Setting up Fault Tolerant Hypervisors

Stratus recommends that any hypervisors tagged as fault-tolerant run the Stratus ftServer with the KVM, VMware, or Hyper-V hypervisor. These locations can be ensured to have total redundancy. You must ensure duplicate networks and if using central storage, that the storage has adequate Quality of Service (QoS).

Once the hypervisor software has been added, you must make the node discoverable to the OpenStack installation. For more information, refer to:

http://docs.openstack.org/havana/install-guide/install/apt/content/nova-compute.html

All hypervisors discovered by OpenStack appear in the Stratus Management Console in the Hypervisors page.

Physical Connectivity

In order to achieve the highest level of availability, the physical network connectivity must be redundant. Stratus suggests the following:

• Hardware configuration should include at least two NICs with enough ports in the cards to support the networking needs
• Bond two ports from two different physical cards to protect against NIC port failure and NIC failure
• Connect each port of the bonded network of the server to separate switches to protect against switch failures

In the following diagram, the server contains two NICs, each with two ports. The ports are marked as Card1Port1 through Card2Port2.

Bond:

• Card1Port1 and Card2Port1, and:
• Card1Port2 to Card2Port2

And connect the ports to two separate switches as shown.

Switch Configuration

Configure VLANs on the switch to match the configuration discussed in the previous section. Assign a temporary IP address, and verify network connectivity.

Storage Considerations

Consider the following recommendations when configuring storage:

1. Protect the boot volume with either RAID 5 or RAID 6. A single-disk boot volume can result in system failure, and will result in a failure of all instances if the disk fails.
2. Shared storage must be highly available, as all instances live on the shared storage.
3. Cinder block shared storage should be highly available. Unless a high-end storage solution is used on the backend, the disks must at least be RAID 5 or RAID 6 for protection from physical disk failures.