Lab architecture

One region, two availability zones

This Openstack Lab is composed of one region (fr1) and two availability zones (az1 and az2).

In the first availability zone :

• 3 compute nodes (aka hypervisors) for Nova wihtin a Nova Cell called cell1
• 1 Ceph cluster for Cinder and Glance
• 3 bare metal nodes for Ironic

In the second availability zone :

• 3 compute nodes wihtin a Nova Cell called cell2
• 1 Ceph cluster (Cinder only)

Lab architecture

Networks

The different networks are :

• fr1-az1 : provider network in AZ1 (flat, i.e. without VLAN, shared, i.e. available for everybody and external, i.e. managed outside Openstack)
• fr1-az2 : provider network in AZ2
• ironic : external VLAN for bare metal nodes (only available in AZ1)
• octavia : external VLAN for Octavia management (only available in AZ1)

Networks

Resources

For the needs of the lab, the actual hardware used is :

• 4 VMs for Openstack controllers on a workstation (10 GB of RAM and 3 vCPUs for controller1 and controller2, 2 GB of RAM and 1 vCPU for ctrl-cell1 and ctrl-cell2)
• 3 VMs acting as hyperconverged (compute + Ceph) nodes in AZ1 on Hypervisor 1 (Supermicro X10SL7-F with Xeon E3-1265L v3, 32 GB RAM ECC DDR3). Each VM has 10 GB of RAM, 2 vCPUs, one virtual disk for OS and one virtual disk for Ceph OSD
• 3 VMs acting as bare metal nodes in AZ1 on Hypervisor 2 (HP Elitedesk 800 g2 mini with Core i5-6500t, 32 GB RAM DDR4)
• 3 VMs acting as hyperconverged (compute + Ceph) nodes in AZ2 on Hypervisor 3 (Supermicro X9SCM-F with Xeon E3-1265L v2, 32 GB RAM ECC DDR3)

The bare metal nodes are "simulated" thanks to Virtual BMC. Ironic uses among others the IPMI protocol to turn on/off/change the boot order of nodes.

VM distribution