Cluster

Cluster Basic Operations

You can understand the basic operations supported by clusters from the perspective of CRUD (Create, Read, Update, Delete).

Create a Cluster
Edit a Cluster
View Cluster Capacity Information
Delete a Cluster

Create a Cluster

The platform provides multiple entry points for creating clusters. You can create a cluster from the following two main entry points:

In the navigation bar on the left side of the platform page, right-click the target data center and click New Cluster.
In the navigation bar on the left side of the platform page, select the target data center. Then, on the right side of the platform page, click Actions > New Cluster, or on the Cluster and Host sub-page, click New Cluster.

ZStack ZSphere supports the following major categories of information configuration:

Basic Information Configuration: Includes name and description, associated data center, CPU architecture, and dynamic resource scheduling configuration

Name: Name of the cluster
Description: Description of the cluster
Data Center: Data center where the cluster is located
CPU Architecture: Supports creating x86 or aarch architecture clusters. The architecture of hosts added to the cluster must match that of the cluster.
DRS: Monitors CPU/memory load of all hosts in the cluster and dynamically adjusts the virtual machine services running on hosts based on scheduling policies. This feature is disabled by default. If enabled, manual scheduling and automatic scheduling are supported. For details, refer to Cluster DRS

Advanced Configuration: Includes cluster networking, cluster over-provisioning, host settings within the cluster, and virtual machine settings within the cluster:

Cluster Network: Configure dedicated VDI network CIDR and VM migration network CIDR for the cluster. If not set or ineffective, the management network is used by default.
Cluster Overcommitment: Controls the number of virtual CPUs and the amount of virtual memory allocated to virtual machines:
- Allocatable Virtual CPUs = Physical CPU Total Threads in the Cluster × CPU Overcommit Ratio
- Allocatable Virtual Memory Capacity = (Physical Memory Capacity − Reserved Capacity) × Memory Overcommit Ratio
Host Settings: Customize host characteristics within the cluster:
- Host CPU Model Check: Whether the system checks for consistency between the CPU model of the source host and the CPU models of hosts in the current cluster during live migration or host addition. Default is disabled. If enabled, inconsistent models prevent the corresponding operation.
- ignore_msrs Option: Whether the ignore_msrs option is enabled in the host KVM kernel module. Default is disabled.
- Host Zero Copy: Whether the host CPU enables Zero Copy. Default is not enabled. Enabling reduces the number of copies of data between kernel space and user space, reducing CPU usage time and improving virtual NIC performance.
- Huge Pages: Whether hosts in the cluster enable huge pages (each page is 2MB) and allocate huge page space to virtual machines. Default is not enabled.
- Host Reserved Memory: The amount of memory reserved on the KVM host to ensure the host system can run normally. When huge pages are enabled in the cluster, hosts must reserve at least 4GB of memory.
Virtual Machine Settings: Customize virtual machine characteristics within the cluster:
- VM HA: Automatic restart mechanism for virtual machines after shutdown. Default is disabled. Details can be found in VM HA.
- VM Cross-Cluster HA: In shared storage (NFS, SAN, and distributed storage) scenarios, whether high availability migration across clusters is supported when changing hosts or when hosts enter maintenance mode.
- VM CPU Model: Whether the virtual machine CPU model is set to match the host CPU model, inheriting some or all of the host's CPU features to meet different business needs. Default is unset (none) for x86_64 architecture, and passthrough for aarch architecture.
- Hyper-V Virtualization: Whether the virtual machine enables Hyper-V emulation functionality. Default is disabled. Mainly used for nested virtualization scenarios with Windows systems.
- Video Card Type on Boot: The default graphics card type used when starting virtual machines, providing basic/high-definition/high-performance video functionality. Default is vga for x86 architecture; virtio is the default and only option for aarch architecture.
- vNIC Multi-queue Upgrading: Uses multiple queues for virtual machine NICs to receive and transmit network packets to improve network PPS and bandwidth performance.

After clicking OK, the cluster will be created.

Edit a Cluster

If you need to modify the name or description of an existing cluster, you can do so on the target cluster page by clicking ActionsEdit Name and Description. In the dialog box that appears, you can modify the corresponding information.

