cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Benedict (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-7954) Skip paxos for CAS if RF=1
Date Fri, 19 Sep 2014 15:43:36 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-7954?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14140755#comment-14140755
] 

Benedict commented on CASSANDRA-7954:
-------------------------------------

To clarify, if the owning node is down, with RF=1 we cannot fulfil any consistency level with
hinting, and cannot possibly perform our read-before-write, so the coordinators absolutely
require the owning node to be up to fulfil the request. If the request is sent by the coordinator
to the owning node, and applied conditionally locally by the owning node, with the result
of the conditional update being returned to the coordinator, no paxos is necessary and the
semantics and guarantees are the same.

> Skip paxos for CAS if RF=1
> --------------------------
>
>                 Key: CASSANDRA-7954
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-7954
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: Benedict
>            Priority: Minor
>              Labels: performance
>
> If RF=1, there's no point doing a paxos round - we get just as much safety and atomicity
by simply conditionally applying the update locally



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message