[
https://issues.apache.org/jira/browse/ASTERIXDB-1747?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15730024#comment-15730024
]
Yingyi Bu commented on ASTERIXDB-1747:
--------------------------------------
Also this item:
-- have a separate interface method, e.g., called prepare, to pre-distribute a
stored job, instead of using job flag to specialize it.
The input jobId in the startJob method is sufficient to tell it's a stored
job or not.
Moreover, I think the lack of full lifecyle support is a design bug which can
cause memory leaks. Therefore, I marked this issue a "bug". Actually this
issue contains three subtasks:
-- Full lifecyle support (bug)
-- Avoid serialization (improvement)
-- Interface cleanup (improvement)
> Improve Asterix Capabilities to pre-distribute jobs
> ---------------------------------------------------
>
> Key: ASTERIXDB-1747
> URL: https://issues.apache.org/jira/browse/ASTERIXDB-1747
> Project: Apache AsterixDB
> Issue Type: Bug
> Reporter: Steven Jacobs
> Assignee: Steven Jacobs
>
> Currently, we are able to store the ActivityClusterGraph for a job at the NC
> level. We need to add the following:
> Don't bother passing the bytes around when the graph is already at the NC.
> Don't serialize the bytes at the CC side when they aren't needed at the NC.
> Remove the ActivityClusterGraph from the NC when the job is no longer needed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
