1
0
-1

We are trying to do full installation of Casablanca version of ONAP on Openstack using Rancher approach. We are using a total of 15 VMs with 1 being Rancher VM (16GB RAM, 8 VCPUs, 80GB Hard disk) and 14 Kubernetes cluster VMs (each VM with 16GB RAM, 8 VCPUs, 100GB Hard disk). 

Considering the hard disk space being reserved upto 1.48TB might be a problem in future. With reference to link: https://docs.onap.org/en/casablanca/guides/onap-developer/settingup/index.html#installing-onap, it says full installation of ONAP needs 160GB of HD as whole (See under Requirements). 

Can someone suggest me on exact configuration details(Hard disk in specific) of each K8s cluster VM. A clear documentation would be of great help!

TIA!!

    CommentAdd your comment...

    1 answer

    1.  
      1
      0
      -1

      Tejaswi Tanneru It should be 160GB per VM. If any node goes down k8s will create the pods in any other node. So, that involves downloading the docker image. We have to update a note there saying that HD requirement is per node but not for whole cluster (We can't put a monolithic number for cluster since number of hosts may change)

      1. Abhay Narayan Katare

        kranthi guttikonda One quick thing, when we are saying K8s cluster size is 14 VMs so ideally space required sould be 14*160G ( considering all in a sibgle host ), right ? 

      2. kranthi guttikonda

        Abhay Narayan Katare It doesn't matter from where the VM storage comes from. All the VMs could be from single host or scattered across multiple computes or storage is from cinder backend like LVM or HITACHI (..etc). For k8s every VM or server as a node. So no matter how many hosts we have its better to match the HDD size. Here is the take, if you have a VM with 32 CPUs, 128GB RAM and 160GB HDD and another VM with 8 CPUs, 16GB RAM, 80 GB harddisk your k8s default scheduler will try to place most of the pods in first VM because if has more resources. So, eventhough you have less HDD space your installation will succeed. This is just an example. However only problems comes when the first VM goes down then k8s will try to put some of the pods into 2nd VM and it may fail while downloading docker images itself because of HDD space. 

      CommentAdd your comment...