[
https://issues.apache.org/jira/browse/HBASE-17016?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15634425#comment-15634425
]
Andrew Purtell edited comment on HBASE-17016 at 11/3/16 9:59 PM:
-----------------------------------------------------------------
bq. We could also look at moving to HdrHistogram for all of our histogram usage.
+1
Really fine grained latency histograms would be nice to have, if cheap enough.
Otherwise, external monitoring based on the Canary can still achieve this, at
cost/benefit of rolling whatever you need.
So remove now, then _maybe_ put back reimplemented with HdirHistogram, if
workable?
was (Author: apurtell):
bq. We could also look at moving to HdrHistogram for all of our histogram usage.
+1
Really fine grained latency histograms would be nice to have, if cheap enough.
Otherwise, external monitoring based on the Canary can still achieve this, at
cost/benefit of rolling whatever you need.
So remove now, then put back reimplemented with HdirHistogram?
> Remove per-region latency histogram metrics
> -------------------------------------------
>
> Key: HBASE-17016
> URL: https://issues.apache.org/jira/browse/HBASE-17016
> Project: HBase
> Issue Type: Task
> Reporter: Andrew Purtell
> Fix For: 2.0.0, 1.4.0
>
>
> Follow up from HBASE-10656, where [~enis] says:
> {quote}
> the main problem is that we have A LOT of per-region metrics that are latency
> histograms. These latency histograms create many many Counter / LongAdder
> objects. We should get rid of per-region latencies and maybe look at reducing
> the per-region metric overhead.
> {quote}
> Let's consider removing the per-region latency histograms until we have a low
> impact (in time and space) accounting.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
