A host provides compute, network, and storage resources
for VM instances.
Hosts are core resources on the Cloud. VM instances are running on hosts.
Figure 1. Host
Add a Host
On the main menu of ZStack Cube Ultimate, choose Resource Center > Hardware > Computing Facility > Host. On the Host page, click Add
Host. Then, the Add Host page is displayed.
ZStack Cube Ultimate allows you to add a host by one of the following
methods:
Manual addition
Import template
Manual Addition
With this method, you can specify an IP address or an IP range to add hosts. You can
add up to 500 hosts at a time. To use this method, set the following parameters:
Name: Enter a name for the host.
The name must be 1 to 128 characters in length and can
contain Chinese characters, letters, digits, spaces, hyphens (-), underscores
(_), periods (.), parenthesis (), colons (:), and plus signs (+) and cannot
begin or end with spaces.
Description: Optional. Enter a description for the
host.
Virtualization Technology: Select the hypervisor type
of the server. You can select KVM or X-Dragon.
If the server applies the KVM virtualization technology, select
KVM.
If the server applies the X-Dragon architecture, select
X-Dragon.
Tag: Optional. Attach one or more tags to the
host.
Cluster: Select the cluster where the host is located.
Note:
Hosts that apply the KVM virtualization technology can be added
only to KVM-based clusters.
You can enable network acceleration support for a KVM-based
cluster. If you add a host to a cluster for which network
acceleration support is enabled, note that:
If the L2 network associated with the cluster uses the
Smart NIC network acceleration mode, the host must be
attached with a smart NIC of the specified model.
Hosts of the X-Dragon architecture can be added only to
X-Dragon-based clusters.
X-Dragon-based clusters do not support network acceleration
support.
Host Info: Configure the host information such as the
host IP and SSH port according to your actual needs. You can configure the
host information by entering an IP address or an IP range.
If you select the IP Address method, set the
following parameters:
IP Address: Enter the IP address of
the host.
SSH Port: Enter the SSH port of the
host. Default: 22.
SSH Username: Enter the SSH user name
of the host. Default: root. You can enter a user name
according to your actual needs.
SSH Password: Enter an SSH password
for the user name.
If you select the IP Range method, set the
following parameters:
IP Range: Enter the planed IP range
for the host.
SSH Port: Enter the SSH port for the
host. Default: 22.
Note: Make sure that all hosts in the
specified IP range share the same SSH port.
SSH Username: Enter the user name of
the host. Default: root. You can enter a user name according
to your actual needs.
Note: Make sure that all hosts in the
specified IP range share the same user name.
SSH Password: Enter a password for
the user name.
Note: Make sure that all user name in the
specified IP range share the same password.
Scan Host IOMMU Setting: Optional. Choose whether to
scan the IOMMU setting of the host. The IOMMU setting is used in scenarios
such as peripheral device passthrough and virtualization. By default, the
host IOMMU setting is not scanned.
Note: If you choose to scan the host IOMMU
setting, note the following:
Scanning the host IOMMU setting will traverse available GPU
devices on the host and physical NICs that can be virtualized.
Before you scan the setting, make sure that Intel VT-d or AMD
IOMMU is enabled in the host BIOS.
Before you can use GPU passthrough, vGPU virtualization, and
SR-IOV, scan the host IOMMU setting.
If you are scanning the host IOMMU setting for the first time,
restart the host to make the setting take effect.
Intel EPT Hardware Assist: Optional. Choose whether
to enable Intel EPT hardware assist for Intel CPUs to improve the CPU
performance. By default, Intel EPT hardware assist is enabled.
Note:
If the CPU model of your Intel server is too old, you might fail
to create VM instances or cannot open the console properly of
the newly created VM instances. In this case, disable the Intel
EPT hardware assist feature.
Disabling Intel EPT Hardware Assist will lower the VM
performance.
Figure 2. Manual Addition
Import Template
With this method, download the CSV template file provided by the Cloud, fill in the
host information according to the specified format, and upload the file. You can add
hosts in bulk with this method. The procedures are as follows:
Download the configuration template.
