[
https://issues.apache.org/jira/browse/HADOOP-10673?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ming Ma updated HADOOP-10673:
-----------------------------
Attachment: HADOOP-10673-6.patch
Thanks, Tsuyoshi. [~lohit] asked a good question how the increase of the number
of metrics will impact metrics subsystem. The most common one might be
StandbyException, which can be thrown regardless of methods. Other exceptions
are likely to be specific to certain methods. Based on the current number of
metrics, JMX size; number of methods, a very rough estimate can put the
increase to around 10-15% for NN.
So to be safe, we can aggregate by exception class name instead; that will
still provide useful information (from the exception name, we can find the
possible methods).
Here is the updated patch.
> Update rpc metrics when the call throws an exception
> ----------------------------------------------------
>
> Key: HADOOP-10673
> URL: https://issues.apache.org/jira/browse/HADOOP-10673
> Project: Hadoop Common
> Issue Type: Bug
> Reporter: Ming Ma
> Assignee: Ming Ma
> Attachments: HADOOP-10673-2.patch, HADOOP-10673-4.patch,
> HADOOP-10673-5.patch, HADOOP-10673-6.patch, HADOOP-10673.patch
>
>
> Currently RPC metrics isn't updated when the call throws an exception. We can
> either update the existing metrics or have a new set of metrics in the case
> of exception.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
