[
https://issues.apache.org/jira/browse/HIVE-20814?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18054672#comment-18054672
]
Dayakar M commented on HIVE-20814:
----------------------------------
Hi [~zabetak],
When *‘{_}hive.allow.udf.load.on.demand{_}’* configuration (default
{*}_false_{*}) is enabled then the udf/function used in the query if not
available in the FunctionRegistry then it will try to fetch it from HMS, if
found then will be registered to the FunctionRegistry. So the functionality is
already available in the HS2.
But if the udf/function is updated or dropped then there is no way to notify to
the HS2, in this case the old version of UDF available in the FunctionRegistry
will work so user has to explicitly call *RELOAD FUNCTION;* to reload all the
UDFs from HMS to the FunctionRegistry.
As update/drop udf feature will rarely used one in that case user can
explicitly call RELOAD FUNCTION in other HS2 instances.
So as per my understanding we no need to do anything for this Jira as the
requested feature is already available.
Please let me know if you have any other suggestions.
Thanks,
Dayakar
> New create function between HiveServer2 is not synchronized
> ------------------------------------------------------------
>
> Key: HIVE-20814
> URL: https://issues.apache.org/jira/browse/HIVE-20814
> Project: Hive
> Issue Type: Improvement
> Components: Beeline
> Affects Versions: 1.1.0
> Reporter: leozhang
> Priority: Major
> Attachments: image-2018-10-26-10-23-48-101.png,
> image-2018-10-26-10-24-16-669.png, image-2018-10-26-10-24-54-904.png,
> image-2018-10-26-10-26-00-591.png, image-2018-10-26-10-27-32-291.png
>
>
> I am using CDH open source version 5.15.0, where the Hive version is 1.1.0.
> I have 3 Metastore services and 3 HiveServer2 services in the cluster,
> Metastore and HiveServer2 don't currently have HA turned on. The cluster
> opens the Sentry service and configures the Hive Auxiliary JARs Directory
> attribute.
> Now I am having a problem, I am using beeline to connect to HiveServer2 on
> node 1 to create a function that is successful and can be queried normally.
> But if I now connect to HiveServer2 of node 2 via Beeline, I can't see the
> function I just created on node 1 through _show function_. I have to restart
> HiveServer2 or use _reload function_ command on node 2 to see the function
> created on node 1 just now.
> Is there a way to implement the function that I can create on any node? I
> don't know the reason and solution for this problem, I hope I can get help,
> thank you!
>
> exp:
> connect dn1.test.com:10000 to create function _zzytest_trans_ , use _show
> functions_ I can see it
> !image-2018-10-26-10-24-16-669.png!
> !image-2018-10-26-10-24-54-904.png!
>
> At this point I used beeline to connect to dn4.test.com:10000 and use the
> show functions command to view the function and I can't find the
> zzytest_trans function. I have to restart HiveServer2 or use _reload
> function_ command on the dn4.test.com node to see the zzytest_trans function.
> !image-2018-10-26-10-26-00-591.png!
> !image-2018-10-26-10-27-32-291.png!
>
>
>
>
>
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
