Red Hat Training

A Red Hat training course is available for Red Hat OpenStack Platform

Bare Metal Provisioning

Red Hat OpenStack Platform 11

Install, Configure, and Use the Bare Metal Service (Ironic)

OpenStack Documentation Team

Abstract

This guide provides procedures for installing, configuring, and using the Bare Metal service in the overcloud of a Red Hat OpenStack Platform environment.

Preface

This document provides instructions for installing and configuring the Bare Metal service (ironic) in the overcloud, and using the service to provision and manage physical machines for end users.

The Bare Metal service components are also used by the Red Hat OpenStack Platform director, as part of the undercloud, to provision and manage the bare metal nodes that make up the OpenStack environment (the overcloud). For information on how the director uses the Bare Metal service, see Director Installation and Usage.

Chapter 1. About the Bare Metal Service

The OpenStack Bare Metal service (ironic) provides the components required to provision and manage physical machines for end users. The Bare Metal service in the overcloud interacts with the following OpenStack services:

  • OpenStack Compute (nova) provides scheduling, tenant quotas, IP assignment, and a user-facing API for virtual machine instance management, while the Bare Metal service provides the administrative API for hardware management.
  • OpenStack Identity (keystone) provides request authentication and assists the Bare Metal service in locating other OpenStack services.
  • OpenStack Image service (glance) manages images and image metadata.
  • OpenStack Networking (neutron) provides DHCP and network configuration.
  • OpenStack Object Storage (swift) is used by certain drivers to expose temporary URLs to images.

The Bare Metal service uses iPXE to provision physical machines. The following diagram outlines how the OpenStack services interact during the provisioning process when a user launches a new machine with the default drivers.

The PXE Provisioning Process

Chapter 2. Planning for Bare Metal Provisioning

This chapter outlines the requirements for setting up the Bare Metal service, including installation assumptions, hardware requirements, and networking requirements.

2.1. Installation Assumptions

This guide assumes you have installed the director on the undercloud node, and are ready to install the Bare Metal service along with the rest of the overcloud. For more information on installing the director, see Installing the Undercloud.

Note

The Bare Metal service in the overcloud is designed for a trusted tenant environment, as the bare metal nodes can access the control plane network of your OpenStack installation.

2.2. Hardware Requirements

Overcloud Requirements

The hardware requirements for an overcloud with the Bare Metal service are the same as for the standard overcloud. For more information, see Overcloud Requirements in the Director Installation and Usage guide.

Bare Metal Machine Requirements

The hardware requirements for bare metal machines that will be provisioned vary depending on the operating system you are installing. For Red Hat Enterprise Linux 7, see the Red Hat Enterprise Linux 7 Installation Guide. For Red Hat Enterprise Linux 6, see the Red Hat Enterprise Linux 6 Installation Guide.

All bare metal machines to be provisioned require the following:

  • A NIC to connect to the bare metal network.
  • A power management interface (for example, IPMI) connected to a network reachable from the ironic-conductor service. If you are using the SSH driver for testing purposes, this is not required. By default, ironic-conductor runs on all of the controller nodes, unless you are using composable roles and running ironic-conductor elsewhere.
  • PXE boot on the bare metal network. Disable PXE boot on all other NICs in the deployment.

2.3. Networking requirements

The bare metal network:

This is a private network that the Bare Metal service uses for:

  • The provisioning and management of bare metal machines on the overcloud.
  • Cleaning bare metal nodes before and between deployments.
  • Tenant access to the bare metal nodes.

The bare metal network provides DHCP and PXE boot functions to discover bare metal systems. This network must use a native VLAN on a trunked interface so that the Bare Metal service can serve PXE boot and DHCP requests.

The bare metal network must reach the control plane network:

The bare metal network must be routed to the control plane network. If you define an isolated bare metal network, the bare metal nodes will not be able to PXE boot.

Note

The Bare Metal service in the overcloud is designed for a trusted tenant environment, as the bare metal nodes have direct access to the control plane network of your OpenStack installation.

Network tagging:

  • The control plane network (the director’s provisioning network) is always untagged.
  • The bare metal network must be untagged for provisioning, and must also have access to the Ironic API.
  • Other networks may be tagged.

Overcloud controllers:

The controller nodes with the Bare Metal service must have access to the bare metal network.

Bare metal nodes:

The NIC which the bare metal node is configured to PXE-boot from must have access to the bare metal network.

2.3.1. The Default Bare Metal Network

In this architecture, the bare metal network is separated from the control plane network. The bare metal network also acts as the tenant network.

  • The bare metal network is created by the OpenStack operator. This network requires a route to the director’s provisioning network.
  • Ironic users have access to the public OpenStack APIs, and to the bare metal network. Since the Bare metal network is routed to the director’s provisioning network, users also have indirect access to the control plane.
  • Ironic uses the bare metal network for node cleaning.

Default bare metal network architecture diagram

Default bare metal network architecture diagram

Chapter 3. Deploying an Overcloud with the Bare Metal Service

For full details about overcloud deployment with the director, see Director Installation and Usage. This chapter only covers deployment steps specific to ironic.

3.1. Creating the Ironic Template

Use an environment file to deploy the overcloud with the Bare Metal service enabled. A template is located on the director node at /usr/share/openstack-tripleo-heat-templates/environments/services/ironic.yaml.

Filling in the template

Additional configuration can be specified either in the provided template or in an additional yaml file, for example ~/templates/ironic.yaml.

  • For a hybrid deployment with both bare metal and virtual instances, you must add AggregateInstanceExtraSpecsFilter to the list of NovaSchedulerDefaultFilters. If you have not set NovaSchedulerDefaultFilters anywhere, you can do so in ironic.yaml. For an example, see Section 3.3, “Example Templates”.

