You are viewing a plain text version of this content. The canonical link for it is here.
Posted to hdfs-dev@hadoop.apache.org by "Guo Ruijing (JIRA)" <ji...@apache.org> on 2014/05/28 11:49:05 UTC

[jira] [Created] (HDFS-6454) Configurable Block Placement Policy

Guo Ruijing created HDFS-6454:
---------------------------------

             Summary: Configurable Block Placement Policy
                 Key: HDFS-6454
                 URL: https://issues.apache.org/jira/browse/HDFS-6454
             Project: Hadoop HDFS
          Issue Type: Improvement
          Components: namenode
            Reporter: Guo Ruijing


In existing implementation,  block choose priority is localhost/remote rack/local rack/ramdon.
in BlockPlacementPolicyDefault, network topology is /rack/host.
In BlockPlacementPolicyWithNodeGroup, network topology is /rack/nodegroup/host.

This JIRA is to propose block choose priority can be configurable as:

<property>
  <name>dfs.block.replicator.priority</name>
  <value>0, 2, 1, *</value>
  <description>
       default network topology is /level2/level1
       nodegroup network topology is /level3/level2/level1. choose priority can be 0(localhost), 3(remote rack), 2(local rack), *(any host)
  </description>
</property>

Another example (one VM includes serveral dockers/containers) so network topology can be /rack/nodegroup/container/host. in this case, block replicator priority can be
0(localhost), 4(remote rack), 3(local rack), *(any host)



--
This message was sent by Atlassian JIRA
(v6.2#6252)