NI-testbed

Testbed for NI.

Testbed topology

 +------------+            +------------+
 | Controller |            | Monitoring |
 +-----+------+            +-----+------+
       |                         |
-------++-------------+----------+------+-------- Network
        |             |                 |
 +------+----+  +-----+-----+      +----+------+
 | Compute 1 |  | Compute 2 | ...  | Compute N |
 +-----------+  +-----------+      +-----------+

Installation

Clone the repo

This repo has several submodules.

git clone https://github.com/tu-nv/ni-testbed
git submodule update --init --recursive

Prepare env

NI testbed with VMs

We recommend (as we only tested on) Ubuntu 18.04 LTS kernel.

• Install vagrant.

  wget https://releases.hashicorp.com/vagrant/2.2.8/vagrant_2.2.8_x86_64.deb
  sudo apt install ./vagrant_2.2.8_x86_64.deb

• Ensure kvm and nested virtualization are enabled, then install qemu and libvirt.
• Install vagrant-libvirt.
• Install OpenvSwitch and create bridge ovsbr0

  sudo apt install openvswitch-switch
  sudo ovs-vsctl add-br ovsbr0
  sudo ifconfig ovsbr0 192.168.3.1/24 up

• Provide internet access to ovsbr0 network

  sudo sysctl -w net.ipv4.ip_forward=1
  sudo iptables -t nat -I POSTROUTING -s 192.168.3.0/24 -j MASQUERADE
  sudo iptables -P FORWARD ACCEPT
  sudo iptables -t nat -P POSTROUTING ACCEPT

• Provision VMs.

  vagrant up --provider=libvirt --no-parallel

NI testbed with physical servers

If you use physical servers instead of VMs, prepair nodes as described in testbed topology. In our setup, all node run Ubuntu 16.04 server. Nodes should support KVM.

Install devstack

• Install ansible 2.7 via pip (install as root) and sshpass:

  sudo apt install -y python-pip sshpass
  sudo pip install ansible==2.7.16

• Use vagrant inventory as the example and modify to suit your custom setup.

  cd inventories
  cp vagrant your_inventory_file

• Bootstrap

  ansible-playbook -i inventories/<your_inventory_file> bootstrap.yml

• Setup devstack

  ansible-playbook -i inventories/<your_inventory_file> openstack.yml

Add more compute node to the cluster

• Add an entry of new compute node to the equivalent intentory file

  [computes]
  141.223.xx.xx HOST_NAME='NI-Compute-xx-xx' DATA_INTERFACE='<if_name>'
  

• Run ansible

  ansible-playbook -i inventories/physical_tesbed_001 openstack.yml --limit 141.223.xx.xx

Firewall

Currently, we applied some firewall rules to the monitoring node. Check roles/firewall for details.

Issues

• Currenly, Graylog requires manual adding rsyslog input by going to Dashboard -> System/Inputs.
• If use networking-ovs-dpdk then need to setup reserved_huge_pages, otherwise there will be error Insufficient free host memory pages. Custom conf file to set reserved_huge_pages.