You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Lars Hofhansl (JIRA)" <ji...@apache.org> on 2013/10/09 07:13:42 UTC

[jira] [Commented] (HBASE-9731) updatesBlockedSeconds RegionServer metric should not be a histogram

    [ https://issues.apache.org/jira/browse/HBASE-9731?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13790031#comment-13790031 ] 

Lars Hofhansl commented on HBASE-9731:
--------------------------------------

We can relatively safely change this because:
# It's not useful the way it is.
# We (Salesforce) added this, and are as far as we know the only ones planning to use this.


> updatesBlockedSeconds RegionServer metric should not be a histogram
> -------------------------------------------------------------------
>
>                 Key: HBASE-9731
>                 URL: https://issues.apache.org/jira/browse/HBASE-9731
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.94.13
>
>
> I just noticed that updatesBlockedSeconds is declared as a histogram (in 0.94), which makes little sense (neither mean, nor any percentiles are useful, as it is an every increasing number).
> This metric measures the total time the region server has blocked writes due to flushes that could not proceed due to too many store files... In that it should be a number. Unless there are objections I will make this so.
> Also, it looks like this metric was lost in 0.96 ([~eclark], [~stack], fyi).



--
This message was sent by Atlassian JIRA
(v6.1#6144)