hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Junping Du (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HDFS-4884) Provide a way to enforce new Block Placement Policy to existing data after switch from old one
Date Thu, 06 Jun 2013 23:22:20 GMT

     [ https://issues.apache.org/jira/browse/HDFS-4884?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Junping Du updated HDFS-4884:
-----------------------------

    Description: The BlockPlacementPolicy (BPP) is extensible and already multiple instances
in system: BlockPlacementPolicyDefault, BlockPlacementPolicyWithNodeGroup, BlockPlacementPolicyRAID,
etc. When cluster is switched from one BPP to another BPP, HDFS will not check if block locations
conform the new BPP. Now, you can manually fsck on specific directory to identify replica
placement policy is violated for some blocks but have no way to fix it so far. We should provide
way to fix it.  (was: The BlockPlacementPolicy (BPP) is extensible and already multiple instances
in system: BlockPlacementPolicyDefault, BlockPlacementPolicyWithNodeGroup, BlockPlacementPolicyRAID,
etc. When cluster is switched from one BPP to another BPP, HDFS will not check if block locations
conform the new BPP. Now, you can manually fsck on specific directory to identify replica
placement policy is violated for some blocks but have no way to fix it so far. We should provide
way to fix it. Also, in the long term, we should allow multiple BPPs co-existing on the same
HDFS cluster for different purpose of data, i.e. we should allow some old, infrequently accessed
data/directory to under RAID policy so that can save space. )
    
> Provide a way to enforce new Block Placement Policy to existing data after switch from
old one
> ----------------------------------------------------------------------------------------------
>
>                 Key: HDFS-4884
>                 URL: https://issues.apache.org/jira/browse/HDFS-4884
>             Project: Hadoop HDFS
>          Issue Type: Improvement
>    Affects Versions: 1.2.0, 2.0.4-alpha
>            Reporter: Junping Du
>            Assignee: Junping Du
>
> The BlockPlacementPolicy (BPP) is extensible and already multiple instances in system:
BlockPlacementPolicyDefault, BlockPlacementPolicyWithNodeGroup, BlockPlacementPolicyRAID,
etc. When cluster is switched from one BPP to another BPP, HDFS will not check if block locations
conform the new BPP. Now, you can manually fsck on specific directory to identify replica
placement policy is violated for some blocks but have no way to fix it so far. We should provide
way to fix it.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message