You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@ambari.apache.org by Cyril Scetbon <cy...@free.fr> on 2016/04/17 04:16:53 UTC

Cluster CPU widget is wrong

Hi,

I'm using HDP 2.4 to run Yarn applications. When I look at my current NodeManagers CPU, the consumption is low. However on Ambari interface, the default "Cluster CPU“ widget displays a high CPU (91%). When I check the expression used, it uses cpu_idle._sum (see http://picpaste.com/Screen_Shot_2016-04-16_at_22.13.43-gS4XsIGb.png <http://picpaste.com/Screen_Shot_2016-04-16_at_22.13.43-gS4XsIGb.png>) which is a metric that I can't find if I try to create a new widget http://picpaste.com/Screen_Shot_2016-04-16_at_22.09.56-DJK3v4Wd.png <http://picpaste.com/Screen_Shot_2016-04-16_at_22.09.56-DJK3v4Wd.png> 
Can it be the reason why the CPU consumption displayed is totally wrong ? Is it a known bug ? 

Thanks

Re: Cluster CPU widget is wrong

Posted by Cyril Scetbon <cy...@free.fr>.
Sure.

https://issues.apache.org/jira/browse/AMBARI-15936

> On Apr 16, 2016, at 23:09, Siddharth Wagle <sw...@hortonworks.com> wrote:
> 
> Hi Cyril,
> 
> Could you kindly open an Apache Jira with details like:
> 
> - Ambari version
> - Number of cores
> - OS for the host
> 
> Note: AMS supports (._avg, ._sum, ._max, ._min) as the aggregate functions for any metrics requested.
> 
> We have made numerous fixes recently to AMS including new Grafana UI scheduled for 2.2.2 release. I haven't found any recent known issues for CPU charts, following are relevant ones that are known:
> 
> https://issues.apache.org/jira/browse/AMBARI-12519 <https://issues.apache.org/jira/browse/AMBARI-12519>
> https://issues.apache.org/jira/browse/AMBARI-12825 <https://issues.apache.org/jira/browse/AMBARI-12825>
> https://issues.apache.org/jira/browse/AMBARI-8662 <https://issues.apache.org/jira/browse/AMBARI-8662>
> 
> 
> BR,
> Sid
> 
> From: Cyril Scetbon <cy...@free.fr>
> Sent: Saturday, April 16, 2016 7:16 PM
> To: user@ambari.apache.org
> Subject: Cluster CPU widget is wrong
>  
> Hi,
> 
> I'm using HDP 2.4 to run Yarn applications. When I look at my current NodeManagers CPU, the consumption is low. However on Ambari interface, the default "Cluster CPU“ widget displays a high CPU (91%). When I check the expression used, it uses cpu_idle._sum (see http://picpaste.com/Screen_Shot_2016-04-16_at_22.13.43-gS4XsIGb.png <http://picpaste.com/Screen_Shot_2016-04-16_at_22.13.43-gS4XsIGb.png>) which is a metric that I can't find if I try to create a new widget http://picpaste.com/Screen_Shot_2016-04-16_at_22.09.56-DJK3v4Wd.png <http://picpaste.com/Screen_Shot_2016-04-16_at_22.09.56-DJK3v4Wd.png> 
> Can it be the reason why the CPU consumption displayed is totally wrong ? Is it a known bug ? 
> 
> Thanks
> 
> 
>

Re: Cluster CPU widget is wrong

Posted by Siddharth Wagle <sw...@hortonworks.com>.
Hi Cyril,


Could you kindly open an Apache Jira with details like:


- Ambari version

- Number of cores

- OS for the host


Note: AMS supports (._avg, ._sum, ._max, ._min) as the aggregate functions for any metrics requested.


We have made numerous fixes recently to AMS including new Grafana UI scheduled for 2.2.2 release. I haven't found any recent known issues for CPU charts, following are relevant ones that are known:


https://issues.apache.org/jira/browse/AMBARI-12519

https://issues.apache.org/jira/browse/AMBARI-12825

https://issues.apache.org/jira/browse/AMBARI-8662



BR,

Sid


________________________________
From: Cyril Scetbon <cy...@free.fr>
Sent: Saturday, April 16, 2016 7:16 PM
To: user@ambari.apache.org
Subject: Cluster CPU widget is wrong

Hi,

I'm using HDP 2.4 to run Yarn applications. When I look at my current NodeManagers CPU, the consumption is low. However on Ambari interface, the default "Cluster CPU" widget displays a high CPU (91%). When I check the expression used, it uses cpu_idle._sum (see http://picpaste.com/Screen_Shot_2016-04-16_at_22.13.43-gS4XsIGb.png) which is a metric that I can't find if I try to create a new widget http://picpaste.com/Screen_Shot_2016-04-16_at_22.09.56-DJK3v4Wd.png
Can it be the reason why the CPU consumption displayed is totally wrong ? Is it a known bug ?

Thanks