    Note

    If you are using SR-IOV, NovaSchedulerDefaultFilters is already set in tripleo-heat-templates/environments/neutron-sriov.yaml. Append AggregateInstanceExtraSpecsFilter to this list.

  • The type of cleaning that occurs before and between deployments is set by IronicCleaningDiskErase. By default, this is set to ‘full’ by puppet/services/ironic-conductor.yaml. Setting this to ‘metadata’ can substantially speed up the process, as it only cleans the partition table, however, since the deployment will be less secure in a multi-tenant environment, you should only do this in a trusted tenant environment.
  • You can add drivers with the IronicEnabledDrivers parameter. By default, pxe_ipmitool, pxe_drac and pxe_ilo are enabled.

For a full list of configuration parameters, see the section Bare Metal in the Overcloud Parameters guide.

3.2. Network Configuration

Create a bridge called br-baremetal for ironic to use. You can specify this in an additional template:

~/templates/network-environment.yaml

parameter_defaults:
  NeutronBridgeMappings: datacentre:br-ex,baremetal:br-baremetal
  NeutronFlatNetworks: datacentre,baremetal

You can either configure this bridge in the provisioning network (control plane) of the controllers, so you can reuse this network as the bare metal network, or add a dedicated network. The configuration requirements are the same, however the bare metal network cannot be VLAN-tagged, as it is used for provisioning.

~/templates/nic-configs/controller.yaml

network_config:
    -
      type: ovs_bridge
          name: br-baremetal
          use_dhcp: false
          members:
            -
              type: interface
              name: eth1

3.3. Example Templates

The following is an example template file. This file may not meet the requirements of your environment. Before using this example, make sure it does not interfere with any existing configuration in your environment.

~/templates/ironic.yaml

parameter_defaults:

    NovaSchedulerDefaultFilters:
        - RetryFilter
        - AggregateInstanceExtraSpecsFilter
        - AvailabilityZoneFilter
        - RamFilter
        - DiskFilter
        - ComputeFilter
        - ComputeCapabilitiesFilter
        - ImagePropertiesFilter

    IronicCleaningDiskErase: metadata

In this example:

  • The AggregateInstanceExtraSpecsFilter allows both virtual and bare metal instances, for a hybrid deployment.
  • Disk cleaning that is done before and between deployments only erases the partition table (metadata).

3.4. Deploying the Overcloud

To enable the Bare Metal service, include your ironic environment files with -e when deploying or redeploying the overcloud, along with the rest of your overcloud configuration.

For example:

$ openstack overcloud deploy \
  --templates \
  -e ~/templates/node-info.yaml \
  -e /usr/share/openstack-tripleo-heat-templates/environments/network-isolation.yaml \
  -e ~/templates/network-environment.yaml \
  -e /usr/share/openstack-tripleo-heat-templates/environments/services/ironic.yaml \
  -e ~/templates/ironic.yaml \

For more information about deploying the overcloud, see Creating the Overcloud with the CLI Tools and Including Environment Files in Overcloud Creation.

3.5. Testing the Bare Metal Service

You can use the OpenStack Integration Test Suite to validate your Red Hat OpenStack deployment. For more information, see the OpenStack Integration Test Suite Guide.

Additional Ways to Verify the Bare Metal Service:

  1. Set up the shell to access Identity as the administrative user:

    $ source ~/overcloudrc
  2. Check that the nova-compute service is running on the controller nodes:

    $ openstack compute service list -c Binary -c Host -c Status
  3. If you have changed the default ironic drivers, make sure the required drivers are enabled:

    $ openstack baremetal driver list
  4. Ensure that the ironic endpoints are listed:

    $ openstack catalog list

Chapter 4. Configuring for the Bare Metal Service After Deployment

This section describes the steps necessary to configure your overcloud after deployment.

4.1. Configuring OpenStack Networking

Configure OpenStack Networking to communicate with the Bare Metal service for DHCP, PXE boot, and other requirements. The procedure below configures OpenStack Networking for a single, flat network use case for provisioning onto bare metal. The configuration uses the ML2 plug-in and the Open vSwitch agent. Only flat networks are supported.

This procedure creates a bridge using the bare metal network interface, and drops any remote connections.

All steps in the following procedure must be performed on the server hosting OpenStack Networking, while logged in as the root user.

Configuring OpenStack Networking to Communicate with the Bare Metal Service

  1. Set up the shell to access Identity as the administrative user:

    $ source ~/overcloudrc
  2. Create the flat network over which to provision bare metal instances:

    $ openstack network create \
      --provider-network-type flat \
      --provider-physical-network baremetal \
      --share NETWORK_NAME

    Replace NETWORK_NAME with a name for this network. The name of the physical network over which the virtual network is implemented, (in this case baremetal), was set earlier in ~/templates/network-environment.yaml, with the parameter NeutronBridgeMappings.