Click Download
Template to download the configuration template file in
CSV format.
Figure 3. Configuration Template File
Fill in the host information according to the specified format.
The
configuration template contains a header and a row of sample. You can
delete or overwrite the sample as needed.
Set the following
parameters in the template:
Name: Enter a name for the host. If null,
the Cloud allocates a name (HOST-host IP address) to the host by
default.
Description: Optional. Enter a
description for the host.
Cluster: Enter the UUID of the cluster
where the host is located.
Note: You can add only KVM-based
clusters in the template.
Host IP: Enter an IP address or an IP
range for the host.
IP Address: Enter the specified IP address of the
host.
IP Range: Enter the specified IP range to bulk add hosts.
Note:
If you enter multiple IP ranges for the host, use
commas (,) to separate the ranges. ^ means
that the IP range is not included. For
example,
Scan Host IOMMU Setting: Optional. Choose
whether to scan the IOMMU setting of the host. The IOMMU setting
is used in scenarios such as peripheral device passthrough and
virtualization. By default, the host IOMMU setting is not
scanned.
Note: If you choose to scan the host IOMMU setting,
note the following:
To enable this feature, enter YES, Yes, yes, Y, or
y. To disable this feature, enter NO, No, no, N, or
n. If null, this feature is disabled.
Before you can use GPU passthrough, vGPU
virtualization, and SR-IOV, scan the host IOMMU
setting.
Scanning the host IOMMU setting will traverse
available GPU devices on the host and physical NICs
that can be virtualized. Before you scan the
setting, make sure that Intel VT-d or AMD IOMMU is
enabled in the host BIOS.
If you are scanning the host IOMMU setting for the
first time, restart the host to make the setting
take effect.
Intel EPT Hardware Assist: Optional.
Specify whether to enable Intel EPT hardware assist for Intel
CPUs to improve the CPU performance.
Note: When you set this
parameter, note the following:
If you enter YES, Yes, yes, Y, or y, the Intel EPT
hardware assist will be enabled. If you enter NO,
No, no, N, or n, the Intel EPT hardware assist will
be disabled. If null, this feature is disabled.
If the CPU model of your Intel server is too old,
you might fail to create VM instances or cannot open
the console properly of the newly created VM
instances. In this case, disable the Intel EPT
hardware assist feature.
Disabling Intel EPT Hardware Assist will lower the
VM performance.
SSH Port: Enter the SSH port for the
host. If null, the Cloud will use port 22 by default.
SSH User Name: Enter the SSH user name of
the host.
SSH Password: Enter an SSH password for
the user name.
Upload the configuration file.
After you fill in the host information,
verify that the grammar is correct, and then upload the file to the
Cloud.
Figure 4. Import Template
Add hosts by using the template.
Click OK. Then,
the Cloud add hosts according to the configuration file.
Manage a Host
On the main menu of ZStack Cube Ultimate, choose Resource Center > Hardware > Computing Facility > Host. Then, the Host page appears.
The following table lists the actions that you can perform on a host.
Action
Description
Preconditioned Host Status
Add Host
Add one or more hosts.
/
Edit Host
Edit the name and description of a host.
/
Enable Host
Enable a host.
Disabled
Disable Host
Disable a host.
Note: Disabling a host does not
affect the resources already running on the host. However, the
host is no longer available as a candidate when you apply for
new resources.
Enabled
Reconnect Host
Reconnect a host.
Note: In most cases, host
reconnection can be implemented when the host configurations are
updated. For example, if the memory or volume capacity of a hots
is updated, you can reconnect the host to update the
database.
ALL
Enter Maintenance Mode
Enter the maintenance mode. If a host is in
maintenance mode, you can power off the host or troubleshoot
failures.
If the primary storage is of the LocalStorage type, VM
instances on a host are stopped when the host enters
maintenance mode.
If the primary storage is of a shared storage type, VM
instances on a host are automatically migrated when the host
enters maintenance mode.
