[
https://issues.apache.org/jira/browse/CASSANDRA-5897?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13913484#comment-13913484
]
Tyler Hobbs commented on CASSANDRA-5897:
----------------------------------------
Thanks, Daniel!
Overall this looks pretty good. I would still check if the dc, rack, or
preferLocal have changed before invalidating cached rings and gossiping.
Although ResourceWatch only triggers a reload if the mod time on the file goes
up, that doesn't necessarily mean the dc or rack have changed.
And a couple of nitpicks:
* You don't need curly braces for single-line "if" statements
* Name the constant {{DEFAULT_REFRESH_PERIOD_IN_SECONDS}} instead of
{{DEFAULT_REFRESH_PERIOD_IN_S}}
* Although this case should be rare, I would log at error instead of debug if
there's a ConfigurationException setting up the watcher.
> GossipingPropertyFileSnitch does not auto-reload local rack/dc
> --------------------------------------------------------------
>
> Key: CASSANDRA-5897
> URL: https://issues.apache.org/jira/browse/CASSANDRA-5897
> Project: Cassandra
> Issue Type: Bug
> Reporter: Jeremy Hanna
> Assignee: Daniel Shelepov
> Fix For: 2.1
>
> Attachments: trunk-5897.txt
>
>
> With the property file snitch, I can change the rack and dc of a node in
> cassandra-topology.properties from DC1/RAC1 to DC2/RAC2 while the server is
> running and that configuration is reloaded a short time later.
> With the gossiping property file snitch, I change the
> cassandra-rackdc.properties when the server is running from DC1/RAC1 to
> DC2/RAC2 while the server is running and it never reloads. It requires a
> restart.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
