Ironic - Bare Metal provisioning

Overview

Ironic is the OpenStack service for handling bare metal, i.e., the physical machines. It can work standalone as well as with other OpenStack services (notably, Neutron and Nova).

Pre-deployment Configuration

Enable Ironic in /etc/kolla/globals.yml:

enable_ironic: "yes"

In the same file, define a network interface as the default NIC for dnsmasq and define a network to be used for the Ironic cleaning network:

ironic_dnsmasq_interface: "eth1"
ironic_cleaning_network: "public1"

Finally, define at least one DHCP range for Ironic inspector:

ironic_dnsmasq_dhcp_ranges:
  - range: "192.168.5.100,192.168.5.110"

Another example of a single range with a router (multiple routers are possible by separating addresses with commas):

ironic_dnsmasq_dhcp_ranges:
  - range: "192.168.5.100,192.168.5.110"
    routers: "192.168.5.1"

Together with an router there can be provided the NTP (time source) server. For example it can be the same address as default router for the range:

ironic_dnsmasq_dhcp_ranges:
  - range: "192.168.5.100,192.168.5.110"
    routers: "192.168.5.1"
    ntp_server: "192.168.5.1"

To support DHCP relay, it is also possible to define a netmask in the range. It is advisable to also provide a router to allow the traffic to reach the Ironic server.

ironic_dnsmasq_dhcp_ranges:
  - range: "192.168.5.100,192.168.5.110,255.255.255.0"
    routers: "192.168.5.1"

Multiple ranges are possible, they can be either for directly-connected interfaces or relays (if with netmask):

ironic_dnsmasq_dhcp_ranges:
  - range: "192.168.5.100,192.168.5.110"
  - range: "192.168.6.100,192.168.6.110,255.255.255.0"
    routers: "192.168.6.1"

The default lease time for each range can be configured globally via ironic_dnsmasq_dhcp_default_lease_time variable or per range via lease_time parameter.

In the same file, specify the PXE bootloader file for Ironic Inspector. The file is relative to the /var/lib/ironic/tftpboot directory. The default is pxelinux.0, and should be correct for x86 systems. Other platforms may require a different value, for example aarch64 on Debian requires debian-installer/arm64/bootnetaa64.efi.

ironic_dnsmasq_boot_file: pxelinux.0

Ironic inspector also requires a deploy kernel and ramdisk to be placed in /etc/kolla/config/ironic/. The following example uses coreos which is commonly used in Ironic deployments, though any compatible kernel/ramdisk may be used:

$ curl https://tarballs.opendev.org/openstack/ironic-python-agent/dib/files/ipa-centos9-master.kernel \
  -o /etc/kolla/config/ironic/ironic-agent.kernel

$ curl https://tarballs.opendev.org/openstack/ironic-python-agent/dib/files/ipa-centos9-master.initramfs \
  -o /etc/kolla/config/ironic/ironic-agent.initramfs

You may optionally pass extra kernel parameters to the inspection kernel using:

ironic_inspector_kernel_cmdline_extras: ['ipa-lldp-timeout=90.0', 'ipa-collect-lldp=1']

in /etc/kolla/globals.yml.

Configure conductor’s HTTP server port (optional)

The port used for conductor’s HTTP server is controlled via ironic_http_port in /etc/kolla/globals.yml:

ironic_http_port: "8089"

Attach ironic to external keystone (optional)

In multi-regional deployment keystone could be installed in one region (let’s say region 1) and ironic - in another region (let’s say region 2). In this case we don’t install keystone together with ironic in region 2, but have to configure ironic to connect to existing keystone in region 1. To deploy ironic in this way we have to set variable enable_keystone to "no".

enable_keystone: "no"

It will prevent keystone from being installed in region 2.

To add keystone-related sections in ironic.conf, it is also needed to set variable ironic_enable_keystone_integration to "yes"

ironic_enable_keystone_integration: "yes"

Avoiding problems with high availability

Note

This section assumes that you have not yet deployed the Nova Compute Ironic service. If you have already deployed multiple instances of the service and have one or more baremetal nodes registered, the following operations are non-trivial. You will likely have to use the nova-manage command (or pre-Caracal edit the DB) to ensure that all Ironic nodes are registered with a single Nova Compute Ironic instance. This is an advanced subject and is not covered here. Stop now if you don’t know what you are doing.

