Have 4 physical servers behind a UTM device which also acts hosts DNS service and acts as the gateway. This device is connected to a managed switch where all the physical servers are connected. IP of the UTM device or gateway & dns-server is 10.1.1.1
All nodes except contoller node will host compute and cinder servers, so configured two physical partitions in these nodes. Second partition for cinder volumes.
Step 1 : Install Ubuntu server 18.04 server (minimal) in all nodes.
DNS is updated with (in case of DNS services not available, have to update /etc/hosts with the following in all nodes so that each node and reach the other user hostname)
10.1.1.30 node1 node1.xxxxx.net 10.1.1.40 node2 node2.xxxxx.net 10.1.1.50 node3 node3.xxxxx.net 10.1.1.60 controller controller.xxxxx.net
Step 2 : Configure the FQDN to match the DNS entries in each node – Could not configure the FQDN during installation.
sudo hostnamectl set-hostname controller.xxxx.net sudo hostnamectl set-hostname node1.xxxx.net
Commented out the following in /etc/hosts
# 127.0.1.1 controller
Step 3 : Changes to network management – Optional and can skip to next step – Not mandatory – Just a personal choice.
In all nodes disable Consistent Network Device Naming, update /etc/default/grub entry as follows :
Update grub is essential for the changes to take effect during next boot.
Enable networking and disable netplan
sudo apt install -y ifupdown
On controller node, which has only two NICs configure one for management and one for provider network. Edit /etc/network/interfaces and update the contents as follows :
iface lo inet loopback
allow-hotplug eth0 auto eth0 iface eth0 inet static address 10.1.1.25 netmask 255.255.255.0 broadcast 10.1.1.255 gateway 10.1.1.1 dns-nameservers 10.1.1.1 184.108.40.206 allow-hotplug eth1 auto eth1 iface eth1 inet manual up ip link set dev $IFACE up down ip link set dev $IFACE down
On the other nodes, which have 4 NICs, bond the first three interfaces as a single bonded interface and the fourth one for provider network.
Note : On the switch (Cisco) had configured LAG for the 3 interfaces and also enabled LACP.
Edit /etc/network/interfaces and update with following contents :
source /etc/network/interfaces.d/* auto lo iface lo inet loopback auto eth0 iface eth0 inet manual bond-master bond0 auto eth1 iface eth1 inet manual bond-master bond0 auto eth2 iface eth2 inet manual bond-master bond0 auto bond0 iface bond0 inet static address 10.1.1.30 netmask 255.255.255.0 broadcast 10.1.1.255 gateway 10.1.1.1 dns-nameservers 10.1.1.1 220.127.116.11 slaves eth0 eth1 eth2 bond_mode 6 bond-miimon 100 bond_downdelay 200 bond_updelay 200 auto eth3 iface eth3 inet manual up ip link set dev $IFACE up down ip link set dev $IFACE down
Install ifenslave package and enable bonding module (only where bonding is planned – other than controller all nodes – in my case)
sudo apt install -y ifenslave sudo modprobe bonding echo "bonding" >> /etc/modules
Disable and mask NetworkManager and netplan in all nodes – enable “networking”.
Copy the following commands in say nmchanges.sh. Add execute permission “chmod +x nmchanges.sh” and finally execute the script with nohup so that momentary network disruptions do not impact the ssh session
sudo systemctl stop systemd-networkd.socket systemd-networkd networkd-dispatcher systemd-networkd-wait-online sudo systemctl disable systemd-networkd.socket systemd-networkd networkd-dispatcher systemd-networkd-wait-online sudo systemctl mask systemd-networkd.socket systemd-networkd networkd-dispatcher systemd-networkd-wait-online sudo apt -y purge nplan netplan.io sudo systemctl unmask networking sudo systemctl enable networking
Reboot the system so that the network management changes take effect.
Step 4 : Disable snapd and motd (Message of the day). Optional – can skip to next step – Just a personal choice.
sudo apt purge -y snapd
Disable message of the day. Edit /etc/default/motd-news and set ENABLE=0. In addition to that we can disable the motd.timer
sudo systemctl disable motd-news.timer sudo systemctl mask motd-news.timer
Comment out the following lines in /etc/pam.d/sshd as shown below
# session optional pam_motd.so motd=/run/motd.dynamic # session optional pam_motd.so noupdate # session optional pam_mail.so standard noenv # 
Remove the execute permission for few motd scripts
sudo chmod -x /etc/update-motd.d/10-help-text /etc/update-motd.d/50-motd-news /etc/update-motd.d/80-livepatch
sudo service sshd restart
Step 5 : Configure timezone and update DNS resolver configuration
timedatectl set-timezone Asia/Kolkata
Update DNS resolver configuration and restart – Edit /etc/systemd/resolved.conf. Uncomment and update DNS entry – 10.1.1.1 is the IP address of the UTM device which hosts DNS services also.
[Resolve] DNS=10.1.1.1 systemctl restart systemd-resolved
Step 6 : Install and configure chrony for time synchronization
apt install -y chrony
Comment out pool entries and add one server entry (public NTP server in cotnroller node and controller.xxxx.net in compute nodes).
server time.google.com iburst
Add the following to allow compute nodes to synch from controller (in controller node only)
Restart chrony services
systemctl restart chronyd.service
Step 7 : Other miscellaneous configuration updates
Avoid running out of file descriptor handles – Add the following at the end of /etc/security/limits.conf
* nproc hard 65535 * nproc soft 65535 * nofiles hard 65535 * nofiles soft 65535
Disable swap – Optional. In my case all compute nodes have 192G RAM and the dedicated controller has 48G RAM and hence a personal choice.
swapoff -a rm -f /swap.img
Edit /etc/fstab and remove the line related to swap – so that the swap disabling gets permanent.
Optionally reboot the system – Not mandatory – Just a personal choice.
Step 8 : Add openstack packages repository and install openstack client. “crudinit” tool will be helpful while updating config files.
add-apt-repository -y cloud-archive:train apt update apt -y upgrade apt install -y python3-openstackclient crudini