[
https://issues.apache.org/jira/browse/AMBARI-25195?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
ASF GitHub Bot updated AMBARI-25195:
------------------------------------
Labels: pull-request-available (was: )
> Add option to graceful rolling restart for HBase RegionServers
> --------------------------------------------------------------
>
> Key: AMBARI-25195
> URL: https://issues.apache.org/jira/browse/AMBARI-25195
> Project: Ambari
> Issue Type: Improvement
> Components: ambari-server, ambari-web
> Affects Versions: 2.7.3
> Reporter: Eungsop Yoo
> Priority: Minor
> Labels: pull-request-available
> Attachments: rr-rs.png
>
>
> Currently Ambari supports rolling restart for HBase RegionServers, just like
> for DataNodes and NodeManagers. But the feature has not been optimized for
> HBase. I think it has mainly two shortcomings.
> # WAL Splitting
> #- The current way of rolling start causes WAL splitting. And the WAL
> splitting causes some down times for the regions related to this process.
> # Region Churning
> #- After rolling restarting, the assignment of regions to RegionServers is
> drastically different from before. It causes some performance degradation.
> So I suggest improving the feature of rolling starting for HBase
> RegionServers to be more graceful. The specific process is described below.
> # Disable HBase Balancer
> # For RS in RegionServers
> ## Put RS into Drainig Mode
> ## Unload all regions of the RS to the others
> ## Stop RS
> ## Remove RS from Drainig Mode
> ## Start RS
> ## Reload the regions from the others
> # Enable HBase Balancer
> And the view of rolling restart is modified just like below.
> !rr-rs.png!
> The PR of this issue will be created soon.
> How do you think about this suggestion?
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