  3. Create the subnet on the flat network:

    $ openstack subnet create \
      --network NETWORK_NAME \
      --subnet-range NETWORK_CIDR \
      --ip-version 4 \
      --gateway GATEWAY_IP \
      --allocation-pool start=START_IP,end=END_IP \
      --dhcp SUBNET_NAME

    Replace the following values:

    • Replace SUBNET_NAME with a name for the subnet.
    • Replace NETWORK_NAME with the name of the provisioning network you created in the previous step.
    • Replace NETWORK_CIDR with the Classless Inter-Domain Routing (CIDR) representation of the block of IP addresses the subnet represents. The block of IP addresses specified by the range started by START_IP and ended by END_IP must fall within the block of IP addresses specified by NETWORK_CIDR.
    • Replace GATEWAY_IP with the IP address or host name of the router interface that will act as the gateway for the new subnet. This address must be within the block of IP addresses specified by NETWORK_CIDR, but outside of the block of IP addresses specified by the range started by START_IP and ended by END_IP.
    • Replace START_IP with the IP address that denotes the start of the range of IP addresses within the new subnet from which floating IP addresses will be allocated.
    • Replace END_IP with the IP address that denotes the end of the range of IP addresses within the new subnet from which floating IP addresses will be allocated.
  4. Attach the network and subnet to the router to ensure the metadata requests are served by the OpenStack Networking service.

    $ openstack router create ROUTER_NAME

    Replace ROUTER_NAME with a name for the router.

  5. Add the Bare Metal subnet to this router:

    $ openstack router add subnet ROUTER_NAME BAREMETAL_SUBNET

    Replace ROUTER_NAME with the name of your router and BAREMETAL_SUBNET with the ID or subnet name that you previously created. This allows the metadata requests from cloud-init to be served and the node configured.

4.2. Configuring Node Cleaning

By default, the Bare Metal service is set to use a network named provisioning for node cleaning. However, network names are not unique in OpenStack Networking, so it is possible for a tenant to create a network with the same name, causing a conflict with the Bare Metal service. Therefore, it is recommended to use the network UUID instead.

  1. Configure cleaning by providing the provider network UUID on the controller running the Bare Metal Service:

    ~/templates/ironic.yaml

    parameter_defaults:
        IronicCleaningNetwork: UUID

    Replace UUID with the UUID of the bare metal network created in the previous steps.

    You can find the UUID using openstack network show:

    openstack network show NETWORK_NAME -f value -c id
    Note

    This configuration must be done after the initial overcloud deployment, because the UUID for the network isn’t available beforehand.

  2. Apply the changes by redeploying the overcloud with the openstack overcloud deploy command as described in Section 3.4, “Deploying the Overcloud”.

    Note

    It is possible to avoid redeploying the overcloud by updating the ironic.conf file on each controller. However, manually updating the ironic.conf file of an OpenStack director installation is not supported. These instructions are only provided for convenience.

    1. Uncomment the following line and replace <None> with the UUID of the bare metal network:

      cleaning_network = <None>
    2. Restart the Bare Metal service:

      # systemctl restart openstack-ironic-conductor.service

    Redeploying the overcloud with openstack overcloud deploy will revert any manual changes, so make sure you have added the cleaning configuration to ~/templates/ironic.yaml (described in the previous step) before the next time you use openstack overcloud deploy.

4.2.1. Manual Node Cleaning

To manually initiate node cleaning, the node must be in the manageable state.

Node cleaning has two modes:

Metadata only clean - Removes partitions from all disks on a given node. This is a faster clean cycle, but less secure since it only erases partition tables. Only use this mode on trusted tenant environments.

Full clean - Removes all data from all disks, using either ATA secure erase or by shredding. This can take several hours to complete.

To initiate a metadata clean:

$ openstack baremetal node clean UUID \
    --clean-steps [{"interface": "deploy", "step": "erase_devices_metadata"}]

To initiate a full clean:

$ openstack baremetal node clean UUID \
    --clean-steps [{"interface": "deploy", "step": "erase_devices"}]

Replace UUID with the UUID of the node you would like cleaned.

After a successful cleaning, the node state returns to manageable. If the state is clean failed, check the last_error field for the cause of failure.

4.3. Creating the Bare Metal Flavor

You need to create a flavor to use as a part of the deployment. The specifications (memory, CPU, and disk) of this flavor must be equal to or less than what your bare metal node provides.

  1. Set up the shell to access Identity as the administrative user:

    $ source ~/overcloudrc
  2. List existing flavors:

    $ openstack flavor list
  3. Create a new flavor for the Bare Metal service:

    $ openstack flavor create \
      --id auto --ram RAM \
      --vcpus VCPU --disk DISK \
      --property baremetal=true \
      --public baremetal

    Replace RAM with the amount of memory, VCPU with the number of vCPUs and DISK with the disk storage value. The property baremetal is used to distinguish bare metal from virtual instances.

  4. Verify that the new flavor is created with the respective values:

    $ openstack flavor list

4.4. Creating the Bare Metal Images

The deployment requires two sets of images:

  • The deploy image is used by the Bare Metal service to boot the bare metal node and copy a user image onto the bare metal node. The deploy image consists of the kernel image and the ramdisk image.
  • The user image is the image deployed onto the bare metal node. The user image also has a kernel image and ramdisk image, but additionally, the user image contains a main image. The main image is either a root partition, or a whole-disk image.

    • A whole-disk image is an image that contains the partition table and boot loader. The Bare Metal service does not control the subsequent reboot of a node deployed with a whole-disk image as the node supports localboot.
    • A root partition image only contains the root partition of the operating system. If using a root partition, after the deploy image is loaded into the Image service, you can set the deploy image as the node’s boot image in the node’s properties. A subsequent reboot of the node uses netboot to pull down the user image.

The examples in this section use a root partition image to provision bare metal nodes.

4.4.1. Preparing the Deploy Images

You do not have to create the deploy image as it was already used when the overcloud was deployed by the undercloud. The deploy image consists of two images - the kernel image and the ramdisk image as follows:

ironic-python-agent.kernel
ironic-python-agent.initramfs

These images are often in the home directory, unless you have deleted them, or unpacked them elsewhere. If they are not in the home directory, and you still have the rhosp-director-images-ipa package installed, these images will be in the /usr/share/rhosp-director-images/ironic-python-agent*.tar file.

