HBase Metrics never expired in ganglia leads to storage leaking

[柳刘 <cd...@jd.com>]

Did somebody else encountered this problem? Can we HBase fix it? thanks

Yang He shared an issue with you
----------------------------------
can somebody review this ? thranks. this issue is significant for hbase,since region metrics is invalidate when region splited;
> In the hadoop metrics framework,The default value of dmax is 0, which means gmond process will never delete this metirc although it is disappeared.
> ---------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: HADOOP-10161
> URL: https://issues.apache.org/jira/browse/HADOOP-10161
> Project: Hadoop Common
> Issue Type: Improvement
> Components: metrics
> Affects Versions: 2.2.0
> Reporter: Yang He
> Attachments: HADOOP-10161_0_20131211.patch, hadoop-metrics.properties, hadoop-metrics2.properties
>
>
> The property of dmax in ganglia configration is metric dead time , when one metrics is disappeared ,so no more value of the metric will be emtic to the gmond,after 'dmax' seconds,then gmond will destroy the metric in memery . in hadoop metrics framework ,the default value is 0,which means the gmond will never destroy the metric althought the metric is disappeared ,the gmetad daemon also does not delete the rrdtool file ,and there does not have a method to configure the defualt value of dmax for all metrics . 
--
This message was sent by Atlassian JIRA
(v6.1.4#6159)





柳刘