In a Nutanix cluster, I/O latency can be affected by several factors. Given the scenario where a 32-node hybrid cluster is experiencing I/O latency, one possible cause could be the Information Life Cycle Management (ILM) constantly moving data due to curator scans1.
In a Nutanix cluster, ILM is responsible for managing data placement across different storage tiers. It promotes frequently accessed data to higher-performance tiers (like SSDs) and demotes less frequently accessed data to lower-performance tiers (like HDDs)1. This process involves moving data around, which can cause I/O operations to take longer, leading to increased latency1.
If the working set size (WSS) exceeds the SSD tier capacity, ILM will have to constantly move data between the SSD and HDD tiers to accommodate new data. This constant data movement can lead to increased I/O latency1.
Therefore, if curator scans caused by ILM are constantly moving data, it could be the reason for the observed I/O latency in this scenario.
References:
* Disk I/O Latency on a Nutanix Cluster | Nutanix Community
* Nutanix Volumes - Recommendations And Best Practices
* Nutanix Support & Insights
* Expand Cluster with Nodes with Higher NIC SPEED (25 GBPS) | Nutanix ...
* [Scaling Up VM Storage Performance on AHV - Volume Group ... - Nutanix]
* [About support for Nutanix segmented iSCSI network | Managing AHV ...]

In a Nutanix environment, if the login process using an Active Directory (AD) user account takes more than a minute to complete, it could be due to the configuration of the AD lookups.
Active Directory lookups can be configured to use either Recursive or Non-Recursive mode. In Recursive mode, the system will look up each group that a user is a member of, then each group that those groups are members of, and so on. This can lead to a large number of lookups and potentially slow down the login process1.
On the other hand, in Non-Recursive mode, the system only looks up the direct groups that a user is a member of. This can significantly reduce the number of lookups and speed up the login process1.
Therefore, the administrator should confirm that Active Directory lookups are configured to use Non-Recursive mode to resolve this issue.
References:
* Active Directory Authentication is slow - Takes minutes to Logon | Nutanix Community
* Configuring Authentication | Nutanix Community
* Solving Slow Logon Issues in Active Directory Environments

The consultant should be aware that the subnet 192.168.5.0/24 is reserved for internal connectivity within a Nutanix cluster. This subnet is typically used for the internal operations of the Nutanix cluster, including CVM backplane communications. It is critical to ensure that this subnet does not overlap with any external network configurations to avoid networking conflicts that can affect cluster operations and communications.References: Nutanix Networking Best Practices, Nutanix Cluster Configuration Guide.

To ensure additional resiliency in a 20-node ESXi deployment capable of surviving the concurrent loss of two nodes, the consultant should configure the system with a storage reservation and HA (High Availability) admission control each set to 5%, combined with a cluster redundancy factor (RF) of 3. RF3 provides the capability to tolerate the failure of up to two nodes without data loss, thus enhancing the cluster's resilience.
This configuration is optimal for maintaining cluster availability and integrity during multiple node failures.References: Nutanix Bible, Nutanix University