Extract the images and upload them to the Image service:

$ openstack image create \
  --container-format aki \
  --disk-format aki \
  --public \
  --file ./ironic-python-agent.kernel bm-deploy-kernel
$ openstack image create \
  --container-format ari \
  --disk-format ari \
  --public \
  --file ./ironic-python-agent.initramfs bm-deploy-ramdisk

4.4.2. Preparing the User Image

The final image that you need is the user image that will be deployed on the bare metal node. User images also have a kernel and ramdisk, along with a main image.

  1. Download the Red Hat Enterprise Linux KVM guest image from the Customer Portal (requires login).
  2. Define DIB_LOCAL_IMAGE as the downloaded image:

    $ export DIB_LOCAL_IMAGE=rhel-server-7.4-x86_64-kvm.qcow2
  3. Create the user images using the diskimage-builder tool:

    $ disk-image-create rhel7 baremetal -o rhel-image

    This extracts the kernel as rhel-image.vmlinuz and initial ramdisk as rhel-image.initrd.

  4. Upload the images to the Image service:

    $ KERNEL_ID=$(openstack image create \
      --file rhel-image.vmlinuz --public \
      --container-format aki --disk-format aki \
      -f value -c id rhel-image.vmlinuz)
    $ RAMDISK_ID=$(openstack image create \
      --file rhel-image.initrd --public \
      --container-format ari --disk-format ari \
      -f value -c id rhel-image.initrd)
    $ openstack image create \
      --file rhel-image.qcow2   --public \
      --container-format bare \
      --disk-format qcow2 \
      --property kernel_id=$KERNEL_ID \
      --property ramdisk_id=$RAMDISK_ID \
      rhel-image

4.5. Adding Physical Machines as Bare Metal Nodes

There are two methods to enroll a bare metal node:

  1. Prepare an inventory file with the node details, import the file into the Bare Metal service, then make the nodes available.
  2. Register a physical machine as a bare metal node, then manually add its hardware details and create ports for each of its Ethernet MAC addresses. These steps can be performed on any node which has your overcloudrc file.

Both methods are detailed in this section.

After enrolling the physical machines, Compute is not immediately notified of new resources, because Compute’s resource tracker synchronizes periodically. Changes will be visible after the next periodic task is run. This value, scheduler_driver_task_period, can be updated in /etc/nova/nova.conf. The default period is 60 seconds.

4.5.1. Enrolling a Bare Metal Node With an Inventory File

  1. Create a file overcloud-nodes.yaml, including the node details. Multiple nodes can be enrolled with one file.

    nodes:
        - name: node0
          driver: pxe_ipmitool
          driver_info:
            ipmi_address: <IPMI_IP>
            ipmi_username: <USER>
            ipmi_password: <PASSWORD>
          properties:
            cpus: <CPU_COUNT>
            cpu_arch: <CPU_ARCHITECTURE>
            memory_mb: <MEMORY>
            local_gb: <ROOT_DISK>
            root_device:
                serial: <SERIAL>
          ports:
            - address: <PXE_NIC_MAC>

    Replace the following values:

    • <IPMI_IP> with the address of the Bare Metal controller.
    • <USER> with your username.
    • <PASSWORD> with your password.
    • <CPU_COUNT> with the number of CPUs.
    • <CPU_ARCHITECTURE> with the type of architecture of the CPUs.
    • <MEMORY> with the amount of memory in MiB.
    • <ROOT_DISK> with the size of the root disk in GiB.
    • <MAC_ADDRESS> with the MAC address of the NIC used to PXE boot.

      You only need to include root_device if the machine has multiple disks. Replace <SERIAL> with the serial number of the disk you would like used for deployment.

  2. Set up the shell to use Identity as the administrative user:

    $ source ~/overcloudrc
  3. Import the inventory file into ironic:

    $ openstack baremetal create overcloud-nodes.yaml
  4. The nodes are now in the enroll state. Make them available by specifying the deploy kernel and deploy ramdisk on each node:

    $ openstack baremetal node set NODE_UUID \
      --driver-info deploy_kernel=KERNEL_UUID \
      --driver-info deploy_ramdisk=INITRAMFS_UUID

    Replace the following values:

    • Replace NODE_UUID with the unique identifier for the node. Alternatively, use the node’s logical name.
    • Replace KERNEL_UUID with the unique identifier for the kernel deploy image that was uploaded to the Image service. Find this value with:

      $ openstack image show bm-deploy-kernel -f value -c id
    • Replace INITRAMFS_UUID with the unique identifier for the ramdisk image that was uploaded to the Image service. Find this value with:

      $ openstack image show bm-deploy-ramdisk -f value -c id
  5. Check that the nodes were successfully enrolled:

    $ openstack baremetal node list

    There may be a delay between enrolling a node and its state being shown.

4.5.2. Enrolling a Bare Metal Node Manually

  1. Set up the shell to use Identity as the administrative user:

    $ source ~/overcloudrc
  2. Add a new node:

    $ openstack baremetal node create --driver pxe_impitool --name NAME

    To create a node you must specify the driver name. This example uses pxe_impitool. To use a different driver, you must enable it by setting the IronicEnabledDrivers parameter. For more information on supported drivers, see Appendix A, Bare Metal Drivers.

    Important

    Note the unique identifier for the node.

