Node cleaning¶

Enabling automated cleaning¶

To enable automated cleaning, ensure that your ironic.conf is set as follows:

[conductor]
automated_clean=true

This will enable the default set of cleaning steps, based on your hardware and ironic hardware types used for nodes. This includes, by default, erasing all of the previous tenant’s data.

You may also need to configure a Cleaning Network.

Cleaning steps¶

Cleaning steps used for automated cleaning are ordered from higher to lower priority, where a larger integer is a higher priority. In case of a conflict between priorities across interfaces, the following resolution order is used: Power, Management, Deploy, and RAID interfaces.

You can skip a cleaning step by setting the priority for that cleaning step to zero or ‘None’.

You can reorder the cleaning steps by modifying the integer priorities of the cleaning steps.

See How do I change the priority of a cleaning step? for more information.

Setup¶

In order for manual cleaning to work, you may need to configure a Cleaning Network.

Starting manual cleaning via API¶

Manual cleaning can only be performed when a node is in the manageable state. The REST API request to initiate it is available in API version 1.15 and higher:

PUT /v1/nodes/<node_ident>/states/provision

(Additional information is available here.)

This API will allow operators to put a node directly into cleaning provision state from manageable state via ‘target’: ‘clean’. The PUT will also require the argument ‘clean_steps’ to be specified. This is an ordered list of cleaning steps. A cleaning step is represented by a dictionary (JSON), in the form:

{
    "interface": "<interface>",
    "step": "<name of cleaning step>",
    "args": {"<arg1>": "<value1>", ..., "<argn>": <valuen>}
}

The ‘interface’ and ‘step’ keys are required for all steps. If a cleaning step method takes keyword arguments, the ‘args’ key may be specified. It is a dictionary of keyword variable arguments, with each keyword-argument entry being <name>: <value>.

If any step is missing a required keyword argument, manual cleaning will not be performed and the node will be put in clean failed provision state with an appropriate error message.

If, during the cleaning process, a cleaning step determines that it has incorrect keyword arguments, all earlier steps will be performed and then the node will be put in clean failed provision state with an appropriate error message.

An example of the request body for this API:

{
  "target":"clean",
  "clean_steps": [{
    "interface": "raid",
    "step": "create_configuration",
    "args": {"create_nonroot_volumes": false}
  },
  {
    "interface": "deploy",
    "step": "erase_devices"
  }]
}

In the above example, the node’s RAID interface would configure hardware RAID without non-root volumes, and then all devices would be erased (in that order).

Starting manual cleaning via “openstack baremetal” CLI¶

Manual cleaning is available via the openstack baremetal node clean command, starting with Bare Metal API version 1.15.

The argument --clean-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 clean steps. Using ‘-‘ to signify stdin is common in Unix utilities.

The following examples assume that the Bare Metal API version was set via the OS_BAREMETAL_API_VERSION environment variable. (The alternative is to add --os-baremetal-api-version 1.15 to the command.):

export OS_BAREMETAL_API_VERSION=1.15

Examples of doing this with a JSON string:

openstack baremetal node clean <node> \
    --clean-steps '[{"interface": "deploy", "step": "erase_devices_metadata"}]'

openstack baremetal node clean <node> \
    --clean-steps '[{"interface": "deploy", "step": "erase_devices"}]'

Or with a file:

openstack baremetal node clean <node> \
    --clean-steps my-clean-steps.txt

Or with stdin:

cat my-clean-steps.txt | openstack baremetal node clean <node> \
    --clean-steps -

In-band¶

In-band steps are performed by ironic making API calls to a ramdisk running on the node using a deploy interface. Currently, all the deploy interfaces support in-band cleaning. By default, ironic-python-agent ships with a minimal cleaning configuration, only erasing disks. However, you can add your own cleaning steps and/or override default cleaning steps with a custom Hardware Manager.

Out-of-band¶

Out-of-band are actions performed by your management controller, such as IPMI, iLO, or DRAC. Out-of-band steps will be performed by ironic using a power or management interface. Which steps are performed depends on the hardware type and hardware itself.

For Out-of-Band cleaning operations supported by iLO hardware types, refer to Node Cleaning Support.

How are cleaning steps ordered?¶

For automated cleaning, cleaning steps are ordered by integer priority, where a larger integer is a higher priority. In case of a conflict between priorities across hardware interfaces, the following resolution order is used:

1. Power interface
2. Management interface
3. Deploy interface
4. RAID interface

For manual cleaning, the cleaning steps should be specified in the desired order.

How do I skip a cleaning step?¶

For automated cleaning, cleaning steps with a priority of 0 or None are skipped.

How do I change the priority of a cleaning step?¶

For manual cleaning, specify the cleaning steps in the desired order.

For automated cleaning, it depends on whether the cleaning steps are out-of-band or in-band.

Most out-of-band cleaning steps have an explicit configuration option for priority.

Changing the priority of an in-band (ironic-python-agent) cleaning step requires use of a custom HardwareManager. The only exception is erase_devices, which can have its priority set in ironic.conf. For instance, to disable erase_devices, you’d set the following configuration option:

[deploy]
erase_devices_priority=0

To enable/disable the in-band disk erase using ilo hardware type, use the following configuration option:

[ilo]
clean_priority_erase_devices=0

The generic hardware manager first tries to perform ATA disk erase by using hdparm utility. If ATA disk erase is not supported, it performs software based disk erase using shred utility. By default, the number of iterations performed by shred for software based disk erase is 1. To configure the number of iterations, use the following configuration option:

[deploy]
erase_devices_iterations=1

What cleaning step is running?¶

To check what cleaning 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 clean_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 clean failed state.

Should I disable automated cleaning?¶

Automated cleaning is recommended for ironic deployments, however, there are some tradeoffs to having it enabled. For instance, ironic cannot deploy a new instance to a node that is currently cleaning, and cleaning can be a time consuming process. To mitigate this, we suggest using disks with support for cryptographic ATA Security Erase, as typically the erase_devices step in the deploy interface takes the longest time to complete of all cleaning steps.

Why can’t I power on/off a node while it’s cleaning?¶

During cleaning, nodes may be performing actions that shouldn’t be interrupted, such as BIOS or Firmware updates. As a result, operators are forbidden from changing power state via the ironic API while a node is cleaning.