Note: If the primary storage is of a shared storage type, you can configure
the policy on failures of VM migration triggered when the host
enters maintenance mode.
On the main menu of ZStack Cube Ultimate, choose Settings > Platform Setting > Global Setting > Advanced > Host, and configure the Policy for VM Migration
Failure in Entering Host Maintenance Mode
setting:
StopVmOnMigrationFailure: When the host
enters maintenance mode, VM instances are
automatically migrated to another host. If a VM
instance fails to be migrated, the VM instance is
forcibly stopped.
JustMigrate: When the host enters maintenance
mode, VM instances are automatically migrated to
another host. If a VM instance still runs on the
host, the host fails to enter the maintenance
mode.
Connected
Add Bond
Aggregates the physical NIC ports on the host, so
as to realize a high availability or a load balance. Supports two
bond mode: active-backup and LACP.
The active-backup mode allows you to bind 1-2 physical NIC
ports, and we recommend that you bind 2 ports, one acting as
the primary port and the other as the secondary port. By
default, all network flows are handled by the primary port.
Once the primary port fails, the secondary port replaces it
to handle network flows.
The LACP mode allows you to bind 1-8 physical NIC ports, and
we recommend that you bind at least 2 ports. Bound ports
share the same speed and duplex setting. The network flows
are evenly sent to these ports to handle, thus realizing a
load balance. A bond of this mode determines its network
export based on a hash computation. Three hash policies are
supported: layer2+3, layer 3+4, and layer2.
layer2+3: Picks out a port to send data packets
based on the hash computation on the source MAC
address, destination MAC address, and IP
address.
layer3+4: Picks out a port to send data packets
based on the hash computation on the IP address and
port. TCP/IP stacks are supported.
layer2: Picks out a port to send data packets based
on the hash computation on the source MAC address
and destination MAC address.
Note: Make sure that the NIC ports added to a same bond have a
same speed rate.
ALL
Power on
Power on a host.
Powered off
Power off
Power off a host.
Note:
Powering off a host may cause VM instances, VPC
vRouters, SLB instances, and other resources running on
the host powered off. To ensure your business stability
and security, we recommend that you make the host enter
the maintenance mode before you perform the
operation.
If the host is used also as a Monitor node of Ceph
primary storage, powering it off will stop the monitor
node service of the primary storage, affect the
stability of the storage cluster, and thus may cause
data loss.
If the host is designated as a migration server,
powering it off will make migration jobs running on the
host end up failure.
Powered on/Unknown
Restart
Restart a host.
Note:
Restarting a host may affect normal running of VM
instances, VPC vRouters, SLB instances, and other
resources on the host. To ensure your business stability
and security, we recommend that you make the host enter
the maintenance mode before you perform the
operation.
If the host is used also as a Monitor node of Ceph
primary storage, restarting it will stop the monitor
node service of the primary storage, affect the
stability of the storage cluster, and thus may cause
data loss.
If the host is designated as a migration server,
restarting it will make migration jobs running on the
host end up failure.
Powered on/Unknown
Enter Web Terminal
Enter a web terminal of a host to perform
operations on the host.
Note: The following reasons may make you fail
to enter the Web Terminal of a host:
The host architecture is ARM.
The host system does not support this action or a system
error occurs.
Powered on/Unknown
Attach Tag
Attach a tag to one or more hosts.
Note:
You can attach a maximum of 50 tags to a resource. The
number of resources to which a tag can be attached is
not limited.
You can attach one or more tags to one or more
resources.
A tag created by a tenant can be attached only to the
resources owned by the tenant. The admin can attach a
tag to all resources.
The admin can detach or delete tags created by a
tenant.
You can sort tags attached to resources by attachment
time or tag name. Tag names are sorted based on symbols,
digits, Chinese characters, and English letters in
descending order. To configure the sorting method, go to
the main menu of ZStack Cube Ultimate,
choose Settings > Platform Setting > Global Setting > Basic, and then configure the Tag Sort
Method setting. By default, resources are sorted
by tagName.
