org/jira/browse/TEZ-1384
> Project: Apache Tez
> Issue Type: Sub-task
> Reporter: Jeff Zhang
> Assignee: Jeff Zhang
> Attachments: Tez-1384.patch
>
>
> Currently each entity (DAG, Vertex, Task, TaskAttempt) has some common recovery code like log history event and restore from history event. These are 2 opposite aspects of recovery. One for store status while the other is for restore status. This jira is for putting these pieces of code together ( in an inner class ). In this way, it is easy to maintain and cut down the possibility that one field is not stored or restored.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)