cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Brandon Williams (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-8175) Extend existing endpoint check to also check when restarting normally
Date Thu, 23 Oct 2014 16:16:34 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-8175?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14181523#comment-14181523
] 

Brandon Williams commented on CASSANDRA-8175:
---------------------------------------------

Adding this without CASSANDRA-8072 figured out would make me very nervous.

> Extend existing endpoint check to also check when restarting normally
> ---------------------------------------------------------------------
>
>                 Key: CASSANDRA-8175
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-8175
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: Brandon Williams
>            Assignee: Brandon Williams
>             Fix For: 2.1.2
>
>
> In CASSANDRA-5571 we added a check in shadow gossip to see if there was an existing node
in gossip, but only if we bootstrap.  The scenario I want to prevent here is similar, but
slightly different.  If someone restores a node to their development environment for whatever
reason, they have a good chance of accidentally joining that node to the production cluster
since it will know about the production cluster from the system table, and now this node will
also have a higher generation than the existing one, causing a host id collision and knocking
it out of the ring (except in its own view.)  Instead what I'd like to do is always do a shadow
gossip round at startup, and check for conflicting host ids.  Obviously we'll want to bypass
this for replacement.



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

Mime
View raw message