[
https://issues.apache.org/jira/browse/HBASE-15967?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15318677#comment-15318677
]
Bryan Beaudreault commented on HBASE-15967:
-------------------------------------------
Hey [~stack], here's a gist of the current iteration of our metrics source:
https://gist.github.com/bbeaudreault/bbec28c502d12148f8657aa3f9fecd4f This is
still in code review so may change. We're also still performance testing. Feel
free to use as needed, if you want. There's some extra complication in there,
because we also wanted to be able to track rpc usage percentages by user, which
we're using to detect problem clients (using more than their fair share of rpc
time) and automatically put them into a separate detention queue or apply a
quota, depending on the situation.
We didn't use the DynamicMetricsRegistry, instead we used yammer metrics for
calculating an exponentially weighted moving average and use the 1-min average.
We directly hook into metrics2 by creating a MetricsSource. At read-time
(calls to getMetrics) we use yammer's EWMA#rate() function to return the
accumulated nano idle time, basically converting it into nanosIdle/nanos over
the time range. We divide that by the # of handlers in the pool to get an
average, and subtract from 100. The moving average is updated by calls to
countIdlePeriod() which is called from the RpcExecutor
> Metric for active ipc Readers and make default fraction of cpu count
> --------------------------------------------------------------------
>
> Key: HBASE-15967
> URL: https://issues.apache.org/jira/browse/HBASE-15967
> Project: HBase
> Issue Type: Sub-task
> Reporter: stack
> Assignee: stack
> Attachments: HBASE-15967.master.001.patch
>
>
> Our ipc Readers are hard coded at 10 regardless since . Running w/ less
> Readers, we go faster..(e.g. 12 Readers has us doing 135k with workloadc and
> 6 readers has us doing 145k).. .but hard to tell what count of Readers are
> needed since no metric.
> This issue changes Readers to be 1/4 the installed CPUs or 8, whichever is
> the minimum, and then adds a new hbase.regionserver.ipc.runningReaders metric
> so you have a chance seeing whats needed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
