Intro

Mission: Deploy Kolla-ansible release Train with 1 controller, 2 computes and external Ceph (business purpose - test bug with live-migration in Train);

Prerequisites

• Working Ceph (3 nodes) within same network and root access;
• 3 VM for Openstack env (for test, 8 cpu/12ram/100GB local + 3 network interfaces(mgmt/storage/external));
• Deployment VM Ubuntu 18 (1 cpu/1ram/30GB);

Deployment

Prepare deployment host

SSH to deployment VM and begin:
cd /opt git clone https://github.com/openstack/kolla-ansible.git cd kolla-ansible git checkout stable/train apt update; apt -y install virtualenv virtualenv /opt/venv source /opt/venv/bin/activate pip install -r requirements.txt pip install ./ #install kolla-ansible package mkdir -p /etc/kolla chown $USER:$USER /etc/kolla cp -r etc/kolla/* /etc/kolla cp ansible/inventory/* .

Prepare target hosts

1. Configure all network interfaces (for data plane and control)
2. Install python-minimal
3. Configure passwordless access from deploymen VM

Configuration of interfaces

cat /etc/netplan/01-netcfg.yaml
network: version: 2 renderer: networkd ethernets: ens160: addresses: [ 10.220.104.53/24 ] gateway4: 10.220.104.254 nameservers: addresses: - "8.8.8.8" ens192: addresses: [ 10.220.103.53/24 ] ens256: addresses: [ 10.220.102.53/24 ]
netplan apply

Install python-minimal

apt update; apt -y install python-minimal

Configure passwordless access

deployhost:
ssh-keygen; cat /root/.ssh/id_rsa.pub (copy in buffer)
target1-3:
echo $(key) >> /root/.ssh/authorized_keys

Don’t forget to enable hardware virtualization

On compute nodes verify:
grep -r vmx /proc/cpuinfo

Also, check that nova user exists in Ceph, or create it

ceph auth list; ceph auth add client.nova ceph auth caps client.nova mon 'profile rbd' osd 'profile rbd pool=images, profile rbd pool=vms, profile rbd pool=volumes, profile rbd pool=backups'

Create files for ceph integration in /etc/kolla/config (without cinder-backup)

mkdir -p /etc/kolla/config/{nova,cinder,glance}
tree /etc/kolla/config/
. ├── cinder │   ├── ceph.conf │   ├── cinder-volume │   │   └── ceph.client.cinder.keyring │   └── cinder-volume.conf ├── glance │   ├── ceph.client.glance.keyring │   ├── ceph.conf │   └── glance-api.conf └── nova ├── ceph.client.cinder.keyring ├── ceph.client.nova.keyring ├── ceph.conf └── nova-compute.conf
paste there keyrings, ceph.conf from ceph and other stuff, more info here - https://docs.openstack.org/kolla-ansible/queens/reference/external-ceph-guide.html

Generate passwords

kolla-genpwd
Optionally, change keystone_admin_password there to something shorter for convenience;

Configure variables

vim /etc/kolla/globals.yml
/etc/kolla/globals.yml
vim /opt/kolla-ansible/multinode /opt/kolla-ansible/multinode

Other checks

In inventory file (multinode, for example), your nodes also should be placed under [baremetal:children] section (bootstrap).

Bootstrap

kolla-ansible -i multinode bootstrap-servers

Deploy

kolla-ansible deploy -i multinode

Post-Deploy

kolla-ansible post-deploy

Check basic stuff

If you installed kolla in venv, u also need openstack clients.
apt install -y build-essential python-dev source /opt/venv/bin/activate pip install openstackclient
Now, check basic statuses:
cinder service-list; nova service-list; neutron agent-list

Other stuff

Customize roles

If you are using venv as I do then path is:
/opt/venv/share/kolla-ansible/ansible/roles

Mirror images in registry

for img in $(docker images | awk '{print $1":"$2}' | grep -v REPOSITORY | tr '\n' ' '); do docker tag $img harbor.cognoz/$img; docker push harbor.cognoz/$img; done

Destroing Environment

cd /opt/kolla-ansible source /etc/kolla/admin-openrc.sh source /opt/venv/bin/activate nova list --all-tenants #first, you need to delete all instances kolla-ansible -i multinode destroy --yes-i-really-really-mean-it

Links

https://docs.openstack.org/kolla-ansible/latest/user/operating-kolla.html