[
https://issues.apache.org/jira/browse/CASSANDRA-11830?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15795017#comment-15795017
]
Paulo Motta commented on CASSANDRA-11830:
-----------------------------------------
This seems like an unintended side-effect of skipping the system keyspaces just
for the repair command on CASSANDRA-5483 as noted by [[email protected]], but
it ended up skipping for all nodetool commands that are executed without
explicit keyspace set. So we can restore the previous behavior for all
commands, except repair.
> "nodetool flush" not flushing system keyspace
> ---------------------------------------------
>
> Key: CASSANDRA-11830
> URL: https://issues.apache.org/jira/browse/CASSANDRA-11830
> Project: Cassandra
> Issue Type: Bug
> Components: Tools
> Environment: Ubuntu 14.04, Oracle Java 8, Cassandra 2.2.6
> Reporter: Dominik Keil
>
> I'm regularly splitting off maintenance systems from our QA cluster by adding
> a new node to a new "datacenter", joining it, then stopping and removing it
> (adapting the schema before and after accordingly).
> In order to not mix up the systems I rename the cluster in the newly created
> maintenance system by updating cluster_name in system.local
> In the past, when running "nodetool flush" then restarting Cassandra, this
> worked as expected.
> However, this time it did not. After restarting Cassandra the old cluster
> name was in place every time. However, explicitly flushing the system
> keyspace using "nodetool flush system" did work as expected.
> Is this a bug or did the default behaviour of "nodetool flush" change at some
> point?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
