[
https://issues.apache.org/jira/browse/CASSANDRA-15600?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17121134#comment-17121134
]
Kornel Pal commented on CASSANDRA-15600:
----------------------------------------
[~e.dimitrova] you have added a generateSplits method to
ReplicationAwareTokenAllocator, but it does not seem to be used. New comments
suggest that splits are supposed to be used, but generateRandomTokens is used
instead. Could you please double check whether some intended changes are
missing from the commit.
> Algorithmic token allocation does not handle the racks = RF case well
> ---------------------------------------------------------------------
>
> Key: CASSANDRA-15600
> URL: https://issues.apache.org/jira/browse/CASSANDRA-15600
> Project: Cassandra
> Issue Type: Bug
> Components: Feature/Virtual Nodes
> Reporter: Branimir Lambov
> Assignee: Ekaterina Dimitrova
> Priority: Normal
> Fix For: 4.0, 4.0-alpha4
>
> Attachments: CASSANDRA-15600 unit.txt, CASSANDRA-15600-jvm.txt.zip,
> Screen Shot 2020-03-02 at 6.55.41 PM.png
>
>
> When the number of racks in a data centre is equal to the replication factor,
> the node allocation algorithm may calculate the target ownership incorrectly
> and allocate unsuitable tokens. This causes some inefficiency when the number
> of nodes in the DC is small, and significant problems if the number of nodes
> in each rack is different.
> In this case (racks count equal to replication factor) the load within each
> rack is effectively distributed independently. The allocation algorithm
> should reflect that, restricting the allocation space to the rack (instead of
> the DC) and using the RF=1 solution.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
