Recommended Guidelines For Maximum Storage Utilization On A Cluster

  • 12 October 2020
  • 0 replies
  • 1836 views

Userlevel 3
Badge +2

Customers can observe cluster issues when they use more than 90 percent of the total available storage on the cluster. This article provides general guidelines on the maximum storage utilization on a cluster.

 

The total available storage capacity on a Nutanix cluster is the sum of all the available storage space across the nodes.

Customers should not try to use all the available space in the cluster. Sufficient space should always be available in the cluster in case a failure occurs. For example, when a disk fails in the cluster, the extent groups that were present on that disk have to be copied to other disks to ensure that all the data is at the configured replication factor. If the disk space is unavailable for this activity, then there is a risk of data loss if more failures occur with the data in an under-replicated state.

 

The formula for calculating the maximum recommended usage for clusters is one of the following.

  • Recommended maximum utilization of a cluster with containers using replication factor (RF)=2 M = 0.9 x (T - N1)

  • Recommended maximum utilization of a cluster with containers using replication factor (RF)=3 M = 0.9 x (T - [N1 + N2])

SbB8Ai5snSm46uf20e8pzPNLNmvBH9ASdlq7RvsC3qu0sIC6a8pALD6hBIUKbUKj5ybE_KTNo-MX-K4PGs6pke2kBhGEuABemqD9_9x-fWRPQuJ2jy2T_voWW9EcWfPLE54Rx052

Please note that the formula takes the following into consideration.

  1. Space is reserved in the cluster to tolerate one full node failure.

  2. 10 percent space is reserved as a buffer for various actions such as adding or removing nodes or VMs as needed.

  3. The physical storage capacity of a node can be found by under Hardware > Diagram. Click a specific node in Diagram and see the value of Storage Capacity in the Prism web console.  

Example

Note: This example shows the physical storage utilization. The actual logical utilization is less due to replication.

For RF=2, the logical available space is half of the physical available space.

This example assumes a five-node cluster consisting of the following.

  • One (1) four-node Model NX-3050

    • The physical storage capacity available is 4.48 TBs per node, or 17.92 TBs total

  • One (1) node of a Model NX-6060

    • The physical storage capacity is 13.91 TBs per node

The total physical storage capacity available to use on this cluster is: (4 x 4.48) + 13.91 = 31.83 TB

  • T = 31.83

  • N1 = 13.91

  • N2 = 4.48

The recommended maximum utilization of this cluster when using Replication Factor of 2

  • M = 0.9 x (T - N1 )

  • 0.9 x ( 31.83 - 13.91) = 16.128 TB

The recommended maximum utilization of this cluster when using Replication Factor of 3

  • M = 0.9 x (T - [N1 + N2])

  • 0.9 x (31.83 - [13.91 + 4.48]) = 12.09 TB

Note: The values for storage used in the example are not the actual values which can be compared with the actual storage available for a Node Model NX-3050.

 

 

For more information on guidelines, please refer to the following documents:

Recommended guidelines for maximum storage utilization on a cluster

NCC Health Check

Calculation for: Cluster does not have enough capacity X for usage Z to tolerate any (or 2) node failure

Troubleshooting "Disk Space Usage High" alerts regarding HDD usage


This topic has been closed for comments