{{ secondMenu.name }}
Description
A network can be unhealthy due to hardware or software issues (e.g., loosely seated NICs, loosely seated or too hot optical modules), which can greatly degrade performance and affect the system’s running (for example, some VM networks encountering latency, or networks between clusters encountering high latency or too many packet losses).
Network health monitoring focuses on physical links from NICs to switch ports. When an unhealthy issue (packet loss, latency) occurs, it can proactively disconnect or isolate the failed link and alert the user to avoid business impact.
As shown in the figure below, the hosts in a cluster are arrayed in rings, with each detecting up to five. The figure shows the host links to be probed by hosts A, B, and C.
Prerequisites
N/A
Precautions
| Limitation |
Description |
| Number of Aggregate Interfaces |
Storage area network: Network health monitoring is supported only if standard link aggregation is enabled and the number of physical interfaces is less than or equal to 2. Network health monitoring is not supported if the vs-bond mode of earlier versions is used. Non-storage area network: Network health monitoring is supported if the number of member physical interfaces is less than or equal to 2. If network health issues are detected in scenarios with link aggregation disabled, alerts are reported, but physical interfaces are not deactivated. |
| Aggregation Mode |
Health issue alerting: Alerts are reported if network health issues are detected in the following aggregation modes: active/standby, manual load balancing, and LACP. Health issue handling: For virtual storage area networks, health issues detected in the following modes are handled: active/standby, manual load balancing, and LACP. For non-virtual storage area networks, health issues detected in manual load balancing mode are not handled. |
| Switch Configuration |
Load balancing policy: Load balancing based on the source address or the source and destination addresses at Layer 2, Layer 2+Layer 3, and Layer 2+Layer 3+Layer 4 are supported. If combined load balancing policies are supported, you need to select the load balancing mode based on the source and destination addresses or source MAC address (e.g., NVIDIA and Mellanox switches). |
| Virtual Storage Area Network |
Network health monitoring is not supported for a datastore with 2 nodes or with 2+1 (2 HCI nodes+1 witness node) and the witness node of a stretched datastore. |
| Cluster Communication Domain |
Network health monitoring is not supported in cluster communication domains. |
| Interface Multiplexing |
Network health monitoring is not supported for a storage network interface with interface multiplexing enabled. The network health monitoring settings (including the thresholds and actions) of a storage area network are synchronized from other network planes. |
| Network Segments in IP Pool of VXLAN Network |
Network health monitoring is not supported in scenarios where different network segments of a VXLAN network communicate over Layer 3. |
| System Resource Load |
Network health monitoring is not supported for an overloaded HCI node (with a CPU idle rate lower than 10% or available memory size less than 2 GB). |
Steps
Go to Reliability > Host Health Monitoring > Health Monitoring to configure latency and packet loss rate thresholds for communication interfaces in the Network Health area. Alerts will be triggered when the specified thresholds are reached.
{{ $t('index.defaultHeader.chromeBrowserTip') }}