[
https://issues.apache.org/jira/browse/HADOOP-16266?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16826289#comment-16826289
]
Erik Krogen commented on HADOOP-16266:
--------------------------------------
v003 patch is looking really good! The remaining things are mostly pretty
minor, besides fixing the test. I think leaving the change to make the metrics
time unit to be configurable for a later JIRA is a good move. But this sets up
the framework for it to be done.
* In the {{UnsupportedOperationException}} we throw within
{{DecayRpcScheduler}}, can we provide a helpful message like "This method has
been deprecated, please use the other addResponseTime"
* In {{RpcScheduler}}, can we mark the old {{addResponseTime}} as
{{@Deprecated}} and give a pointer to the new version.
* Can we make the {{default}} implementation of {{addResponseTime}} actually a
useful implementation:
{code}
default void addResponseTime(String name, Schedulable obj,
ProcessingDetails details) {
addResponseTime(name, obj.getPriorityLevel(),
(int) details.get(ProcessingDetails.Timing.QUEUE,
TimeUnit.MILLISECONDS),
(int) details.get(ProcessingDetails.Timing.PROCESSING,
TimeUnit.MILLISECONDS));
}
{code}
* In the description for {{RpcLockWaitTimeAvgTime}} within {{Metrics.md}}, you
mention "the lock", which is the case for HDFS, but other uses (such as the RM)
may have multiple locks. We should make it more like "waiting for lock
acquisition"
* Can we add {{@InterfaceAudience.Private}} to {{ProcessingDetails}}
* In {{doPurge}}, can we use {{TimeUnit.NANOSECONDS.convert()}} instead of
dividing by a constant
> Add more fine-grained processing time metrics to the RPC layer
> --------------------------------------------------------------
>
> Key: HADOOP-16266
> URL: https://issues.apache.org/jira/browse/HADOOP-16266
> Project: Hadoop Common
> Issue Type: Improvement
> Components: ipc
> Reporter: Christopher Gregorian
> Assignee: Christopher Gregorian
> Priority: Minor
> Labels: rpc
> Attachments: HADOOP-16266.001.patch, HADOOP-16266.002.patch,
> HADOOP-16266.003.patch
>
>
> Splitting off of HDFS-14403 to track the first part: introduces more
> fine-grained measuring of how a call's processing time is split up.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
