[
https://issues.apache.org/jira/browse/HIVE-28662?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Zhihua Deng updated HIVE-28662:
-------------------------------
Description:
In the Hive ACID world, we must enable the metastore.compactor.initiator.on and
metastore.compactor.cleaner.on on HMS to trigger the table compaction
automatically. In a real warehouse, we might have multiple HMS instances
behind, or even deploy for different purposes, running the tasks like these on
all instances could be a waste of resources, make the HMS scale out not so
easily.
The HIVE-26509 introduces the dynamic leader election. Compared to the old one,
the administrator doesn't need to know the deployment beforehand, and a shared
properties can be used across all the HMS instances, simple yet effective.
was:
In the Hive ACID world, we must enable the metastore.compactor.initiator.on and
metastore.compactor.cleaner.on on HMS to trigger the table compaction
automatically. In a real warehouse, we might have multiple HMS instances
behind, or even deploy for different purposes, running the tasks like these on
all these instances could be a waste of resources, make the HMS scale out not
so easily.
The HIVE-26509 introduces the dynamic leader election. Compared to the old one,
the administrator doesn't need to know the deployment beforehand, and a shared
properties can be used across all the HMS instances, simple yet effective.
> Enable the dynamic leader election for HMS
> ------------------------------------------
>
> Key: HIVE-28662
> URL: https://issues.apache.org/jira/browse/HIVE-28662
> Project: Hive
> Issue Type: Improvement
> Security Level: Public(Viewable by anyone)
> Reporter: Zhihua Deng
> Assignee: Zhihua Deng
> Priority: Major
>
> In the Hive ACID world, we must enable the metastore.compactor.initiator.on
> and metastore.compactor.cleaner.on on HMS to trigger the table compaction
> automatically. In a real warehouse, we might have multiple HMS instances
> behind, or even deploy for different purposes, running the tasks like these
> on all instances could be a waste of resources, make the HMS scale out not so
> easily.
> The HIVE-26509 introduces the dynamic leader election. Compared to the old
> one, the administrator doesn't need to know the deployment beforehand, and a
> shared properties can be used across all the HMS instances, simple yet
> effective.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