  3. Update the node driver information to allow the Bare Metal service to manage the node:

    $ openstack baremetal node set NODE_UUID \
      --driver_info PROPERTY=VALUE \
      --driver_info PROPERTY=VALUE

    Replace the following values:

    • Replace NODE_UUID with the unique identifier for the node. Alternatively, use the node’s logical name.
    • Replace PROPERTY with a required property returned by the ironic driver-properties command.
    • Replace VALUE with a valid value for that property.
  4. Specify the deploy kernel and deploy ramdisk for the node driver:

    $ openstack baremetal node set NODE_UUID \
      --driver-info deploy_kernel=KERNEL_UUID \
      --driver-info deploy_ramdisk=INITRAMFS_UUID

    Replace the following values:

    • Replace NODE_UUID with the unique identifier for the node. Alternatively, use the node’s logical name.
    • Replace KERNEL_UUID with the unique identifier for the .kernel image that was uploaded to the Image service.
    • Replace INITRAMFS_UUID with the unique identifier for the .initramfs image that was uploaded to the Image service.
  5. Update the node’s properties to match the hardware specifications on the node:

    $ openstack baremetal node set NODE_UUID \
      --property cpus=CPU \
      --property memory_mb=RAM_MB \
      --property local_gb=DISK_GB \
      --property cpu_arch=ARCH

    Replace the following values:

    • Replace NODE_UUID with the unique identifier for the node. Alternatively, use the node’s logical name.
    • Replace CPU with the number of CPUs.
    • Replace RAM_MB with the RAM (in MB).
    • Replace DISK_GB with the disk size (in GB).
    • Replace ARCH with the architecture type.
  6. OPTIONAL: Configure the node to reboot after initial deployment from a local boot loader installed on the node’s disk, instead of using PXE from ironic-conductor. The local boot capability must also be set on the flavor used to provision the node. To enable local boot, the image used to deploy the node must contain grub2. Configure local boot:

    $ openstack baremetal node set NODE_UUID \
      --property capabilities="boot_option:local"

    Replace NODE_UUID with the unique identifier for the node. Alternatively, use the node’s logical name.

  7. Inform the Bare Metal service of the node’s network card by creating a port with the MAC address of the NIC on the provisioning network:

    $ openstack baremetal port create --node NODE_UUID MAC_ADDRESS

    Replace NODE_UUID with the unique identifier for the node. Replace MAC_ADDRESS with the MAC address of the NIC used to PXE boot.

  8. If you have multiple disks, set the root device hints. This informs the deploy ramdisk which disk it should use for deployment.

    $ openstack baremetal node set NODE_UUID \
      --property root_device={"PROPERTY": "VALUE"}

    Replace with the following values:

    • Replace NODE_UUID with the unique identifier for the node. Alternatively, use the node’s logical name.
    • Replace PROPERTY and VALUE with details about the disk you want used for deployment, for example root_device='{"size": 128}'

      The following properties are supported:

      • model (String): Device identifier.
      • vendor (String): Device vendor.
      • serial (String): Disk serial number.
      • hctl (String): Host:Channel:Target:Lun for SCSI.
      • size (Integer): Size of the device in GB.
      • wwn (String): Unique storage identifier.
      • wwn_with_extension (String): Unique storage identifier with the vendor extension appended.
      • wwn_vendor_extension (String): Unique vendor storage identifier.
      • rotational (Boolean): True for a rotational device (HDD), otherwise false (SSD).
      • name (String): The name of the device, for example: /dev/sdb1 Only use this for devices with persistent names.

        Note

        If you specify more than one property, the device must match all of those properties.

  9. Validate the node’s setup:

    $ openstack baremetal node validate NODE_UUID
    +------------+--------+---------------------------------------------+
    | Interface  | Result | Reason                                      |
    +------------+--------+---------------------------------------------+
    | boot       | False  | Cannot validate image information for node  |
    |            |        | a02178db-1550-4244-a2b7-d7035c743a9b        |
    |            |        | because one or more parameters are missing  |
    |            |        | from its instance_info. Missing are:        |
    |            |        | ['ramdisk', 'kernel', 'image_source']       |
    | console    | None   | not supported                               |
    | deploy     | False  | Cannot validate image information for node  |
    |            |        | a02178db-1550-4244-a2b7-d7035c743a9b        |
    |            |        | because one or more parameters are missing  |
    |            |        | from its instance_info. Missing are:        |
    |            |        | ['ramdisk', 'kernel', 'image_source']       |
    | inspect    | None   | not supported                               |
    | management | True   |                                             |
    | network    | True   |                                             |
    | power      | True   |                                             |
    | raid       | True   |                                             |
    | storage    | True   |                                             |
    +------------+--------+---------------------------------------------+

    Replace NODE_UUID with the unique identifier for the node. Alternatively, use the node’s logical name. The output of the command above should report either True or None for each interface. Interfaces marked None are those that you have not configured, or those that are not supported for your driver.

    Note

    Interfaces may fail validation due to missing 'ramdisk', 'kernel', and 'image_source' parameters. This result is fine, because the Compute service populates those missing parameters at the beginning of the deployment process.

4.6. Using Host Aggregates to Separate Physical and Virtual Machine Provisioning

OpenStack Compute uses host aggregates to partition availability zones, and group together nodes with specific shared properties. When an instance is provisioned, Compute’s scheduler compares properties on the flavor with the properties assigned to host aggregates, and ensures that the instance is provisioned in the correct aggregate and on the correct host: either on a physical machine or as a virtual machine.

The procedure below describes how to do the following:

  • Add the property baremetal to your flavors, setting it to either true or false.
  • Create separate host aggregates for bare metal hosts and compute nodes with a matching baremetal property. Nodes grouped into an aggregate inherit this property.

