Node Deployment

Node Deployment

Overview

Node deployment is performed by the Bare Metal service to prepare a node for use by a workload. The exact work flow used depends on a number of factors, including the hardware type and interfaces assigned to a node.

Deploy Steps

The Bare Metal service implements deployment by collecting a list of deploy steps to perform on a node from the Power, Deploy, Management, BIOS, and RAID interfaces of the driver assigned to the node. These steps are then ordered by priority and executed on the node when the node is moved to the deploying state.

Nodes move to the deploying state when attempting to move to the active state (when the hardware is prepared for use by a workload). For a full understanding of all state transitions into deployment, please see Ironic’s State Machine.

The Bare Metal service added support for deploy steps in the Rocky release.

Order of execution

Deploy steps are ordered from higher to lower priority, where a larger integer is a higher priority. If the same priority is used by deploy steps on different interfaces, the following resolution order is used: Power, Management, Deploy, BIOS, and RAID interfaces.

Core steps

Certain default deploy steps are designated as ‘core’ deploy steps. The following deploy steps are core:

deploy.deploy

In this step the node is booted using a provisioning image, and the user image is written to the node’s disk. It has a priority of 100.

Writing a Deploy Step

Please refer to Developing a new Deploy Step.

FAQ

What deploy step is running?

To check what deploy step the node is performing or attempted to perform and failed, run the following command; it will return the value in the node’s driver_internal_info field:

openstack baremetal node show $node_ident -f value -c driver_internal_info

The deploy_steps field will contain a list of all remaining steps with their priorities, and the first one listed is the step currently in progress or that the node failed before going into deploy failed state.

Troubleshooting

If deployment fails on a node, the node will be put into the deploy failed state until the node is deprovisioned. A deprovisioned node is moved to the available state after the cleaning process has been performed successfully.

Strategies for determining why a deploy step failed include checking the ironic conductor logs, checking logs from the ironic-python-agent that have been stored on the ironic conductor, or performing general hardware troubleshooting on the node.

Deploy Templates

Starting with the Stein release, with Bare Metal API version 1.55, deploy templates offer a way to define a set of one or more deploy steps to be executed with particular sets of arguments and priorities.

Each deploy template has a name, which must be a valid trait. Traits can be either standard or custom. Standard traits are listed in the os_traits library. Custom traits must meet the following requirements:

  • prefixed with CUSTOM_

  • contain only upper case characters A to Z, digits 0 to 9, or underscores

  • no longer than 255 characters in length

Deploy step format

An invocation of a deploy step is defined in a deploy template as follows:

{
    "interface": "<name of the driver interface>",
    "step": "<name of the step>",
    "args": {
        "<arg1>": "<value1>",
        "<arg2>": "<value2>"
    },
    "priority": <priority of the step>
}

A deploy template contains a list of one or more such steps. Each combination of interface and step may only be specified once in a deploy template.

Matching deploy templates

During deployment, if any of the traits in a node’s instance_info.traits field match the name of a deploy template, then the steps from that deploy template will be added to the list of steps to be executed by the node.

When using the Compute service, any traits in the instance’s flavor properties or image properties are stored in instance_info.traits during deployment. See Scheduling based on traits for further information on how traits are used for scheduling when the Bare Metal service is used with the Compute service.

Note that there is no ongoing relationship between a node and any templates that are matched during deployment. The set of matching deploy templates is checked at deployment time. Any subsequent updates to or deletion of those templates will not be reflected in the node’s configuration unless it is redeployed or rebuilt. Similarly, if a node is rebuilt and the set of matching deploy templates has changed since the initial deployment, then the resulting configuration of the node may be different from the initial deployment.

Overriding default deploy steps

A deploy step is enabled by default if it has a non-zero default priority. A default deploy step may be overridden in a deploy template. If the step’s priority is a positive integer it will be executed with the specified priority and arguments. If the step’s priority is zero, the step will not be executed.

If a core deploy step is included in a deploy template, it can only be assigned a priority of zero to disable it.

