Working with Nutanix Metro Availability – part 2

Capacity planning for Nutanix Metro Availability (stretch cluster) is not that difficult as you could think. During capacity planning exercise, you have to consider all capacity aspects, compute (RAM and CPU on compute cluster) and storage (Nutanix storage on both datacenters) and network (datacenter interconnect). I will try to explain as easy as it is only possible in this post.

Capacity planning in HCI solution is different from 3-Tier capacity planning. The reason being is compute and storage are tide together very closely. When compute node is not available (hardware failure or maintenance), not only compute resources are gone from the resource pool (CPU and RAM) but also storage. Read here more about Nutanix Metro Availability.

Lets focus first on scenario where all VMs are protected by Nutanix Metro Availability.

Scenario:

• 5 nodes in datacenter 1 – same compute and storage capacity
  • 100TB RAW capacity / 20TB per node
• 5 nodes in datacenter 2 – same compute and storage capacity
  • 100TB RAW capacity / 20TB per node
• n+1 , on vSphere compute cluster, fault tolerance on each cluster in a datacenter – to avoid a situation where after DR event you have to compromise performance during standard maintenance activities like patching or hardware issues.
• RF=2 on Nutanix cluster (which means cluster can survive maximum of 1 simultaneous node\disk failure
• all VMs protected by Nutanix Metro Availability

Let’s do the math and how much resources we should reserve to make sure we have enough capacity to run workloads on a single site.

VMware vSphere reservations

In general math formula should be as follows:

((total number of nodes) – (cluster fault tolerance factor * 2)) / 2 = percentage reservation
(10 (100%) – 2 (20%))/2 = 40%

Based on the above formula, you have to reserve 60% (CPU and RAM) resource to be able to support workloads after failover to the surviving datacenter. In addition, each cluster in the datacenter can afford one node failure.

To set up vSphere resource reservations edit vSphere cluster settings and tune up Admission Control settings.

Nutanix storage reservation

Storage capacity planning is as important as compute (CPU and RAM). You have to make sure you have enough space to host all VM data, a copy of that data (data resiliency) and data coming from remote Nutanix cluster which is part of the Metro Cluster. Let’s do the math.

((Total RAW capacity) – 10%) – (RAW storage capacity of the biggest node in the Nutanix cluster) = maximum cluster storage utilization
(Total RAW capacity) – (maxium cluster storage utilization) = minimum cluster storage free space

Let’s do the math based on our scenario:

(100TB – 10%) – 20TB = 70TB – maximum cluster storage utilization
100TB – 70TB = 30TB – minimum cluster storage free space

Meaning you have to make sure clusters, not going below 30TB of free storage utilization. You can check it out from Prism Central or Prism Element.
NOTE: Storage utilization includes data replicated from the second cluster.

To reserve capacity on Nutanix cluster you can use “Advertise capacity” option on Nutanix container. What it does is advertise container capacity to the hypervisor. Setting up advertise capacity to the hypervisor, you are enabling “storage quota” your cluster.