Creating a Host Aggregate

  1. Set the baremetal property to true on the baremetal flavor.

    $ openstack flavor set baremetal --property baremetal=true
  2. Set the baremetal property to false on the flavors used for virtual instances.

    $ openstack flavor set FLAVOR_NAME --property baremetal=false
  3. Create a host aggregate called baremetal-hosts:

    $ openstack aggregate create --property baremetal=true baremetal-hosts
  4. Add each controller node to the baremetal-hosts aggregate:

    $ openstack aggregate add host baremetal-hosts HOSTNAME
    Note

    If you have created a composable role with the NovaIronic service, add all the nodes with this service to the baremetal-hosts aggregate. By default, only the controller nodes have the NovaIronic service.

  5. Create a host aggregate called virtual-hosts:

    $ openstack aggregate create --property baremetal=false virtual-hosts
  6. Add each compute node to the virtual-hosts aggregate:

    $ openstack aggregate add host virtual-hosts HOSTNAME
  7. If you did not add the following Compute filter scheduler when deploying the overcloud, add it now to the existing list under scheduler_default_filters in /etc/nova/nova.conf:

    AggregateInstanceExtraSpecsFilter

Chapter 5. Administering Bare Metal Nodes

This chapter describes how to provision a physical machine on an enrolled bare metal node. Instances can be launched either from the command line or from the OpenStack dashboard.

5.1. Launching an Instance Using the Command Line Interface

Use the openstack command line interface to deploy a bare metal instance.

Deploying an Instance on the Command Line

  1. Set up the shell to access Identity as the administrative user:

    $ source ~/overcloudrc
  2. Deploy the instance:

    $ openstack server create \
      --nic net-id=NETWORK_UUID \
      --flavor baremetal \
      --image IMAGE_UUID \
      INSTANCE_NAME

    Replace the following values:

    • Replace NETWORK_UUID with the unique identifier for the network that was created for use with the Bare Metal service.
    • Replace IMAGE_UUID with the unique identifier for the disk image that was uploaded to the Image service.
    • Replace INSTANCE_NAME with a name for the bare metal instance.
  3. Check the status of the instance:

    $ openstack server list --name INSTANCE_NAME

5.2. Launch an Instance Using the Dashboard

Use the dashboard graphical user interface to deploy a bare metal instance.

Deploying an Instance in the Dashboard

  1. Log in to the dashboard at http[s]://DASHBOARD_IP/dashboard.
  2. Click Project > Compute > Instances
  3. Click Launch Instance.

    • In the Details tab, specify the Instance Name and select 1 for Count.
    • In the Source tab, select an Image from Select Boot Source, then click the + (plus) symbol to select an operating system disk image. The chosen image will move to Allocated.
    • In the Flavor tab, select baremetal.
    • In the Networks tab, use the + (plus) and - (minus) buttons to move required networks from Available to Allocated. Ensure that the shared network created for the Bare Metal service is selected here.
  4. Click Launch Instance.

Chapter 6. Troubleshooting the Bare Metal Service

The following sections contain information and steps that may be useful for diagnosing issues in a setup with the Bare Metal service enabled.

6.1. PXE Boot Errors

Permission Denied Errors

If you are getting a permission denied error on the console of your Bare Metal service node, make sure you have applied the appropriate SELinux context to the /httpboot and /tftpboot directories as follows:

# semanage fcontext -a -t httpd_sys_content_t "/httpboot(/.*)?"
# restorecon -r -v /httpboot
# semanage fcontext -a -t tftpdir_t "/tftpboot(/.*)?"
# restorecon -r -v /tftpboot

Boot Process Freezes at /pxelinux.cfg/XX-XX-XX-XX-XX-XX

On the console of your node, if it looks like you are getting an IP address and then the process stops as shown below:

PXE Process Freezes During Boot

This indicates that you might be using the wrong PXE boot template in your ironic.conf file.

$ grep ^pxe_config_template ironic.conf
pxe_config_template=$pybasedir/drivers/modules/ipxe_config.template

The default template is pxe_config.template, so it is easy to miss the i to turn this into ipxe_config.template.

6.2. Login Errors After the Bare Metal Node Boots

When you try to log in at the login prompt on the console of the node with the root password that you set in the configurations steps, but are not able to, it indicates you are not booted in to the deployed image. You are probably stuck in the deploy-kernel/deploy-ramdisk image and the system has yet to get the correct image.

To fix this issue, verify the PXE Boot Configuration file in the /httpboot/pxelinux.cfg/MAC_ADDRESS on the Compute or Bare Metal service node and ensure that all the IP addresses listed in this file correspond to IP addresses on the Bare Metal network.

Note

The only network the Bare Metal service node knows about is the Bare Metal network. If one of the endpoints is not on the network, the endpoint will not be able to reach the Bare Metal service node as a part of the boot process.

For example, the kernel line in your file is as follows:

kernel http://192.168.200.2:8088/5a6cdbe3-2c90-4a90-b3c6-85b449b30512/deploy_kernel selinux=0 disk=cciss/c0d0,sda,hda,vda iscsi_target_iqn=iqn.2008-10.org.openstack:5a6cdbe3-2c90-4a90-b3c6-85b449b30512 deployment_id=5a6cdbe3-2c90-4a90-b3c6-85b449b30512 deployment_key=VWDYDVVEFCQJNOSTO9R67HKUXUGP77CK ironic_api_url=http://192.168.200.2:6385 troubleshoot=0 text nofb nomodeset vga=normal boot_option=netboot ip=${ip}:${next-server}:${gateway}:${netmask} BOOTIF=${mac}  ipa-api-url=http://192.168.200.2:6385 ipa-driver-name=pxe_ipmitool boot_mode=bios initrd=deploy_ramdisk coreos.configdrive=0 || goto deploy
Value in the above example kernel lineCorresponding information

http://192.168.200.2:8088

Parameter http_url in /etc/ironic/ironic.conf file. This IP address must be on the Bare Metal network.

