hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sergey Shelukhin (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-9829) make the compaction logging less confusing
Date Fri, 06 Dec 2013 22:07:36 GMT

    [ https://issues.apache.org/jira/browse/HBASE-9829?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13841779#comment-13841779
] 

Sergey Shelukhin commented on HBASE-9829:
-----------------------------------------

There are periodic questions on the thread about why major compactions run when they are disabled
or not according to time table... this tries to clarify. In the similar vein, "large" and
"small" compaction threads get confused with "major" and "minor" compactions, so people assume
large thread == major compaction. They are chosen by size, but the reason behind the choice
is to remove potentially long-blocking compactions from the thread where many small ones may
run (by default, there's only one of each thread, so large compaction would block them), so
I think the naming is allowable.

Time in the thread name is actually not very useful, other than for grepping by number. Maybe
these threads can just be numbered when they are started? "longCompactions-1" is still greppable

> make the compaction logging less confusing
> ------------------------------------------
>
>                 Key: HBASE-9829
>                 URL: https://issues.apache.org/jira/browse/HBASE-9829
>             Project: HBase
>          Issue Type: Improvement
>          Components: Compaction
>            Reporter: Sergey Shelukhin
>            Assignee: Sergey Shelukhin
>            Priority: Minor
>         Attachments: HBASE-9829.patch
>
>
> 1) One of the most popular question from HBase users has got to be "I have scheduled
major compactions to run once per week, why are there so many".
> We need to somehow tell the user, wherever we log that there is a "major" compaction,
whether it's a major compaction because that's what was in the request (from regular major
compaction or user request), or was it just promoted because it took all files. Esp. the latter
should be clear.
> 2) small vs large compaction threads and minor vs major compactions is confusing. Maybe
the threads can be named short and long compactions.
> We



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

Mime
View raw message