[
https://issues.apache.org/jira/browse/FLUME-865?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13160932#comment-13160932
]
Jarek Jarcec Cecho commented on FLUME-865:
------------------------------------------
I would like to present my initial idea how to achieve requested functionality
for discussions. Basically I would introduce a new sink type "failover" that
would consists of arbitrary (non zero) count of sub sinks that would be ordered
somehow. The failover sink would keep an information of current sink and will
be trying to send data to it as long as:
1) there is no problem on the other side. In case of any problem, it would
switch active sink to next one and it would try to sent the event again. In
case that there wouldn't be any sink left to try, event would be simply
returned back to input channel as transaction would be rolled back.
2) higher priority sinks are still not available. It would be nice if the
failover sink would be periodically checking availability of failed sinks and
in case that one of them will go back online it would switch to this sink
immediately. It would be probably beneficial to have this "auto return back"
ability configurable, so that user can turn it off.
Example configuration in form of java property file:
agent01.sinks.m.type = failover
agent01.sinks.m.subsink.1.type = avro
agent01.sinks.m.subsink.1.hostname = first-collector
agent01.sinks.m.subsink.1.port = 12345
agent01.sinks.m.subsink.2.type = avro
agent01.sinks.m.subsink.2.hostname = second-collector
agent01.sinks.m.subsink.2.port = 12345
agent01.sinks.m.channel = input
I'm looking forward to see any comments, objections or better ideas :-)
Jarcec
> Implement failover sink
> ------------------------
>
> Key: FLUME-865
> URL: https://issues.apache.org/jira/browse/FLUME-865
> Project: Flume
> Issue Type: New Feature
> Components: Sinks+Sources
> Affects Versions: NG alpha 2
> Reporter: Jarek Jarcec Cecho
>
> It would be nice if the flume-ng would have ability to failover to different
> sink in case that the active one is not responding (e.g. before failing the
> transaction).
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
