[
https://issues.apache.org/jira/browse/HDFS-13854?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16594456#comment-16594456
]
yanghuafeng commented on HDFS-13854:
------------------------------------
I have checked the time in the NameNodeMetrics, FederationMetrics,
NamenodeBeanMetrics, StateStoreMetrics. Except the NameNodeMetrics, others
don't use the time or use the default value instead.
NameNodeMetrics:
{code:java}
public void addGenerateEDEKTime(long latency)
public void addWarmUpEDEKTime(long latency)
public void addResourceCheckTime(long latency)
public void addEditLogTailTime(long elapsed)
public void addEditLogFetchTime(long elapsed)
public void addNumEditLogLoaded(long loaded)
public void addEditLogTailInterval(long elapsed)
{code}
all of the methods use the Time.monotonicNow() to compute the time metric.
> RBF: The ProcessingAvgTime and ProxyAvgTime should display by JMX with ms
> unit.
> -------------------------------------------------------------------------------
>
> Key: HDFS-13854
> URL: https://issues.apache.org/jira/browse/HDFS-13854
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: federation, hdfs
> Affects Versions: 2.9.0, 3.0.0, 3.1.0
> Reporter: yanghuafeng
> Assignee: yanghuafeng
> Priority: Major
> Attachments: HDFS-13854.001.patch, HDFS-13854.002.patch,
> ganglia_jmx_compare1.jpg, ganglia_jmx_compare2.jpg
>
>
> In the FederationRPCMetrics, proxy time and processing time should be exposed
> to the jmx or ganglia with ms units. Although the method toMS() exists, we
> cannot get the correct proxy time and processing time by jmx and ganglia.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
