[
https://issues.apache.org/jira/browse/HIVE-17595?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16226214#comment-16226214
]
Daniel Dai commented on HIVE-17595:
-----------------------------------
Couple of comments:
1. Can you take a note what problem did you see if updating database
last.repl.id earlier? Is that some bootstrap tasks get skipped?
2. Name EfficientDAGTraversal, do we have a regular DAGTraversal? If not,
probably just leave DAGTraversal is better; DependencyCollectionFunction, might
be better AddDependencyToLeaves?
3. How about createEndReplLogTask? Shall we do it after all tasks as well?
> Correct DAG for updating the last.repl.id for a database during bootstrap load
> ------------------------------------------------------------------------------
>
> Key: HIVE-17595
> URL: https://issues.apache.org/jira/browse/HIVE-17595
> Project: Hive
> Issue Type: Bug
> Components: HiveServer2
> Affects Versions: 3.0.0
> Reporter: anishek
> Assignee: anishek
> Fix For: 3.0.0
>
> Attachments: HIVE-17595.0.patch, HIVE-17595.1.patch,
> HIVE-17595.2.patch
>
>
> We update the last.repl.id as a database property. This is done after all the
> bootstrap tasks to load the relevant data are done and is the last task to be
> run. however we are currently not setting up the DAG correctly for this task.
> This is getting added as the root task for now where as it should be the last
> task to be run in a DAG. This becomes more important after the inclusion of
> HIVE-17426 since this will lead to parallel execution and incorrect DAG's
> will lead to incorrect results/state of the system.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