If you need to modify advanced settings of an existing cluster, such as cluster networking settings, over-provisioning settings, host settings within the cluster, or virtual machine settings, you can do so on the target cluster page by clicking ActionsModify Configuration, or on the Advanced Settings sub-page.

View Cluster Capacity Information

If you need to check the usage and allocation of physical CPU and memory resources for the cluster, as well as the usage and allocation of all sub-resources under the cluster, go to the Overview details page. For more information, please refer to Capacity Monitoring.

Delete a Cluster

If you need to delete an existing cluster, you can do so on the target cluster page by clicking ActionsDelete. You can also delete clusters in bulk on the Cluster and Host page within the data center resources.

Note: Deleting a cluster deletes all hosts within the cluster. If local storage is loaded, it will also delete all virtual machines and snapshots on the hosts. Proceed with caution.

Cluster DRS

Dynamic Resource Scheduling (DRS): Monitors CPU or memory load on hosts at the cluster level and dynamically adjusts virtual machine services running on hosts according to configured scheduling policies.

ZStack ZSphere supports both manual and automatic scheduling strategies, both of which can balance cluster loads and effectively improve platform stability:

Manual scheduling strategy provides scheduling suggestions, allowing you to manually migrate virtual machines according to the suggestions.
Automatic scheduling strategy, where the system automatically performs resource scheduling based on scheduling algorithms.

The basic process for using dynamic resource scheduling is:

Configure the dynamic resource scheduling policy.
Execute dynamic resource scheduling operations.

Configure DRS

Prerequisites: To enable or use the dynamic resource scheduling feature, ensure that the cluster meets the following conditions:

The cluster only attaches ZCE distributed storage or SAN storage.
All hosts in the cluster have consistent CPU models.

Configure Dynamic Resource Scheduling:

The platform provides multiple entry points for configuring dynamic resource scheduling policies. You can configure dynamic resource scheduling policies from the following two main entry points:

In the target cluster DRS tab, click Enable to configure the settings.
In the navigation bar on the left side of the platform page, click Business Reliability > DRS Policy, then in the target cluster, click Actions > Modify Policy to configure the settings.

You can configure dynamic resource scheduling policies as follows:

Resource Type: Default is compute resources, currently not modifiable.
Scheduling Policy: Supports manual and automatic scheduling modes:
- Manual Scheduling: After the CPU utilization or memory utilization of hosts in the cluster reaches the specified threshold, you manually execute resource scheduling based on the scheduling suggestions.
- Automatic Scheduling: After the CPU utilization or memory utilization of hosts in the cluster reaches the specified threshold, the system automatically executes resource scheduling based on scheduling algorithms.
Monitoring Items: Select the monitoring items for hosts, including: CPU utilization, memory utilization, CPU or memory utilization
- CPU Utilization: Define the trigger conditions for the CPU utilization monitoring item.
- Memory Utilization: Define the trigger conditions for the memory utilization monitoring item.
Duration: Define the duration for the threshold, units include: seconds, minutes, hours.
DRS VM Migration Concurrency: Number of virtual machines concurrently migrated during dynamic resource scheduling. Default is 1.
DRS Cluster Scan Interval: Time interval for scanning the cluster balance status during dynamic resource scheduling. Default is 10 minutes.

Note: If any host in the cluster reaches the threshold for the monitoring item trigger condition and meets the duration, it can be determined that the cluster state is imbalanced, and scheduling suggestions will be given.

Execute Dynamic Resource Scheduling Operations

After configuring dynamic resource scheduling, you can perform the following operations:

Manually scan the cluster balance status.
If set to manual scheduling, you can migrate virtual machines to recommended hosts based on scheduling suggestions to balance the cluster load.
You can view scheduling execution history, results, and times in O&M Management > Tasks > Scheduling Tasks. By default, the last 7 days of data are displayed. You can customize the time period to view execution history and search execution history by virtual machine UUID.
You can reconfigure the dynamic resource scheduling policy.
You can disable cluster dynamic resource scheduling.
Note: This may result in an inability to balance host loads in high-load scenarios, affecting business performance. Proceed with caution.

Cluster Basic Operations

Create a Cluster

Edit a Cluster

View Cluster Capacity Information

Delete a Cluster

Cluster DRS

Configure DRS

Execute Dynamic Resource Scheduling Operations

Products

Solutions

Help & Support

Contact Us