You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@mesos.apache.org by "Joseph Wu (JIRA)" <ji...@apache.org> on 2016/11/09 18:57:58 UTC
[jira] [Updated] (MESOS-6479) add ability to execute batch jobs
from TaskGroupInfo proto in execute.cpp and add string flag for
framework-name
[ https://issues.apache.org/jira/browse/MESOS-6479?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Joseph Wu updated MESOS-6479:
-----------------------------
Labels: testing (was: newbie newbie++ testing)
In my opinion, we may not want to introduce these capabilities to {{mesos-execute}}:
* Batch capabilities are already available in some community frameworks, like Chronos and Aurora (and probably others).
* Thus far, the purpose of {{mesos-execute}} has been for single-use commands, usually for testing things on a small/single-node cluster. {{mesos-execute}} sits somewhere between being an example framework and a tool for people to play with. We keep it up to date, but we also want to keep {{mesos-execute}} as simple as possible.
* Breaking apart a {{TaskGroup}} into multiple separate {{Task}}'s effectively negates the concept of a {{TaskGroup}}. You'd be better off running {{mesos-execute}} once per {{Task}}.
> add ability to execute batch jobs from TaskGroupInfo proto in execute.cpp and add string flag for framework-name
> ----------------------------------------------------------------------------------------------------------------
>
> Key: MESOS-6479
> URL: https://issues.apache.org/jira/browse/MESOS-6479
> Project: Mesos
> Issue Type: Improvement
> Components: cli
> Affects Versions: 1.1.0
> Environment: all
> Reporter: Hubert Asamer
> Assignee: Hubert Asamer
> Priority: Trivial
> Labels: testing
> Original Estimate: 4h
> Remaining Estimate: 4h
>
> Extend execute.cpp to use TaskGroupInfo as container for batch jobs to distribute tasks based on available offers. A simple bool cli flag shall enable/disable such a behavior. If enabled the contents of TaskGroupInfo does not cause the execution of tasks within a "pod" (on a single host) but as distributed jobs (on multiple hosts)
> As an addition an optional cli flag for setting the temporary framework name (e.g. to better distinguish between running/finished frameworks) could be useful.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)