5a6cdbe3-2c90-4a90-b3c6-85b449b30512

UUID of the baremetal node in ironic node-list.

deploy_kernel

This is the deploy kernel image in the Image service that is copied down as /httpboot/<NODE_UUID>/deploy_kernel.

http://192.168.200.2:6385

Parameter api_url in /etc/ironic/ironic.conf file. This IP address must be on the Bare Metal network.

pxe_impitool

The IPMI Driver in use by the Bare Metal service for this node.

deploy_ramdisk

This is the deploy ramdisk image in the Image service that is copied down as /httpboot/<NODE_UUID>/deploy_ramdisk.

If a value does not correspond between the /httpboot/pxelinux.cfg/MAC_ADDRESS and the ironic.conf file:

  1. Update the value in the ironic.conf file
  2. Restart the Bare Metal service
  3. Re-deploy the Bare Metal instance

6.3. The Bare Metal Service Is Not Getting the Right Hostname

If the Bare Metal service is not getting the right hostname, it means that cloud-init is failing. To fix this, connect the Bare Metal subnet to a router in the OpenStack Networking service. The requests to the meta-data agent should now be routed correctly.

6.4. Invalid OpenStack Identity Service Credentials When Executing Bare Metal Service Commands

If you are having trouble authenticating to the Identity service, check the identity_uri parameter in the ironic.conf file and make sure you remove the /v2.0 from the keystone AdminURL. For example, identity_uri should be set to http://IP:PORT.

6.5. Hardware Enrollment

Issues with enrolled hardware can be caused by incorrect node registration details. Ensure that property names and values have been entered correctly. Incorrect or mistyped property names will be successfully added to the node’s details, but will be ignored.

Update a node’s details. This example updates the amount of memory the node is registered to use to 2 GB:

$ openstack baremetal node set --property memory_mb=2048 NODE_UUID

6.6. No Valid Host Errors

If the Compute scheduler cannot find a suitable Bare Metal node on which to boot an instance, a NoValidHost error can be seen in /var/log/nova/nova-conductor.log or immediately upon launch failure in the dashboard. This is usually caused by a mismatch between the resources Compute expects and the resources the Bare Metal node provides.

  1. Check the hypervisor resources that are available:

    $ openstack hypervisor stats show

    The resources reported here should match the resources that the Bare Metal nodes provide.

  2. Check that Compute recognizes the Bare Metal nodes as hypervisors:

    $ openstack hypervisor list

    The nodes, identified by UUID, should appear in the list.

  3. Check the details for a Bare Metal node:

    $ openstack baremetal node list
    $ openstack baremetal node show NODE_UUID

    Verify that the node’s details match those reported by Compute.

  4. Check that the selected flavor does not exceed the available resources of the Bare Metal nodes:

    $ openstack flavor show FLAVOR_NAME
  5. Check the output of openstack baremetal node list to ensure that Bare Metal nodes are not in maintenance mode. Remove maintenance mode if necessary:

    $ openstack baremetal node maintenance unset NODE_UUID
  6. Check the output of openstack baremetal node list to ensure that Bare Metal nodes are in an available state. Move the node to available if necessary:

    $ openstack baremetal node provide NODE_UUID

Appendix A. Bare Metal Drivers

A bare metal node can be configured to use one of the drivers enabled in the Bare Metal service. Each driver is made up of a provisioning method and a power management type. Some drivers require additional configuration. Each driver described in this section uses PXE for provisioning; drivers are listed by their power management type.

You can add drivers with the IronicEnabledDrivers parameter in your ironic.yaml file. By default, pxe_ipmitool, pxe_drac and pxe_ilo are enabled.

For the full list of supported plug-ins and drivers, see Component, Plug-In, and Driver Support in Red Hat OpenStack Platform.

A.1. Intelligent Platform Management Interface (IPMI)

IPMI is an interface that provides out-of-band remote management features, including power management and server monitoring. To use this power management type, all Bare Metal service nodes require an IPMI that is connected to the shared Bare Metal network. Enable the pxe_ipmitool driver, and set the following information in the node’s driver_info:

  • ipmi_address - The IP address of the IPMI NIC.
  • ipmi_username - The IPMI user name.
  • ipmi_password - The IPMI password.

A.2. Dell Remote Access Controller (DRAC)

DRAC is an interface that provides out-of-band remote management features, including power management and server monitoring. To use this power management type, all Bare Metal service nodes require a DRAC that is connected to the shared Bare Metal network. Enable the pxe_drac driver, and set the following information in the node’s driver_info:

  • drac_address - The IP address of the DRAC NIC.
  • drac_username - The DRAC user name.
  • drac_password - The DRAC password.

A.3. Integrated Remote Management Controller (iRMC)

iRMC from Fujitsu is an interface that provides out-of-band remote management features including power management and server monitoring. To use this power management type on a Bare Metal service node, the node requires an iRMC interface that is connected to the shared Bare Metal network. Enable the pxe_irmc driver, and set the following information in the node’s driver_info:

  • irmc_address - The IP address of the iRMC interface NIC.
  • irmc_username - The iRMC user name.
  • irmc_password - The iRMC password.

