[
https://issues.apache.org/jira/browse/HIVE-17904?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sergey Shelukhin updated HIVE-17904:
------------------------------------
Attachment: HIVE-17904.patch
The patch. Actually, onDestroy was doing an entirely wrong thing - ID->session
mapping should be maintained for open sessions, it's unrelated to registry.
Registry callback is now propagated to the session and everything handles the
changing endpoint info.
No need to have a WM event for that, WM can just get version from the time of
the error and compare with current one to discard errors for old versions.
> handle internal Tez AM restart in registry and WM
> -------------------------------------------------
>
> Key: HIVE-17904
> URL: https://issues.apache.org/jira/browse/HIVE-17904
> Project: Hive
> Issue Type: Sub-task
> Reporter: Sergey Shelukhin
> Assignee: Sergey Shelukhin
> Attachments: HIVE-17904.patch
>
>
> After the plan update patch is committed. The current code doesn't account
> very well for it; registry may have races, and an event needs to be added to
> WM when some AM resets, at least to make sure we discard the update errors
> that pertain to the old AM.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
