hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ashu Pachauri (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HBASE-17202) Master takes a long time to initialize
Date Wed, 30 Nov 2016 02:19:58 GMT

     [ https://issues.apache.org/jira/browse/HBASE-17202?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Ashu Pachauri updated HBASE-17202:
----------------------------------
    Description: 
On master failover or a full cluster restart, we create the RegionLocationFinder cache by
calculating block locations for every region; this cache is used by the balancer for calculating
region locality. This greatly increases the time it takes for the assignment manager to start
and affects region availability in scenarios of full cluster restart.

We should initialize assignment manager with the bare minimum information and asynchronously
load the block location cache.

  was:
On master failover or a full cluster restart, we create the RegionLocationFinder cache by
calculating block locations for every region used by the balancer. This greatly increases
the time it takes for the assignment manager to start and affects region availability in scenarios
of full cluster restart.

We should initialize assignment manager with the bare minimum information and asynchronously
load the block location cache.


> Master takes a long time to initialize
> --------------------------------------
>
>                 Key: HBASE-17202
>                 URL: https://issues.apache.org/jira/browse/HBASE-17202
>             Project: HBase
>          Issue Type: Improvement
>          Components: Balancer, master
>            Reporter: Ashu Pachauri
>            Assignee: Ashu Pachauri
>
> On master failover or a full cluster restart, we create the RegionLocationFinder cache
by calculating block locations for every region; this cache is used by the balancer for calculating
region locality. This greatly increases the time it takes for the assignment manager to start
and affects region availability in scenarios of full cluster restart.
> We should initialize assignment manager with the bare minimum information and asynchronously
load the block location cache.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message