We have figured out a way to provide  IPs to containers as well as 
network-level policy driven ACLs using Nuage’s VSP and Docker monitor installed 
on each slave node leveraging OVS.


Shafay Latif
> On Aug 10, 2015, at 11:46 PM, Christos Kozyrakis <[email protected]> wrote:
> 
> Hi Trevor, 
> 
> we are working with Project Calico in order to implement two important 
> features (urgently missing in Mesos imho):
> - IPs per container: this will eliminate port conflicts when apps with 
> specific port needs get deployed on the same slave
> - network-level isolation: so that you can control which apps can reach each 
> other and how, within or across slaves. 
> 
> The details will be presented at MesosCon and code released soon after that 
> to the open source. 
> 
> Let me know if you need more info ahead of time. 
> 
> On Mon, Aug 10, 2015 at 11:24 PM, Trevor Powell <[email protected] 
> <mailto:[email protected]>> wrote:
> Anyone have any thoughts on how Mesos may accomplish this use case?
> 
> We have several workloads that span multiple slaves and we want to ensure 
> those work loads can see each other, the internet, and nothing else. 
> Basically we have untrusted groups of work loads. We trust the load to talk 
> to itself across a several slaves. But we don’t trust it to not affect or 
> inspect other work loads on the same slave.  Basically we are looking to 
> place “blinders” on the work load. So it can only see what it needs to see 
> from the network level.
> 
> I have heard of things like weave or Project calico 
> (http://www.projectcalico.org/learn/ <http://www.projectcalico.org/learn/>) . 
> They seem promising. But I ponder what Mesos is looking to do long term.
> 
> -- 
> <11360A2A-682B-4E88-B66D-FF942D0869A1[183].png> <http://www.rms.com/>
> Trevor Alexander Powell
> Sr. Manager, Cloud Engineer & Architecture
> 7575 Gateway Blvd. Newark, CA 94560
> T: +1.510.713.3751 <tel:%2B1.510.713.3751> 
> M: +1.650.325.7467 <tel:%2B1.650.325.7467> 
> www.rms.com <http://www.rms.com/>
> 
> 
> -- 
> Christos

Reply via email to