[
https://issues.apache.org/jira/browse/BIGTOP-1581?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14260960#comment-14260960
]
Biju Nair commented on BIGTOP-1581:
-----------------------------------
One of our use case is to copy multiple log files from the same node into
corresponding HDFS files and be able to manage them as separate services i.e
stop|start the services copying each of the log files. Since the nodes have
configuration like 256 GB RAM, resource is not a concern.
Since the changes made to the init.d script includes the option to start one
flume agent by name, users can use that option to bring up a flume agent
without wasting the resources if that is the concern.
But with the current script, users who need to bring up multiple flume-agents
need to duplicate the script to manage them as different services. This can
also end up polluting the /etc/init.d directory with many init.d scripts.
> Allow multiple Flume agents to be executed as a service using Bigtop init.d
> script
> ----------------------------------------------------------------------------------
>
> Key: BIGTOP-1581
> URL: https://issues.apache.org/jira/browse/BIGTOP-1581
> Project: Bigtop
> Issue Type: Improvement
> Components: Init scripts
> Affects Versions: 0.6.0, 0.7.0, 0.8.0
> Reporter: Biju Nair
> Labels: features
> Fix For: backlog, 0.8.0, 0.9.0, 1.0.0
>
> Attachments: flume-agent.init
>
>
> Using the current init.d script only one Flume agent can be managed as a
> service. In practice there will be more than one Flume agent on a node which
> need be managed. If the init.d script can be improved to accept a agent name
> for the service requests (start|stop|restart|...) and manage them it will
> help many installations.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