ALL
Detach Tag
Detach a tag from a host.
Note:
The admin tags are created and owned by the admin. The
tenant tags are created and owned by tenants.
A tag created by a tenant can be attached only to the
resourced owned by the tenant. The admin can attach a
tag to all resources.
The admin can detach or delete tags created by a
tenant.
Tags created in a project belongs to the
project. All project members, including the project
admin, project managers, and normal project members, can
perform actions on the tags.
You cannot change the owner of a tag.
If you change the owner of a resource, all tenant tags
attached to the resource are detached. The admin tags
are not affected.
After a seamless upgrading of the cloud platform, tags
that were created are automatically updated and
displayed in the latest format. If an error or exception
occurs, refresh the browser or create a new tag.
You can detach a batch of tags from a resource or detach
a tag from a batch of resources.
A tenant can only detach tags from resources owned by
the tenant. The admin can detach tags from all
resources.
ALL
Modify IPMI Info
Modify the IPMI username and password of the
host.
Note: You can only modify the IPMI info of a host that is
in IPMI Unmanaged or Unknown power status.
Unknown
Update SSH Password
Update the SSH password of a host. After the
password is updated, the host is automatically reconnected.
/
Delete Host
Delete a host.
Note: If you delete a host, note
that:
Deleting a host stops all VM instances running on the
host.
If a LocalStorage primary storage is attached to the
cluster to which the host belongs:
Deleting a host also deletes all VM instances
and volumes on the host.
If you add the deleted host to the cloud
platform, the system redeploys the host. If you
did not backup the database, the related business
data cannot be recovered.
If a shared primary storage is attached to the cluster
to which the host belongs:
If you disable HA for VM instances:
Deleting a
host stops the VM instances running on the
host.
If you enable HA for VM instances:
Deleting a
host stops the VM instances for which HA is
enabled and then migrates the VM instances to
other hosts where resources are available. Then
the VM instances are automatically started. During
the process, data security is not affected. If
host resources are insufficient, the VM instances
are stopped upon the host deletion.
Deleting a host does not affect data volumes
stored on shared storage.
/
Host Details
Host Hardware Monitoring
In the ZStack Cube Ultimate main menu, click Resource Center > Hardware Facility > Compute Facility. Click a host name to go to the host details page, and then click the Hardware Monitoring subpage.
This page displays hardware configuration and monitoring information of the host in lists.
Summary: Displays CPU, memory, disk, fan, power supply, temperature sensor, vendor, model, SN, BIOS, and other information.
CPU: Displays CPU model, status, architecture, logical cores, current frequency, current temperature, and cache information.
Memory: Displays available slots, memory slot information, running status, capacity, type, vendor, current speed, and maximum speed.
Disk: Displays disk type, slot number, purpose, status, SSD remaining life, indicator light operations, interface type, capacity, temperature, and HDD rotation speed. Disk statuses include normal, abnormal, offline, and rebuilding. The rebuilding status applies only to disks configured with RAID.
Physical NIC: Displays detected physical NICs. You can passthrough NICs to VM instances, turn indicator lights on or off, and locate physical NICs for troubleshooting.
Fan, Power Supply, and Temperature Sensor: Displays component status and monitoring information.
Note: After turning on a disk or NIC indicator light in the UI, manually turn it off after locating the hardware.
pNUMA Topology
pNUMA Topology: A pNUMA topology (physical NUMA topology) is the topology
of the host NUMA nodes predefined by the CPU vendor based on the host NUMA
architecture.
Concepts
NUMA: Non-uniform memory access (NUMA) is a computer memory
design where the memory access time depends on the memory location relative to the
CPU. Under NUMA, a processor can access its own local memory faster than non-local
memory and thus improves VM performance.
pNUMA node: A pNUMA node (physical NUMA node) is a host NUMA node
predefined based on the host NUMA architecture. It is used to manage the CPUs and
memory of the host. A pNUMA node mainly consists of pCPUs and local memory.
vNUMA node: A vNUMA node (virtual NUMA node) is generated by
passing-through associated pNUMA nodes via CPU pinning. It is used to manage the
CPUs and memory of a VM instance. A vNUMA node mainly consists of vCPUs and local memory.
vNUMA topology: A vNUMA topology (virtual NUMA topology) is the topology
of VM NUMA nodes generated by passing-through associated pNUMA nodes via CPU
pinning.
