cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sylvain Lebresne (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (CASSANDRA-6012) CAS does not always correctly replay inProgress rounds
Date Thu, 12 Sep 2013 15:44:51 GMT

     [ https://issues.apache.org/jira/browse/CASSANDRA-6012?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Sylvain Lebresne updated CASSANDRA-6012:
----------------------------------------

    Attachment:     (was: 0001-Don-t-skip-paxos-old-round-replay-if-there-is-a-value-.txt)
    
> CAS does not always correctly replay inProgress rounds
> ------------------------------------------------------
>
>                 Key: CASSANDRA-6012
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-6012
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: Sylvain Lebresne
>            Assignee: Sylvain Lebresne
>             Fix For: 2.0.1
>
>
> Paxos says that on receiving the result of a prepare from a quorum of acceptors, the
proposer should propose the value of the higher-number proposal accepted amongst the ones
returned by the acceptors, and only propose his own value if no acceptor has send us back
a previously accepted value.
> But in PrepareCallback we only keep the more recent inProgress commit regardless of whether
is has an update. Which means we could ignore a value already accepted by some acceptors if
any of the acceptor send us a more recent ballot than the other acceptor but with no values.
The net effect is that we can mistakenly accept two different values for the same round.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message