[ 
https://issues.apache.org/jira/browse/MESOS-3828?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15006371#comment-15006371
 ] 

Guangya Liu commented on MESOS-3828:
------------------------------------

In my understanding, I think that Mesos should leverage docker itself or some 
other open source project to manage the network for docker containers just like 
what kuberntes is doing now. 

The docker 1.9 now supports multi-host network for now, we can just create an 
over-lay network for all of the docker servers in mesos master via this docker 
feature; or else we can leverage flannel, weave or Calico to help the network 
management for Mesos and Docker integration.

> Strategy for Utilizing Docker 1.9 Multihost Networking
> ------------------------------------------------------
>
>                 Key: MESOS-3828
>                 URL: https://issues.apache.org/jira/browse/MESOS-3828
>             Project: Mesos
>          Issue Type: Story
>          Components: isolation
>    Affects Versions: 0.26.0
>            Reporter: John Omernik
>            Assignee: Timothy Chen
>              Labels: Docker, isolation, mesosphere, network, plugins
>
> This is a user story to discuss the strategy for Mesos to in using the new 
> Docker 1.9 feature: Multihost Networking. 
> http://blog.docker.com/2015/11/docker-multi-host-networking-ga/
> Basically we should determine if this is something we want to work with from 
> a standpoint of container isolation and going forward how can we best 
> integrate. 
> The space for networking in Mesos is growing fast with IP per Container and 
> other networking modules being worked on.  Projects like Project Calico offer 
> services from outside the Mesos community that plug nicely or will plug 
> nicely into Mesos.  
> So how about Multihost networking? An option to work with? With Docker being 
> a first class citizen of Mesos, this is something we should be considering. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to