hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jonathan Gray (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HBASE-1931) Add table of config. for 1, 4, 20 node clusters and table of config. for 1, 2, 8 processor CPU
Date Fri, 23 Oct 2009 15:56:59 GMT

    [ https://issues.apache.org/jira/browse/HBASE-1931?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12769271#action_12769271
] 

Jonathan Gray commented on HBASE-1931:
--------------------------------------

+1 across the board.

I would say that one advantage of documentation vs an automated tool is that it provides a
good opportunity for someone to learn what decisions are being made according to different
hardware/cluster specs (rather than a magic black box).  Many users start out on 1 or 5 nodes
with plans to move to 10, 20, or more.  As we build up a good repository of information, we
can add more as we/users learn.

Another major factor is what the expected load patterns are, especially related to where you
might host your ZK nodes, max number of map/reduce tasks, etc...

Good stuff.

> Add table of config. for 1, 4, 20 node clusters and table of config. for 1, 2, 8 processor
CPU
> ----------------------------------------------------------------------------------------------
>
>                 Key: HBASE-1931
>                 URL: https://issues.apache.org/jira/browse/HBASE-1931
>             Project: Hadoop HBase
>          Issue Type: Task
>            Reporter: stack
>
> Chatting w/ J-D, we should add to 'Getting Started' a simple table that says how to lay
out daemons when you have a cluster of 1, 4, and 20 machines.  Then we'd have another table
that said what kind of config. to use when you have nodes of 1, 2, 4, 4+ processors.  J-D
wanted to go further and add a little jruby script to bin directory that could do a dumbed-down
version of what apurtell suggested a while back, a thing to spit out config.  It would ask
the user a couple of questions and read the local config. and environment -- which wouldn't
be hard if jruby --- then it would spit out recommendation.  It would be easier to maintain
than doc.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message