ambari-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Srimanth Gunturi (JIRA)" <j...@apache.org>
Subject [jira] [Resolved] (AMBARI-10837) HDFS Review: Multiple recommendation API updates for HDFS configs
Date Wed, 29 Apr 2015 23:47:05 GMT

     [ https://issues.apache.org/jira/browse/AMBARI-10837?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Srimanth Gunturi resolved AMBARI-10837.
---------------------------------------
    Resolution: Fixed

+1 for patch. Committed to trunk.

> HDFS Review: Multiple recommendation API updates for HDFS configs
> -----------------------------------------------------------------
>
>                 Key: AMBARI-10837
>                 URL: https://issues.apache.org/jira/browse/AMBARI-10837
>             Project: Ambari
>          Issue Type: Bug
>          Components: ambari-server
>    Affects Versions: 2.1.0
>            Reporter: Myroslav Papirkovskyy
>            Assignee: Myroslav Papirkovskyy
>            Priority: Critical
>             Fix For: 2.1.0
>
>
> HDFS configs review was done and the configs spreadsheet has been updated with various
changes where the following must be fixed.
> * Below configs are to be marked as {{depends_on}} {{namenode_heapsize}}, and their value
should be derived from it (basically ignore document value for below configs). Whenever {{namenode_heapsize}}
changes in UI, below config values should also be updated
> ** namenode_opt_newsize (hadoop-env.sh) = {{namenode_heapsize/8}}
> ** namenode_opt_maxnewsize (hadoop-env.sh) = {{namenode_heapsize/8}}
> * {{dfs.namenode.safemode.threshold-pct}}
> ** minimum = 0.990f
> ** maximum = 1.000f
> ** default = 0.999f
> ** increment-step = 0.001f
> * {{dfs.datanode.failed.volumes.tolerated}} should be {{depends_on}} {{dfs.datanode.data.dir}}.
So if a user adds additional folder in {{dfs.datanode.data.dir}}, then the *value and maximum*
of {{dfs.datanode.failed.volumes.tolerated}} should change accordingly.
> * {{namenode_heapsize}} calculations should take into account host memory limits. Namenode_heapsize
should be the {{host-memory - os-reserverd-memory}}. Also, if there are any other master-components
on the same host, then it should be halved (namenode_heapsize/2). 



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

Mime
View raw message