Create an instance

Create an instance by selecting a boot source, flavor, network, and security group. These components must be configured before you can launch an instance and are not available by default.

When you create an instance, you choose a boot source that has the bootable operating system that you require for your instance, a flavor that has the hardware profile you require for your instance, the network you want to connect your instance to, and any additional storage you need, such as data volumes and ephemeral storage.

Note
With API microversion 2.94, if you pass an optional hostname when creating, updating, or rebuilding an instance, you can use Fully Qualified Domain Names (FQDN) when you specify the hostname. When using FQDN, ensure that [api]dhcp_domain configuration option is set to the empty string for the correct FQDN to appear in the hostname field in the metadata API. By default, the hostname is normalized from the display name and all occurrences of "." are removed from the hostname and replaced with "_".
 Note

To run openstack client commands on the cloud, you must specify the name of the cloud detailed in your clouds.yaml file. You can specify the name of the cloud by using one of the following methods:

  • Use the --os-cloud option with each command, for example: 
    $ openstack flavor list --os-cloud <cloud_name>

    Use this option if you access more than one cloud.

  • Create an environment variable for the cloud name in your bashrc file: 
    `export OS_CLOUD=<cloud_name>`

Prerequisites
Copy link

  • The required image or volume is available as the boot source.
  • A flavor is available that specifies the required number of CPUs, memory, and storage capacity. The flavor settings must meet the minimum requirements for disk and memory size specified by your chosen image, otherwise the instance will fail to launch.
  • The required network is available.

Create an instance from an image
Copy link

Create an instance by using an image as the boot source. The image provides the bootable operating system.

Before you begin
Copy link

  • The administrator has created a project for you and they have provided you with a clouds.yaml file for you to access the cloud.
  • You have installed the python-openstackclient package.

About this task
Copy link

Note
With API microversion 2.94, if you pass an optional hostname when creating, updating, or rebuilding an instance, you can use Fully Qualified Domain Names (FQDN) when you specify the hostname. When using FQDN, ensure that [api]dhcp_domain configuration option is set to the empty string for the correct FQDN to appear in the hostname field in the metadata API. By default, the hostname is normalized from the display name and all occurrences of "." are removed from the hostname and replaced with "_".

Procedure
Copy link

  1. Retrieve the name or ID of the flavor that has the hardware profile that your instance requires: 
    $ openstack flavor list
    Note
    Choose a flavor with sufficient size for the image to successfully boot, otherwise the instance will fail to launch.
  2. Retrieve the name or ID of the image that has the software profile that your instance requires: 
    $ openstack image list

    If the image that you require is not available, you can download or create a new image. For information about how to create or download cloud images, see Virtual machine image management.

    Note

    If you need to attach more than 26 volumes to your instance, then the image you use to create your instance must have the following properties:

    • hw_scsi_model=virtio-scsi
    • hw_disk_bus=scsi
  3. Retrieve the name or ID of the network that you want to connect your instance to: 
    $ openstack network list
  4. Create your instance: 
    $ openstack server create --flavor <flavor> \
  --image <image> --network <network> \
  --wait myInstanceFromImage
    • Replace <flavor> with the name or ID of the flavor that you retrieved in step 1.
    • Replace <image> with the name or ID of the image that you retrieved in step 2.
    • Replace <network> with the name or ID of the network that you retrieved in step 3. You can use the --network option more than once to connect your instance to several networks, as required.

Create an instance from a bootable volume
Copy link

You can create an instance by using a bootable volume as the boot source. Boot your instance from a volume when you need to improve the availability of the instance data if a failure occurs.

Before you begin
Copy link

  • The administrator has created a project for you and they have provided you with a clouds.yaml file for you to access the cloud.
  • You have installed the python-openstackclient package.

About this task
Copy link

Note
When you use a block storage volume for your instance disk data, the block storage volume persists for any instance rebuilds, even when an instance is rebuilt with a new image that requests that a new volume is created.
 Note
With API microversion 2.94, if you pass an optional instance hostname when creating, updating, or rebuilding a server, you can use Fully Qualified Domain Names (FQDN) when you specify the hostname. When using FQDN, ensure that [api]dhcp_domain configuration option is set to the empty string for the correct FQDN to appear in the hostname field in the metadata API. By default, the hostname is normalized from the display name and all occurrences of "." are removed from the hostname and replaced with "_".

Procedure
Copy link

  1. Retrieve the name or ID of the image that has the software profile that your instance requires: 
    $ openstack image list

    If the image that you require is not available, you can download or create a new image. For information about how to create or download cloud images, see Download a RHEL KVM instance image.

    Note

    If you need to attach more than 26 volumes to your instance, then the image you use to create your instance must have the following properties:

    • hw_scsi_model=virtio-scsi
    • hw_disk_bus=scsi
  2. Create a bootable volume from the image: 
    $ openstack volume create --image <image> \
--size <size_gb> --bootable myBootableVolume
    • Replace <image> with the name or ID of the image to write to the volume, retrieved in step 1.
    • Replace <size_gb> with the size of the volume in GB.
  3. Retrieve the name or ID of the flavor that has the hardware profile that your instance requires: 
    $ openstack flavor list
  4. Retrieve the name or ID of the network that you want to connect your instance to: 
    $ openstack network list
  5. Create an instance with the bootable volume: 
    $ openstack server create --flavor <flavor> \
 --volume myBootableVolume --network <network> \
 --wait myInstanceFromVolume
    • Replace <flavor> with the name or ID of the flavor that you retrieved in step 3.
    • Replace <network> with the name or ID of the network that you retrieved in step 4. You can use the --network option more than once to connect your instance to several networks, as required.

