[
https://issues.apache.org/jira/browse/CASSANDRA-9142?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14519271#comment-14519271
]
Marcus Eriksson commented on CASSANDRA-9142:
--------------------------------------------
good point, but on trunk we anticompact and mark sstables as repaired even
after -full repairs - we always send the AnticompactionRequest and we need to
do that to make sure the neighbors remove the parent repair sessions
We should probably add a "dont anticompact"-flag to AnticompactionRequest or
send empty successfulRanges which will have the same effect.
On 2.1 it should be enough to just disallow invoking nodetool with -inc with
any of the local params
> DC Local repair or -hosts should only be allowed with -full repair
> ------------------------------------------------------------------
>
> Key: CASSANDRA-9142
> URL: https://issues.apache.org/jira/browse/CASSANDRA-9142
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Reporter: sankalp kohli
> Assignee: sankalp kohli
> Priority: Minor
> Attachments: trunk_9142.txt
>
>
> We should not let users mix incremental repair with dc local repair or -host
> or any repair which does not include all replicas.
> This will currently cause stables on some replicas to be marked as repaired.
> The next incremental repair will not work on same set of data.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