Local memory: Local memory is the memory that a CPU (pCPU or vCPU)
accesses through the Uncore iMC (Integrated Memory Controller) of the same NUMA
(pNUMA or vNUMA) node. Compared with accessing non-local memory, accessing local
memory has lower latencies.
Fundamentals
After you add a host on ZStack Cube Ultimate, you can view the pNUMA
topology of the host. You can also configure vNUMA for VM instances running on the
host according to the pNUMA topology.
vNUMA configurations are achieved through CPU pinning, which
assigns the vCPUs of a VM instance to specific pCPUs of the host. In vNUMA
configurations, all vCPUs of the VM instance are pinned to pCPUs of the host. In
addition, the pCPUs pinned by a vCPU belong to the same pNUMA node.
After the vNUMA configurations are completed, a vNUMA topology
with one or more vNUMA nodes is generated by passing-through the topology of
associated host pNUMA nodes. Then the vCPUs of the VM instance primarily access the
local memory of the vNUMA node where the vCPUs reside.
Figure 5. vNUMA Configuration Basics
pNUMA Topology
The following describes the pNUMA topology of a host on ZStack Cube Ultimate:
All pNUMA nodes of the host and the VM instances associated with each pNUMA
node are displayed.
Total Memory: The total amount of local memory
that a pCPU can access from the pNUMA node where it resides.
Free Memory: The amount of free local memory
that a pCPU can access from the pNUMA node where it resides.
Both the total memory and free memory are
measured based on physical hardware resources.
Figure 6. pNUMA Topology
You can click the Redirect icon to view the vNUMA topologies of the associated VM
instances. For more information about vNUMA topologies, see vNUMA Topology.
Host QEMU Version
Version Requirement
All hosts in a same cluster should be installed with QEMU of the same
version. If you need to add a host with a QEMU version different from those
of other hosts in the cluster, update the QEMU of the hosts in this cluster
to the Cloud recommended version or create a new cluster to add this new
host.
To avoid function or security patch losses, we recommend that all hosts use
QEMU of the recommended version.
View Host QEMU Version
ZStack Cube Ultimate monitors the QEMU versions of hosts in a cluster.
When a part of the host use QEMU that is not of the Cloud recommended version, a
prompt appears in the cluster details page to inform you to update the host QEMU
version. You can view the QEMU version state of each host by clicking
Check in the prompt. The QEMU state is displayed as To
be Updated if it is not of the recommended version.Figure 7. View Host QEMU Version
Update Host QEMU Version
ZStack Cube Ultimate allows you to update the QEMU version for all
hosts in a cluster. Note that the QEMU update is a risky action. Before you execute
the QEMU update, contact our technical support to evaluate the risk level.
Updating host QEMU version mainly contains following steps:
Check the connection of Cloud services and hardware.
Set VM HA mode as None in Global Setting.
Query the host QEMU version.
Back up the database of the two management nodes.
Upload the latest ISO image to the two management nodes and update the local
repo source.
Use CLI commands to updates QEMU version for hosts in the cluster.
Use CLI commands to query the update process and confirm that the update is
successful.
Note: For detailed update commands, contact our technical supports.
Considerations
After the update, restart or migrate the VM instances, VPC vRouters, and LB
instances on the hosts to make the update take effect.
When the QEMU version of a VM instance, VPC vRouter, or LB instance is not
consistent with that of the host it resides on, its QEMU state is displayed
as To be Updated. You can restart or migrate the VM/VPC vRouter/LB Instance
to update its QEMU version.
Figure 8. VM Instance QEMU State Figure 9. VPC vRouter QEMU State Figure 10. LB Instance QEMU State
Note:
