[
https://issues.apache.org/jira/browse/AMBARI-14882?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Siddharth Wagle updated AMBARI-14882:
-------------------------------------
Attachment: AMBARI-14882.patch
> AMS aggregates Counter values as average over the timeseries (and other
> issues)
> -------------------------------------------------------------------------------
>
> Key: AMBARI-14882
> URL: https://issues.apache.org/jira/browse/AMBARI-14882
> Project: Ambari
> Issue Type: Task
> Components: ambari-metrics
> Affects Versions: 2.0.0
> Reporter: Siddharth Wagle
> Assignee: Siddharth Wagle
> Priority: Critical
> Fix For: 2.2.2
>
> Attachments: AMBARI-14882.patch
>
>
> Current behavior:
> We store the difference between two points if the metric is a counter without
> dividing by the time difference (Assumes fixed sampling frequency, 10 seconds
> for Hadoop daemons). This is similar to the Ganglia sink implementation.
> Expected behavior:
> - Store raw data with type = COUNTER and provide aggregate function called
> ._rate to calculate rate on demand.
> - We need to make sure type attribute on the METRIC_RECORD data capture
> COUNTER / GAUGE. The HadoopMetricsSink needs to change to send this value.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
