[
https://issues.apache.org/jira/browse/AURORA-389?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
David Robinson reopened AURORA-389:
-----------------------------------
The observer is the only Aurora component with a use-case, but the mechanism
could be used for many other things, such as running a monitoring service (like
[collectd|http://collectd.org/] or
[cadvisor|https://github.com/google/cadvisor]) on each host, or a local DNS
resolver, or even a SSH daemon. Most users would run at least one service on
each host, but without this feature they're left to managing (running,
monitoring, updating) the service on their own. Aurora already provides a
management solution, but without this feature it can't support these use-cases.
> provide a mechanism to run a task on every slave
> ------------------------------------------------
>
> Key: AURORA-389
> URL: https://issues.apache.org/jira/browse/AURORA-389
> Project: Aurora
> Issue Type: Task
> Components: Client, Scheduler
> Reporter: David Robinson
> Priority: Minor
>
> Aurora does not provide a mechanism for running a task on every slave. The
> config for a job requires that the number of instances be specified in
> advance, but it's not practical to adjust this number whenever slaves are
> added or removed. If this mechanism were available it could be used to run
> Thermos Observer or other tasks which need to be run on every slave. Without
> this mechanism everyone will need to duplicate Aurora's deployment mechanism
> outside of Aurora. If we could use Aurora to deploy these components we'd
> avoid this duplication.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