To use IPMI to set the boot mode or SCCI to get sensor data, you must complete the following additional steps:

  1. Enable the sensor method in ironic.conf:

    $ openstack-config --set /etc/ironic/ironic.conf \
       irmc sensor_method METHOD

    Replace METHOD with scci or ipmitool.

  2. If you enabled SCCI, install the python-scciclient package:

    # yum install python-scciclient
  3. Restart the Bare Metal conductor service:

    # systemctl restart openstack-ironic-conductor.service
Note

To use the iRMC driver, iRMC S4 or higher is required.

A.4. Integrated Lights-Out (iLO)

iLO from Hewlett-Packard is an interface that provides out-of-band remote management features including power management and server monitoring. To use this power management type, all Bare Metal nodes require an iLO interface that is connected to the shared Bare Metal network. Enable the pxe_ilo driver, and set the following information in the node’s driver_info:

  • ilo_address - The IP address of the iLO interface NIC.
  • ilo_username - The iLO user name.
  • ilo_password - The iLO password.

You must also install the python-proliantutils package and restart the Bare Metal conductor service:

# yum install python-proliantutils
# systemctl restart openstack-ironic-conductor.service

A.5. SSH and Virsh

Note

In Red Hat OpenStack Platform 11, the pxe_ssh driver has been deprecated in favor of VirtualBMC, which uses the pxe_ipmitool driver.

The Bare Metal service can access a host that is running libvirt and use virtual machines as nodes. Virsh controls the power management of the nodes.

Important

The SSH driver is for testing and evaluation purposes only. It is not recommended for Red Hat OpenStack Platform enterprise environments.

To use this power management type, the Bare Metal service must have SSH access to an account with full access to the libvirt environment on the host where the virtual nodes will be set up. Enable the pxe_ssh driver, and set the following information in the node’s driver_info:

  • ssh_virt_type - Set this option to virsh.
  • ssh_address - The IP address of the virsh host.
  • ssh_username - The SSH user name.
  • ssh_key_contents - The contents of the SSH private key on the Bare Metal conductor node. The matching public key must be copied to the virsh host.

A.6. VirtualBMC

VirtualBMC is a utility which can create a virtual baseboard management controller (BMC). A virtual BMC allows you to control virtual machines with the IPMI protocol, just like a physical machine.

Important

VirtualBMC is for testing and evaluation purposes only. It is not recommended for Red Hat OpenStack Platform enterprise environments.

  1. On the hypervisor hosting the virtual machines, install VirtualBMC:

    # yum install python-virtualbmc
  2. Create a BMC for each virtual machine:

    $ vbmc add DOMAIN --port PORT_NUMBER --username USERNAME --password PASSWORD
  3. Start the virtual BMCs:

    $ vbmc start DOMAIN

    The virtual BMCs do not start automatically, so if you reboot the host machine, remember to start the virtual BMCs again.

A.6.1. Migrating from pxe_ssh to VirtualBMC

If you have an existing virtual overcloud using the pxe_ssh driver, it is possible to migrate to pxe_impitool using Virtual BMC.

  1. Install python-virtualbmc and create BMCs for each virtual machine by following the steps above.
  2. Make sure pxe_impitool is enabled (it is enabled by default). If it is not enabled, include it in your ironic.yaml file and run the openstack overcloud deploy command again to apply the changes.

    parameter_defaults:
      IronicEnabledDrivers:
          - pxe_ipmitool
          - pxe_drac
          - pxe_ilo
  3. Update the driver and driver properties on each node:

    $ openstack baremetal node set NODE \
        --driver pxe_ipmitool \
        --driver-info ipmi_address=IP_ADDRESS \
        --driver-info ipmi_port=PORT \
        --driver-info ipmi_username="USERNAME" \
        --driver-info ipmi_password="PASSWORD"
    • Replace NODE with the name or UUID of the node.
    • Replace IP_ADDRESS with the IP address of the virtual host.
    • Replace PORT with the Virtual BMC port.
  4. Validate that the node has updated correctly:

    $ openstack baremetal node validate NODE | grep power
    • Replace NODE with the name or UUID of the node.

Legal Notice

Copyright © 2018 Red Hat, Inc.
The text of and illustrations in this document are licensed by Red Hat under a Creative Commons Attribution–Share Alike 3.0 Unported license ("CC-BY-SA"). An explanation of CC-BY-SA is available at http://creativecommons.org/licenses/by-sa/3.0/. In accordance with CC-BY-SA, if you distribute this document or an adaptation of it, you must provide the URL for the original version.
Red Hat, as the licensor of this document, waives the right to enforce, and agrees not to assert, Section 4d of CC-BY-SA to the fullest extent permitted by applicable law.
Red Hat, Red Hat Enterprise Linux, the Shadowman logo, JBoss, OpenShift, Fedora, the Infinity logo, and RHCE are trademarks of Red Hat, Inc., registered in the United States and other countries.
Linux® is the registered trademark of Linus Torvalds in the United States and other countries.
Java® is a registered trademark of Oracle and/or its affiliates.
XFS® is a trademark of Silicon Graphics International Corp. or its subsidiaries in the United States and/or other countries.
MySQL® is a registered trademark of MySQL AB in the United States, the European Union and other countries.
Node.js® is an official trademark of Joyent. Red Hat Software Collections is not formally related to or endorsed by the official Joyent Node.js open source or commercial project.
The OpenStack® Word Mark and OpenStack logo are either registered trademarks/service marks or trademarks/service marks of the OpenStack Foundation, in the United States and other countries and are used with the OpenStack Foundation's permission. We are not affiliated with, endorsed or sponsored by the OpenStack Foundation, or the OpenStack community.
All other trademarks are the property of their respective owners.