You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ambari.apache.org by "Aravindan Vijayan (JIRA)" <ji...@apache.org> on 2018/05/22 21:41:00 UTC

[jira] [Updated] (AMBARI-23008) AMS: SUM downsampling function works incorrectly

     [ https://issues.apache.org/jira/browse/AMBARI-23008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Aravindan Vijayan updated AMBARI-23008:
---------------------------------------
    Fix Version/s: 2.6.2

> AMS: SUM downsampling function works incorrectly
> ------------------------------------------------
>
>                 Key: AMBARI-23008
>                 URL: https://issues.apache.org/jira/browse/AMBARI-23008
>             Project: Ambari
>          Issue Type: Bug
>          Components: ambari-metrics
>    Affects Versions: 2.6.1
>            Reporter: Aravindan Vijayan
>            Assignee: Aravindan Vijayan
>            Priority: Critical
>              Labels: pull-request-available
>             Fix For: 2.6.2
>
>          Time Spent: 50m
>  Remaining Estimate: 0h
>
> AMS, by default does not sum up across time dimension. However, there are certain metric cases where this is useful.
> *Aggregation Scenario - How it should work*
> Let's say in Host H1 - the metric values for 'topology.streamline-2-test-clone.3-KAFKA.TASK1.--emit-count.STREAM1' are 1,1,0,1,2 in time 10:01 to 10:05 (1 minute values) respectively
> And, in Host H2 - the metric values for the same metric are 1,0,2,2,1 respectively
> AMS will first aggregate across hosts. 'Sum' aggregation -> 2,1,2,3,3 (10:01 - 10:05) and then across time (downsampling) for larger windows. In time based downsampling, the 'sum' value will be 11 (2 + 1 + 2 + 3 + 3) at 10:05.
> Also, such metrics should not be interpolated, because the injected values will cause wrong summed up values.
> * Work*
> A new custom downsampler was added which implements the above logic. 



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)