[Starlingx-discuss] exploration for adding smaller memory node on stx system

[Sun, Austin]

Hi Victor:
     the containerizing is just current approach , as mentioned in ppt shared in google driver, there are other potential approaches.
     If memory is large enough , then not only K8S , but also openstack could be support since openstack are already containerized.
     About external OS package question , I would say currently NO, only need node support docker and Kubernetes. 
     Yes, there might be overlaps this task , containerization and multi-os ,  some items can be shared among those projects. 
     But the target/goal  is not same . 
     Those services are already containerized, so I just use "docker stats xxxxx" to check the memory and cpu usage of those container. 

Thanks.
BR
Austin Sun. 
  

-----Original Message-----
From: Victor Rodriguez <vm.rod25 at gmail.com> 
Sent: Wednesday, October 30, 2019 8:01 PM
To: Sun, Austin <austin.sun at intel.com>
Cc: starlingx-discuss at lists.starlingx.io
Subject: Re: [Starlingx-discuss] exploration for adding smaller memory node on stx system

On Wed, Oct 30, 2019 at 2:42 AM Sun, Austin <austin.sun at intel.com> wrote:
>
> Hi StarlingX Community:
>
>      I have spent some time to explore how to add smaller memory node on stx system.
>
>      The approach I’m trying is containerizing flocking services for worker node.
>

Thanks, Austin, having the capability to add small footprint nodes will open the possibilities for users of StarlingX

The approach is to containerizing flocking services for worker node seems the most straight forward solution. I was wondering if the scope will be just to make the node visible for k8 or also for open stack. I could think that for now just k8. One more question (based on the previous mail I sent about memory ) is there any host OS package is going to be needed? if so we can work in collaboration with the multi os team to handle it. I was wondering if the kernel or systemd might be needed on the worker + the flock services.

One more thing, how do you measure the MB of virtual memory used by the services?


>      And the status/achievement are
>
>      2G memory nodes (ubuntu 18.04 and centos 7.6) were successfully added to multi-node stx system .
>

Great !!

>      this node is unlocked ,  available and joined stx k8s system.
>
>      Sysinv-agent , hbsClient and mtcClient are containerized and deploy this node .
>      Configuration was manually set and some device audit was bypass.
>
>
>
>        There should be still a long way to fully functional worker 
> node on such smaller memory node, such as auto-config ,  devices audit 
> and other containerization services etc.,
>

I had the impression as well that it was going to be a long path, but is definitely important, especially for hybrid edge cloud systems

Thanks

>        The status and plan for net step is shared in [1]
>
>        I created story [2] to track the change if task is ready
>
>
>
> [1] 
> https://drive.google.com/file/d/1nt-Mx2Ebrqr4IBruSoOVmB5cjdOQ9HJR/view
> ?usp=sharing
>
> [2] https://storyboard.openstack.org/#!/story/2006787
>
>
>
> Thanks.
>
> BR
> Austin Sun.
>
>
>
>
>
>
>
>
>
>
>
> _______________________________________________
> Starlingx-discuss mailing list
> Starlingx-discuss at lists.starlingx.io
> http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss