[
https://issues.apache.org/jira/browse/HIVE-23033?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17061637#comment-17061637
]
Naveen Gangam commented on HIVE-23033:
--------------------------------------
[~dlavati] Couple of thoughts and I am not sure which way to go just yet. To my
knowledge, I dont think we ever retro-fixed an issue in a released product. For
example, 3.1 is released and wouldn't be able to fix the issue in the schema
script for 3.1 (hive-schema-3.1.0....sql file) because there is no means to get
this fix to users already using 3.1. So this fix would go instead on branch-3,
the dev branch for next release on 3.x.
* So is there a reason we are making this fix to the 3.1 and 3.2 versions of
the script? If not, I think we should fix it in 4.0 version on the master and
to the 3.2 version of script on branch-3.
* If there is a reason to retro-fix old scripts, then we should also include
the fix to hive-schema-3.0.0.sql as well to make it consistent with other
databases. Looks like this fix was introduced for hive-3.0.
Also for users already on say 3.1, when they upgrade to say 4.0, how will they
pick up this change? I would hope that the NOTIFICATION_SEQUENCE table has some
data based on their usage on 3.1. But if not, such users will have
un-initialized table as well. Should we address this via a change to the
upgrade script to 4.0?
Thanks
> MSSQL metastore schema init script doesn't initialize NOTIFICATION_SEQUENCE
> ---------------------------------------------------------------------------
>
> Key: HIVE-23033
> URL: https://issues.apache.org/jira/browse/HIVE-23033
> Project: Hive
> Issue Type: Bug
> Affects Versions: 3.0.0, 3.1.0, 3.1.1, 3.1.2
> Reporter: David Lavati
> Assignee: David Lavati
> Priority: Major
> Fix For: 4.0.0, 3.1.3
>
> Attachments: HIVE-23033.1.patch
>
>
> * The inital value for this table in the schema scripts was removed in
> HIVE-17566:
> https://github.com/apache/hive/commit/32b7abac961ca3879d23b074357f211fc7c49131#diff-3d1a4bae0d5d53c8e4ea79951ebf5eceL598
> * This was fixed in a number of scripts in HIVE-18781, but not for mssql:
> https://github.com/apache/hive/commit/59483bca262880d3e7ef1b873d3c21176e9294cb#diff-4f43efd5a45cc362cb138287d90dbf82
> * This is as is since then
> When using the schematool, the table gets initialized by other means.
> This could be backported to all active branches for 3.x as well.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
