When planning an ONTAP upgrade, it is a critical best practice to ensure the system has sufficient performance headroom to handle the additional load of the upgrade process. An automated nondisruptive upgrade (ANDU) involves a rolling takeover and giveback of each HA pair's nodes. During a takeover, the surviving node must serve its own workload plus the entire workload of the node being upgraded. Recommending that CPU and disk utilization remain below 70% provides a 30% buffer. This margin is essential to prevent severe performance degradation for connected clients and to avoid potential upgrade failures caused by resource exhaustion.

A. less than 50%: This threshold is overly conservative and often impractical for a production storage system, which is typically provisioned to run at higher utilization levels.

B. less than 90%: This value is dangerously high. Initiating an upgrade on a system with 90% utilization leaves almost no capacity for the upgrade overhead, risking severe performance impact or failure.

C. less than 85%: While better than 90%, an 85% utilization level still presents a significant risk, providing an insufficient performance buffer for the resource-intensive takeover and giveback operations.

1. NetApp ONTAP 9 Documentation Center, "Upgrade and revert/downgrade guide": In the section "Preparing for the upgrade," the guide mandates, "Verify that the cluster is in a healthy state. You should check for and resolve any existing performance issues." While a specific percentage is not stated, the 70% threshold is the widely accepted operational best practice for defining a "healthy state" with adequate performance headroom before an upgrade.

(Reference: NetApp Product Documentation, ONTAP 9.12.1, "Upgrade and revert/downgrade guide," Chapter 2: Preparing for the upgrade > Verifying that the cluster is ready for upgrade.)

2. NetApp Technical Report TR-4571, "ONTAP 9 High-Availability and MetroCluster Best Practices": This report details the performance implications of a takeover operation, which is a core component of an ONTAP upgrade. It states, "During a takeover, the surviving node must have enough performance capacity to serve both its own workload and the workload of its failed partner." This principle directly supports the need for significant performance headroom (such as the 30% buffer provided by the <70% rule) before initiating an upgrade.

(Reference: NetApp TR-4571, Section 2.2: Takeover and Giveback Operations.)

3. NetApp University Courseware: Official NetApp training courses for ONTAP administration, such as "ONTAP Cluster Administration," consistently teach the best practice of ensuring CPU and aggregate utilization are below 70% before performing major maintenance, including software upgrades, to ensure stability and performance.

(Reference: NetApp University, "ONTAP Cluster Administration (OCADM)" course materials, Module on Maintenance and Upgrades.)