[
https://issues.apache.org/jira/browse/CHUKWA-670?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13506237#comment-13506237
]
Sourygna Luangsay commented on CHUKWA-670:
------------------------------------------
Ari has a point.
I see 2 other drawbacks for using ZK for this configuration:
- Chukwa would be more complicated with no real benefit. Having a "central
configuration system"
was what Flume old generation did. In the new generation, they decided to get
rid of it.
On the whole, I think that Chukwa should inspire from the Linux philosophy:
just do one think
and do it (very) well. Such central configuration problem could be solved by
Puppet, Chef or other, no?
And I have no doubt that companies with "large scale system" use one.
- I don't really like the idea of the agent depending on more hadoop
components. Usually, when I
have to install a new Hadoop system in a customer datacenter, Chukwa agent is
the only piece
of sofware I have to install on servers that are already in production. So I
like to say
to the client "Chukwa agent is very easy to install, has little impact on your
system and
has no dependencies apart from java 6". If the ZK system should be
implemented, I believe
it would be great if we could choose between this system or the current
configuration method.
> Use ZooKeeper for collector list and adaptor check list
> -------------------------------------------------------
>
> Key: CHUKWA-670
> URL: https://issues.apache.org/jira/browse/CHUKWA-670
> Project: Chukwa
> Issue Type: New Feature
> Components: Data Collection
> Environment: Java, ZooKeeper, Chukwa
> Reporter: Eric Yang
>
> Agent is currently using local disk to storage check point file, and
> collector list. This can be hard to manage for large scale system that needs
> to be updated more frequently. The proposal is to have a new configuration
> in chukwa agent configuration file to enable ZooKeeper as the location to
> store check point file and collector list.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
