cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Eric Evans (JIRA)" <j...@apache.org>
Subject [jira] Commented: (CASSANDRA-1086) Token range calculator
Date Tue, 18 May 2010 22:48:53 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-1086?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12868882#action_12868882
] 

Eric Evans commented on CASSANDRA-1086:
---------------------------------------

Correct me if I'm wrong (I could be misreading your script), but it seems that you are calculating
the mid-point between the most heavily loaded node and the one that precedes it (what Cassandra's
bootstrap does), and then dividing that by 2. Is that correct?

If so, your token placement will result in the transfer of half as many keys (resulting in
less distribution). Is this the intended behavior?

> Token range calculator
> ----------------------
>
>                 Key: CASSANDRA-1086
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-1086
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Contrib
>            Reporter: Suhail Doshi
>            Priority: Trivial
>             Fix For: 0.6.2
>
>
> We found loadbalance and autobootstrapping to be sub-par in production; also very risky
too if it guesses anything wrong.
> We created a simple Python script that I am sure can be greatly improved to be more automated
but thought we'd throw it up here anyway:
> http://gist.github.com/399336
> In a nutshell: give it your current token ranges, the load for that range, and it'll
suggest you the best token to help load balance your cluster when you decide to add a new
node. We've used it countless times in production ourselves. Our way of contributing some
small back to the community.

-- 
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