Description

The traditional HA feature applies to VM failovers on shared storage. When a node fails, HCI can detect and pull up abnormal VMs on the node within 30 seconds to ensure service continuity. If any VM on a node is restarted due to a network (management network, storage area network, VXLAN network, business network, or aDesk communication network) interruption, or if its KVM process or correction status is abnormal, or if its node is offline for a certain period, the HA feature will be triggered to recover the VM on another node.

HCI 6.10.0 provides HA 2.0, which helps users discover and fix more failure risks (including hardware, platforms, and applications) in advance through comprehensive and optimized detection, actively detecting unhealthy nodes and abnormal networks and migrating VMs to healthy nodes, and providing alerts and tips for HA failures to ensure the effectiveness of the HA mechanism.

Precautions

1. The failover function is only available for virtual machines with HA checked.
2. Only virtual machines whose datastore is virtual storage or external storage can use the HA option.
3. In the HA scenario of many virtual machines, if the cluster resources are insufficient, it leads to HA failure.
4. After the platform license expires, the virtual machine HA does not work.
5. The HA feature cannot be enabled when the management interface and the storage network interface are offline, and when the cluster has only two nodes.
6. It is recommended for core applications to reserve a certain amount of resource space on some nodes in Reliability > Resource Reservation to ensure that there are enough resources to run the virtual machines for core services when HA occurs.
7. When the management and service communication interfaces are faulty, enabling HA only takes effect for the two-node cluster and does not take work in a three-node or more cluster.
8. When the communication interface is faulty, enabling HA only takes effect in a two-node cluster and does not take effect in a three-node or more cluster.
9. The HA-based migration only applies to VMs in a cluster and only migrates VMs to another node, not to another datastore. VMs stored on local storage do not support the HA-based migration.
10. The HA-based migration is not supported for the VMs of the following types: VMs with GPU passthrough mode enabled, VMs on non-shared storage, VMs with CPU enabled, quickly recovered VMs, VMs with low-latency interfaces, and encrypted VMs.

Prerequisites

HCI cluster consists of multiple nodes.

Steps

Go to Reliability > HA.

33. Click the Hardware tab, select Ignore or Restart VM on Healthy Node for Response Method for Encryption Card Failure and GPU Failure of Hardware Failure, and select Ignore or Hot-Migrate VM for Response Method for Hardware Health.
34. Click the Platform tab, select Network Disconnection Duration and Response Method for Network Disconnection, and select Response Method for Frequent Node Crashes and Low Node Reliability of Node Health.

35. Click Health Settings to go to Reliability > Host Health Monitoring > Health Monitoring to configure host and network health checks. For details, refer to Section 8.6.1 Host Health Monitoring.
36. Click the Application tab to learn that when VMs operate abnormally, the platform can automatically restart them for quick business recovery.

37. Click the Advanced Settings tab to configure resource reservation, node health monitoring, and live migration response schedule for the HA service.

38. Click the HA History tab to view the HA triggering records.

39. Select Restart VM on another node if the node fails for HA when creating and editing a VM to enable the HA feature.