hadoop-common-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Kamil Gorlo (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HADOOP-11360) GraphiteSink reports data with wrong timestamp
Date Mon, 08 Dec 2014 13:09:12 GMT

     [ https://issues.apache.org/jira/browse/HADOOP-11360?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Kamil Gorlo updated HADOOP-11360:
---------------------------------
    Description: 
I've tried to use GraphiteSink with metrics2 system, but it looks that timestamp sent to Graphite
is refreshed rarely (about every 2 minutes approx.) no mather how small period is set.

Here is my configuration:

*.sink.graphite.server_host=graphite-relay.host
*.sink.graphite.server_port=2013
*.sink.graphite.metrics_prefix=graphite.warehouse-data-1
*.period=10
nodemanager.sink.graphite.class=org.apache.hadoop.metrics2.sink.GraphiteSink

And here is dumped network traffic to graphite-relay.host (only selected lines, every line
appears in 10 seconds as period suggests):

graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041472
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041472
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041472
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
3 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
4 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
2 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
3 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
2 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
2 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
1 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
1 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041728
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041728


As you can see, AllocatedContainers value is refreshed every 10 seconds, but timestamp is
not.

It looks that the problem is level above (in classes providing MetricsRecord - because timestamp
value is taken from MetricsRecord object provided in argument to putMetrics method in Sink
implementation) which implies that every sink will have the same problem. Maybe I misconfigured
something?

  was:
I've tried to use GraphiteSink with metrics2 system, but it looks that timestamp sent to Graphite
is refreshed rarely (about every 2 minutes approx.) no mather how small period is set.

Here is my configuration:

*.sink.graphite.server_host=graphite-relay.host
*.sink.graphite.server_port=2013
*.sink.graphite.metrics_prefix=graphite.warehouse-data-1
*.period=10
nodemanager.sink.graphite.class=org.apache.hadoop.metrics2.sink.GraphiteSink

And here is dumped network traffic to graphite-relay.host (only select lines, every line appears
in 10 seconds as period suggests):

graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041472
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041472
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041472
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
3 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
4 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
2 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
3 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
2 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
2 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
1 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
1 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041728
graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041728


As you can see, AllocatedContainers value is refreshed every 10 seconds, but timestamp is
not.


> GraphiteSink reports data with wrong timestamp
> ----------------------------------------------
>
>                 Key: HADOOP-11360
>                 URL: https://issues.apache.org/jira/browse/HADOOP-11360
>             Project: Hadoop Common
>          Issue Type: Bug
>          Components: metrics
>            Reporter: Kamil Gorlo
>
> I've tried to use GraphiteSink with metrics2 system, but it looks that timestamp sent
to Graphite is refreshed rarely (about every 2 minutes approx.) no mather how small period
is set.
> Here is my configuration:
> *.sink.graphite.server_host=graphite-relay.host
> *.sink.graphite.server_port=2013
> *.sink.graphite.metrics_prefix=graphite.warehouse-data-1
> *.period=10
> nodemanager.sink.graphite.class=org.apache.hadoop.metrics2.sink.GraphiteSink
> And here is dumped network traffic to graphite-relay.host (only selected lines, every
line appears in 10 seconds as period suggests):
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041472
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041472
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041472
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
3 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
4 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
2 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
3 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
2 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
2 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
1 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
1 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041600
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041728
> graphite.warehouse-data-1.yarn.NodeManagerMetrics.Context=yarn.Hostname=warehouse-data-1.AllocatedContainers
0 1418041728
> As you can see, AllocatedContainers value is refreshed every 10 seconds, but timestamp
is not.
> It looks that the problem is level above (in classes providing MetricsRecord - because
timestamp value is taken from MetricsRecord object provided in argument to putMetrics method
in Sink implementation) which implies that every sink will have the same problem. Maybe I
misconfigured something?



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message