[
https://issues.apache.org/jira/browse/CASSANDRA-9180?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Brandon Williams updated CASSANDRA-9180:
----------------------------------------
Reviewer: Brandon Williams
Assignee: Tyler Hobbs (was: Brandon Williams)
Traced this back to CASSANDRA-6053, at least for bootstrapping. We don't look
at the STATUS, so in the switch statement in SS.onChange when any non-STATUS
state changes (which is guaranteed to happen) we persist the peer. Why
replacing is also affected I'm not sure, since we check for dead state and bail
out, but my guess is STATUS isn't the first state processed and the check
doesn't know the state so it fails.
> Failed bootstrap/replace attempts persist entries in system.peers
> -----------------------------------------------------------------
>
> Key: CASSANDRA-9180
> URL: https://issues.apache.org/jira/browse/CASSANDRA-9180
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Reporter: Brandon Williams
> Assignee: Tyler Hobbs
> Fix For: 2.0.15
>
>
> In working on CASSANDRA-8336, I discovered vanilla C* has this problem. Just
> start a bootstrap or replace and kill it during the ring info gathering
> phase. System.peers, the gift that keeps on giving.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
