Re: OOM: K8S deployment minimum nodes


DeWayne Filppi
 

Perhaps there is some adjustment to the Helm charts needed, because the aai-cassandra container fails claiming it needs 2 nodes.  Granted, I'm using Casablanca code.  Are you aware of any tweaks needed to the charts?


On Thu, Mar 14, 2019 at 1:01 PM Kranthi Guttikonda <kranthi.guttikonda@...> wrote:

Hi DeWayne,

 

I don’t think there is such requirement. You can deploy the whole ONAP in single HP or DELL gen9 v4 or above server (having 256GB RAM, 56 CPU (with hyper threading)). All you need to do is increase the --max-pods in kubelet. I did this with Rancher and kubeadm and able to deploy ONAP successfully in single server.

 

Attached the rancher POST call for reference.

 

Also I did test ONAP Casablanca with helm v2.12.3 and kubelet version 1.13.4 as well. With latest kubeadm you need to adjust the value of maxPods in “/var/lib/kubelet/config.yaml” and restart the kubelet

 

Coming to taints, affinities and nodeSelector fields unless you specify ONAP charts doesn’t have these enabled as default.

 

Thanks,
Kranthi

 

From: <onap-discuss@...> on behalf of "DeWayne Filppi via Lists.Onap.Org" <dewayne=cloudify.co@...>
Reply-To: "onap-discuss@..." <onap-discuss@...>, "dewayne@..." <dewayne@...>
Date: Thursday, March 14, 2019 at 3:37 PM
To: "onap-discuss@..." <onap-discuss@...>
Subject: [onap-discuss] OOM: K8S deployment minimum nodes

 

  Has anyone established the minimum number of compute nodes that ONAP can run on (in a K8S deployment)?  Not counting Rancher or the Kubernetes master.  I know it's at least 2, but was wonder if anyone had established the minimum.  I'm not talking about resource limitations, I'm talking about actual deployments or pods having taints or affinities that require a certain number of nodes.

 

Thanks,

 

DeWayne

 



--

DeWayne Filppi

Director, Sr. Architect CTO Office

+972523609583cloudify.co @cloudifysource
        


Join onap-discuss@lists.onap.org to automatically receive all group messages.