Create an instance with an SR-IOV network interface
Copy link

To create an instance with a single root I/O virtualization (SR-IOV) network interface, you must create the required SR-IOV port.

Before you begin
Copy link

  • The administrator has created a project for you and they have provided you with a clouds.yaml file for you to access the cloud.
  • You have installed the python-openstackclient package.

About this task
Copy link

Note
With API microversion 2.94, if you pass an optional hostname when creating, updating, or rebuilding an instance, you can use Fully Qualified Domain Names (FQDN) when you specify the hostname. When using FQDN, ensure that the [api]dhcp_domain configuration option is set to the empty string for the correct FQDN to appear in the hostname field in the metadata API. By default, the hostname is normalized from the display name and all occurrences of "." are removed from the hostname and replaced with "_".

Procedure
Copy link

  1. Retrieve the name or ID of the flavor that has the hardware profile that your instance requires: 
    $ openstack flavor list
    Note
    Choose a flavor with sufficient size for the image to successfully boot, otherwise the instance will fail to launch.
     Tip
    You can specify the NUMA affinity policy that is applied to your instance for PCI passthrough devices and SR-IOV interfaces, by selecting a flavor that has the policy you require. For more information on the available policies, see Instance PCI NUMA affinity policy in Flavor metadata. If you choose a flavor with a NUMA affinity policy, then the image that you use must have either the same NUMA affinity policy or no NUMA affinity policy.
  2. Retrieve the name or ID of the image that has the software profile that your instance requires: 
    $ openstack image list

    If the image that you require is not available, you can download or create a new image. For information about how to create or download cloud images, see Download a RHEL KVM instance image.

    Tip
    You can specify the NUMA affinity policy that is applied to your instance for PCI passthrough devices and SR-IOV interfaces, by selecting an image that has the policy you require. For more information on the available policies, see Instance PCI NUMA affinity policy inFlavor metadata. If you choose an image with a NUMA affinity policy, then the flavor that you use must have either the same NUMA affinity policy or no NUMA affinity policy.
  3. Retrieve the name or ID of the network that you want to connect your instance to: 
    $ openstack network list
  4. Create the type of port that you require for your SR-IOV interface: 
    $ openstack port create --network <network> \
 --vnic-type <vnic_type> mySriovPort
    • Replace <network> with the name or ID of the network you retrieved in step 3.
    • Replace <vnic_type> with the one of the following values:
      • direct: Creates a direct mode SR-IOV virtual function (VF) port.
      • direct-physical: Creates a direct mode SR-IOV physical function (PF) port.
      • macvtap: Creates an indirect mode SR-IOV VF port that uses MacVTap to expose the virtio interface to the instance.
  5. Create your instance: 
    $ openstack server create --flavor <flavor> \
  --image <image> --port <port> \
  --wait mySriovInstance
    • Replace <flavor> with the name or ID of the flavor that you retrieved in step 1.
    • Replace <image> with the name or ID of the image that you retrieved in step 2.
    • Replace <port> with the name or ID of the port that you created in step 4.

Create an instance with NUMA affinity on the port
Copy link

To create an instance with NUMA affinity on the port, create the port with the required NUMA affinity policy, then specify it when launching.

Before you begin
Copy link

  • The port_numa_affinity_policy extension must be enabled in the Networking service (neutron).
  • The administrator has created a project for you and they have provided you with a clouds.yaml file for you to access the cloud.
  • You have installed the python-openstackclient package.

About this task
Copy link

Note
Port NUMA affinity policies have a higher precedence than flavors, images, and PCI NUMA affinity policies. Cloud operators can set a default NUMA affinity policy for each PCI passthrough device. You can use the instance flavor, image, or port to override the default NUMA affinity policy applied to an instance.
 Note
With API microversion 2.94, if you pass an optional hostname when creating, updating, or rebuilding an instance, you can use Fully Qualified Domain Names (FQDN) when you specify the hostname. When using FQDN, ensure that [api]dhcp_domain configuration option is set to the empty string for the correct FQDN to appear in the hostname field in the metadata API. By default, the hostname is normalized from the display name and all occurrences of "." are removed from the hostname and replaced with "_".

Procedure
Copy link

  1. Verify that the port_numa_affinity_policy extension is enabled in the Networking service (neutron): 
    $ oc get openstackcontrolplane -o yaml | grep -A 10 "neutron:"

    If the port_numa_affinity_policy extension is not listed in the extension_drivers configuration, enable it by adding port_numa_affinity_policy to extension_drivers in the Neutron definition of OpenStackControlPlane:

    neutron:
  template:
    customServiceConfig: |
      [ml2]
      extension_drivers=qos,port_security,dns_domain_ports,port_numa_affinity_policy
    Note

    The configuration method differs when you use a third-party Neutron mechanism driver.
  2. Create a port with the NUMA affinity policy that you require: 
    $ openstack port create --network <network> \
  [--numa-policy-required | --numa-policy-preferred | --numa-policy-legacy] \
  myNUMAAffinityPort
    • Replace <network> with the name or ID of the tenant network that you want to connect your instance to.
    • Use one of the following options to specify the NUMA affinity policy to apply to the port:
      • --numa-policy-required - NUMA affinity policy required to schedule this port.
      • --numa-policy-preferred - NUMA affinity policy preferred to schedule this port.
      • --numa-policy-legacy - NUMA affinity policy using legacy mode to schedule this port.
  3. Create your instance: 
    $ openstack server create --flavor <flavor> \
  --image <image> --port <port> \
  --wait myNUMAAffinityInstance
    • Replace <flavor> with the name or ID of the flavor that has the hardware profile that you require for your instance.
    • Replace <image> with the name or ID of the image that has the software profile that you require for your instance.
    • Replace <port> with the name or ID of the port that you created in step 1.