[
https://issues.apache.org/jira/browse/HADOOP-9420?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13863952#comment-13863952
]
Andrew Wang commented on HADOOP-9420:
-------------------------------------
Hi Liang Xie, thanks for picking this up. Feel free to assign to yourself.
Patch looks basically fine, just a few comments:
* I think a test case isn't too hard here, since we already have tests for
metrics. Would be appreciated, even though it is a small change.
* Can we make the percentiles configurable the same way as in HDFS with a
config key? I've been idly toying with the idea of a lower-overhead percentiles
implementation and defaulting them on everywhere, but until then, it's probably
best to keep it configurable and defaulting to off (especially since affects
every RPC call).
> Add percentile or max metric for rpcQueueTime, processing time
> --------------------------------------------------------------
>
> Key: HADOOP-9420
> URL: https://issues.apache.org/jira/browse/HADOOP-9420
> Project: Hadoop Common
> Issue Type: Bug
> Components: ipc, metrics
> Affects Versions: 2.0.3-alpha
> Reporter: Todd Lipcon
> Assignee: Andrew Wang
> Attachments: HADOOP-9420.txt
>
>
> Currently, we only export averages for rpcQueueTime and rpcProcessingTime.
> These metrics are most useful when looking at timeouts and slow responses,
> which in my experience are often caused by momentary spikes in load, which
> won't show up in averages over the 15+ second time intervals often used by
> metrics systems. We should collect at least the max queuetime and processing
> time over each interval, or the percentiles if it's not too expensive.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
