[ 
https://issues.apache.org/jira/browse/AMBARI-1442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13581298#comment-13581298
 ] 

Tom Beerbower commented on AMBARI-1442:
---------------------------------------

The decision to return temporal vs. point in time metrics is based on the 
query.  If a start time/end time (seconds since epoch) /resolution are provided 
as part of the query then we return temporal results.  The default is to return 
point in time metrics.

The "pointInTime" variable in the json files indicates whether or not point in 
time values can be obtained for that metric.  The "temporal" variable indicates 
whether or not temporal values can be obtained.  None of the metrics in the JMX 
file should be marked as "temporal" : true.  All of the metrics in the Ganglia 
file should be marked as "temporal" : true.  Some of the metrics in the Ganglia 
file are marked as "temporal" : true and "pointInTime" : true.  It makes sense 
to get point in time values from Ganglia only when the metric is not available 
through JMX and we want to return the last data point as the the point in time 
value.  For example, ganglia_properties.json contains ...

{code}
      "metrics/cpu/cpu_speed":{
        "metric":"cpu_speed",
        "pointInTime":true,
        "temporal":true
      },
      "metrics/cpu/cpu_idle":{
        "metric":"cpu_idle",
        "pointInTime":true,
        "temporal":true
      },
{code} 
 
If a request is made and no specific properties/metrics are specified ( same as 
fields=\* ) then each provider that is consulted will set all of the properties 
available to it on the resource being returned.  For example, the JMX property 
provider will make a JMX request to get the JMX data for the requested 
component.  It will populate the component resource to be returned based on the 
data that it gets back, the properties specified in the jmx_properties.json 
file and the properties specified in the request ( none in the case of 
fields=\* ).


                
> Host Component metrics always showing zero values
> -------------------------------------------------
>
>                 Key: AMBARI-1442
>                 URL: https://issues.apache.org/jira/browse/AMBARI-1442
>             Project: Ambari
>          Issue Type: Bug
>            Reporter: Tom Beerbower
>            Assignee: Tom Beerbower
>             Fix For: 1.2.2
>
>         Attachments: AMBARI-1442.patch
>
>
> In spite of the fact that some hadoop jobs that were performed, all fields of 
> shuffle metrics have 0.0 values.
> Used the following request to get shuffle metrics:
> {code}
> api/v1/clusters/clusterName/hosts/hostName/host_components/TASKTRACKER?fields=metrics/mapred/shuffleOutput
> {code}
> Result:
> {code}
> "metrics" : {
> "mapred" : {
> "shuffleOutput" :
> { "shuffle_success_outputs" : 0, "shuffle_handler_busy_percent" : 0, 
> "shuffle_output_bytes" : 0, "shuffle_failed_outputs" : 0, 
> "shuffle_exceptions_caught" : 0 }
> }
> }
> {code}

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Reply via email to