[
https://issues.apache.org/jira/browse/KAFKA-2079?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14907723#comment-14907723
]
Joe Stein commented on KAFKA-2079:
----------------------------------
Not yet, I started a KIP for it
https://cwiki.apache.org/confluence/display/KAFKA/KIP-30+-+Allow+for+brokers+to+have+plug-able+consensus+and+meta+data+storage+sub+systems
I think we need a way to have some plug-able support for different remote
interface/system libraries.
It would be great for the base Kafka code to continue to support how it works
now. The existing code should get refactored and included with the release the
kafka-zkclient-connector.jar (or such) still support how it works now for folks.
We could then have a way for launching other libraries (not much unlike for
metrics) to implement the remote interfaces.
- async watchers
- leader election
- meta data storage
I don't think we should separate out the plug-in separately it should contain
at least those 3 base functionality.
Then folks can work on and build out different implementations and/or
collaborate on implementations.
I still need to write it up some more in the KIP and start a discussion on the
mailing list.
KAFKA-873 could be another implementation to use curator (or maybe it can get
into exhibitor or something). Other implementations folks have brought up are
Akka, Consul and etcd. I think folks can work on and build out and support the
different implementations and the existing Kafka brokers can still work how
they do now. At some point in a subsequent release we could replace the project
released jar with something else.
> Support exhibitor
> -----------------
>
> Key: KAFKA-2079
> URL: https://issues.apache.org/jira/browse/KAFKA-2079
> Project: Kafka
> Issue Type: Improvement
> Reporter: Aaron Dixon
>
> Exhibitor (https://github.com/Netflix/exhibitor) is a discovery/monitoring
> solution for managing Zookeeper clusters. It supports use cases like
> discovery, node replacements and auto-scaling of Zk cluster hosts (so you
> don't have to manage a fixed set of Zk hosts--especially useful in cloud
> environments.)
> The easiest way for Kafka to support connection to Zk clusters via exhibitor
> is to use curator as its client. There is already a separate ticket for this:
> KAFKA-873
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