Creating a deploy template via API

A deploy template can be created using the Bare Metal API:

POST /v1/deploy_templates

Here is an example of the body of a request to create a deploy template with a single step:

{
    "name": "CUSTOM_HYPERTHREADING_ON",
    "steps": [
        {
            "interface": "bios",
            "step": "apply_configuration",
            "args": {
                "settings": [
                    {
                        "name": "LogicalProc",
                        "value": "Enabled"
                    }
                ]
            },
            "priority": 150
        }
    ]
}

Further information on this API is available here.

Creating a deploy template via “openstack baremetal” client

A deploy template can be created via the openstack baremetal deploy template create command, starting with python-ironicclient 2.7.0.

The argument --steps must be specified. Its value is one of:

  • a JSON string

  • path to a JSON file whose contents are passed to the API

  • ‘-‘, to read from stdin. This allows piping in the deploy steps.

Example of creating a deploy template with a single step using a JSON string:

openstack baremetal deploy template create \
    CUSTOM_HYPERTHREADING_ON \
    --steps '[{"interface": "bios", "step": "apply_configuration", "args": {"settings": [{"name": "LogicalProc", "value": "Enabled"}]}, "priority": 150}]'

Or with a file:

openstack baremetal deploy template create \
    CUSTOM_HYPERTHREADING_ON \
    ---steps my-deploy-steps.txt

Or with stdin:

cat my-deploy-steps.txt | openstack baremetal deploy template create \
    CUSTOM_HYPERTHREADING_ON \
    --steps -

Example of use with the Compute service

Note

The deploy steps used in this example are for example purposes only.

In the following example, we have a node with the following node traits:

[
    "CUSTOM_HYPERTHREADING_ON"
]

We also have a flavor, bm-hyperthreading-on, in the Compute service with the following property:

trait:CUSTOM_HYPERTHREADING_ON:required

Creating a Compute instance with this flavor will ensure that the instance is scheduled only to Bare Metal nodes with the CUSTOM_HYPERTHREADING_ON trait.

We could then create a Bare Metal deploy template with the name CUSTOM_HYPERTHREADING_ON and a deploy step that enables Hyperthreading:

{
    "name": "CUSTOM_HYPERTHREADING_ON",
    "steps": [
        {
            "interface": "bios",
            "step": "apply_configuration",
            "args": {
                "settings": [
                    {
                        "name": "LogicalProc",
                        "value": "Enabled"
                    }
                ]
            },
            "priority": 150
        }
    ]
}

When an instance is created using the bm-hyperthreading-on flavor, then the deploy steps of deploy template CUSTOM_HYPERTHREADING_ON will be executed during the deployment of the scheduled node, causing Hyperthreading to be enabled in the node’s BIOS configuration.

To make this example more dynamic, consider adding a second trait to the node:

[
    "CUSTOM_HYPERTHREADING_ON",
    "CUSTOM_HYPERTHREADING_OFF"
]

We could also create a second flavor, bm-hyperthreading-off, with the following property:

trait:CUSTOM_HYPERTHREADING_OFF:required

Finally, we create a deploy template with the name CUSTOM_HYPERTHREADING_OFF and a different set of deploy steps:

{
    "name": "CUSTOM_HYPERTHREADING_OFF",
    "steps": [
        {
            "interface": "bios",
            "step": "apply_configuration",
            "args": {
                "settings": [
                    {
                        "name": "LogicalProc",
                        "value": "Disabled"
                    }
                ]
            },
            "priority": 150
        }
    ]
}

Creating a Compute instance with the bm-hyperthreading-off instance will cause the scheduled node to have Hyperthreading disabled in the BIOS during deployment.

We now have a way to create Compute instances with different configurations, by choosing between different Compute flavors, supported by a single Bare Metal node that is dynamically configured during deployment.

Creative Commons Attribution 3.0 License

Except where otherwise noted, this document is licensed under Creative Commons Attribution 3.0 License. See all OpenStack Legal Documents.