Nova Compute Ironic HA is known to be unstable. Pending a better solution, a workaround is to avoid the feature by running a single Nova Compute Ironic instance. For example:

- [nova-compute-ironic:children]
- nova
+ [nova-compute-ironic]
+ controller1

If you choose to do this, it is helpful to pin the service host name to a ‘synthetic’ constant. This means that if you need to re-deploy the service to another host, the Ironic nodes will automatically use the new service instance. Otherwise you will need to manually move active Ironic nodes to the new service, with either the nova-manage CLI, or pre-Caracal, by editing the Nova database.

The config option to pin the host name is nova_compute_ironic_custom_host and must be set as a group or host var. Note that, unless you know what you are doing, you must not change or set this option if you have already deployed Ironic nodes.

This config option is also useful for Ironic Shards. Whilst these are not explicitly supported by Kolla Ansible, some further information can be found here.

Note that Ironic HA is not affected, and continues to work as normal.

Deployment

Run the deploy as usual:

$ kolla-ansible deploy

Post-deployment configuration

The Ironic documentation describes how to create the deploy kernel and ramdisk and register them with Glance. In this example we’re reusing the same images that were fetched for the Inspector:

openstack image create --disk-format aki --container-format aki --public \
  --file /etc/kolla/config/ironic/ironic-agent.kernel deploy-vmlinuz

openstack image create --disk-format ari --container-format ari --public \
  --file /etc/kolla/config/ironic/ironic-agent.initramfs deploy-initrd

The Ironic documentation describes how to create Nova flavors for bare metal. For example:

openstack flavor create my-baremetal-flavor \
  --ram 512 --disk 1 --vcpus 1 \
  --property resources:CUSTOM_BAREMETAL_RESOURCE_CLASS=1 \
  --property resources:VCPU=0 \
  --property resources:MEMORY_MB=0 \
  --property resources:DISK_GB=0

The Ironic documentation describes how to enroll baremetal nodes and ports. In the following example ensure to substitute correct values for the kernel, ramdisk, and MAC address for your baremetal node.

openstack baremetal node create --driver ipmi --name baremetal-node \
  --driver-info ipmi_port=6230 --driver-info ipmi_username=admin \
  --driver-info ipmi_password=password \
  --driver-info ipmi_address=192.168.5.1 \
  --resource-class baremetal-resource-class --property cpus=1 \
  --property memory_mb=512 --property local_gb=1 \
  --property cpu_arch=x86_64 \
  --driver-info deploy_kernel=15f3c95f-d778-43ad-8e3e-9357be09ca3d \
  --driver-info deploy_ramdisk=9b1e1ced-d84d-440a-b681-39c216f24121

openstack baremetal port create 52:54:00:ff:15:55 \
  --node 57aa574a-5fea-4468-afcf-e2551d464412 \
  --physical-network physnet1

Make the baremetal node available to nova:

openstack baremetal node manage 57aa574a-5fea-4468-afcf-e2551d464412
openstack baremetal node provide 57aa574a-5fea-4468-afcf-e2551d464412

It may take some time for the node to become available for scheduling in nova. Use the following commands to wait for the resources to become available:

openstack hypervisor stats show
openstack hypervisor show 57aa574a-5fea-4468-afcf-e2551d464412

Booting the baremetal

Assuming you have followed the examples above and created the demo resources as shown in the Quick Start for deployment/evaluation, you can now use the following example command to boot the baremetal instance:

openstack server create --image cirros --flavor my-baremetal-flavor \
  --key-name mykey --network public1 demo1

In other cases you will need to adapt the command to match your environment.

Notes

Debugging DHCP

The following tcpdump command can be useful when debugging why dhcp requests may not be hitting various pieces of the process:

tcpdump -i <interface> port 67 or port 68 or port 69 -e -n

Configuring the Web Console

Configuration based off upstream Node web console.

Serial speed must be the same as the serial configuration in the BIOS settings. Default value: 115200bps, 8bit, non-parity.If you have different serial speed.

Set ironic_console_serial_speed in /etc/kolla/globals.yml:

ironic_console_serial_speed: 9600n8

Deploying using virtual baremetal (vbmc + libvirt)

See https://brk3.github.io/post/kolla-ironic-libvirt/