ironic/doc/source/contributor/states.rst
Kaifeng Wang 6df82ee2bc Implementation of inspect wait state
This patch provides implementations to feature of adding inspect wait state.

Changes covered in this patch:

* Added state and transitions, state diagram regenerated.
* inspector and oneview inspect interface now return INSPECTWAIT instead of
  INSPECTING. Move node to inspect wait if inspect interface returns
  INSPECTING or INSPECTWAIT.
* Add a timeout option to conductor, and a periodic task to check timeout
  in the inspect wait state.

Story: #1725211
Task: #10630

Partial-Bug: #1725211
Change-Id: Ie76bfdad5966014a4dae826919ff5705462c743b
2018-04-10 11:21:46 +08:00

10 KiB

Ironic's State Machine

State Machine Diagram

The diagram below shows the provisioning states that an Ironic node goes through during the lifetime of a node. The diagram also depicts the events that transition the node to different states.

Stable states are highlighted with a thicker border. All transitions from stable states are initiated by API requests. There are a few other API-initiated-transitions that are possible from non-stable states. The events for these API-initiated transitions are indicated with '(via API)'. Internally, the conductor initiates the other transitions (depicted in gray).

State Descriptions

enroll (stable state)

This is the state that all nodes start off in when created using API version 1.11 or newer. When a node is in the enroll state, the only thing ironic knows about it is that it exists, and ironic cannot take any further action by itself. Once a node has its driver/interfaces and their required information set in node.driver_info, the node can be transitioned to the verifying state by setting the node's provision state using the manage verb.

verifying

ironic will validate that it can manage the node using the information given in node.driver_info and with either the driver/hardware type and interfaces it has been assigned. This involves going out and confirming that the credentials work to access whatever node control mechanism they talk to.

manageable (stable state)

Once ironic has verified that it can manage the node using the driver/interfaces and credentials passed in at node create time, the node will be transitioned to the manageable state. From manageable, nodes can transition to:

  • manageable (through cleaning) by setting the node's provision state using the clean verb.
  • manageable (through inspecting) by setting the node's provision state using the inspect verb.
  • available (through cleaning if automatic cleaning is enabled) by setting the node's provision state using the provide verb.
  • active (through adopting) by setting the node's provision state using the adopt verb.

manageable is the state that a node should be moved into when any updates need to be made to it such as changes to fields in driver_info and updates to networking information on ironic ports assigned to the node.

manageable is also the only stable state that can be transitioned to, from these failure states:

  • adopt failed
  • clean failed
  • inspect failed
inspecting

inspecting will utilize node introspection to update hardware-derived node properties to reflect the current state of the hardware. Typically, the node will transition to manageable if inspection is synchronous, or inspect wait if asynchronous. The node will transition to inspect failed if error occurred.

inspect wait

This is the provision state used when an asynchronous inspection is in progress. A successfully inspected node shall transition to manageable state.

inspect failed

This is the state a node will move into when inspection of the node fails. From here the node can transitioned to:

  • inspecting by setting the node's provision state using the inspect verb.
  • manageable by setting the node's provision state using the manage verb
cleaning

Nodes in the cleaning state are being scrubbed and reprogrammed into a known configuration.

When a node is in the cleaning state it means that the conductor is executing the clean step (for out-of-band clean steps) or preparing the environment (building PXE configuration files, configuring the DHCP, etc) to boot the ramdisk for running in-band clean steps.

clean wait

Just like the cleaning state, the nodes in the clean wait state are being scrubbed and reprogrammed. The difference is that in the clean wait state the conductor is waiting for the ramdisk to boot or the clean step which is running in-band to finish.

The cleaning process of a node in the clean wait state can be interrupted by setting the node's provision state using the abort verb if the task that is running allows it.

available (stable state)

After nodes have been successfully preconfigured and cleaned, they are moved into the available state and are ready to be provisioned. From available, nodes can transition to:

  • active (through deploying) by setting the node's provision state using the active verb.
  • manageable by setting the node's provision state using the manage verb
deploying

Nodes in deploying are being prepared to run a workload on them. This consists of running a series of tasks, such as:

  • Setting appropriate BIOS configurations
  • Partitioning drives and laying down file systems.
  • Creating any additional resources (node-specific network config, a config drive partition, etc.) that may be required by additional subsystems.
wait call-back

Just like the deploying state, the nodes in wait call-back are being deployed. The difference is that in wait call-back the conductor is waiting for the ramdisk to boot or execute parts of the deployment which need to run in-band on the node (for example, installing the bootloader, or writing the image to the disk).

The deployment of a node in wait call-back can be interrupted by setting the node's provision state using the deleted verb.

deploy failed

This is the state a node will move into when a deployment fails, for example a timeout waiting for the ramdisk to PXE boot. From here the node can be transitioned to:

  • active (through deploying) by setting the node's provision state using either the active or rebuild verbs.
  • available (through deleting and cleaning) by setting the node's provision state using the deleted verb.
active (stable state)

Nodes in active have a workload running on them. ironic may collect out-of-band sensor information (including power state) on a regular basis. Nodes in active can transition to:

  • available (through deleting and cleaning) by setting the node's provision state using the deleted verb.
  • active (through deploying) by setting the node's provision state using the rebuild verb.
  • rescue (through rescuing) by setting the node's provision state using the rescue verb.
deleting

Nodes in deleting state are being torn down from running an active workload. In deleting, ironic tears down and removes any configuration and resources it added in deploying or rescuing.

error (stable state)

This is the state a node will move into when deleting an active deployment fails. From error, nodes can transition to:

  • available (through deleting and cleaning) by setting the node's provision state using the deleted verb.
adopting

This state allows ironic to take over management of a baremetal node with an existing workload on it. Ordinarily when a baremetal node is enrolled and managed by ironic, it must transition through cleaning and deploying to reach active state. However, those baremetal nodes that have an existing workload on them, do not need to be deployed or cleaned again, so this transition allows these nodes to move directly from manageable to active.

rescuing

Nodes in rescuing are being prepared to perform rescue operations. This consists of running a series of tasks, such as:

  • Setting appropriate BIOS configurations.
  • Creating any additional resources (node-specific network config, etc.) that may be required by additional subsystems.
rescue wait

Just like the rescuing state, the nodes in rescue wait are being rescued. The difference is that in rescue wait the conductor is waiting for the ramdisk to boot or execute parts of the rescue which need to run in-band on the node (for example, setting the password for user named rescue).

The rescue operation of a node in rescue wait can be aborted by setting the node's provision state using the abort verb.

rescue failed

This is the state a node will move into when a rescue operation fails, for example a timeout waiting for the ramdisk to PXE boot. From here the node can be transitioned to:

  • rescue (through rescuing) by setting the node's provision state using the rescue verb.
  • active (through unrescuing) by setting the node's provision state using the unrescue verb.
  • available (through deleting) by setting the node's provision state using the deleted verb.
rescue (stable state)

Nodes in rescue have a rescue ramdisk running on them. Ironic may collect out-of-band sensor information (including power state) on a regular basis. Nodes in rescue can transition to:

  • active (through unrescuing) by setting the node's provision state using the unrescue verb.
  • available (through deleting) by setting the node's provision state using the deleted verb.
unrescuing

Nodes in unrescuing are being prepared to transition to active state from rescue state. This consists of running a series of tasks, such as setting appropriate BIOS configurations such as changing boot device.

unrescue failed

This is the state a node will move into when an unrescue operation fails. From here the node can be transitioned to:

  • rescue (through rescuing) by setting the node's provision state using the rescue verb.
  • active (through unrescuing) by setting the node's provision state using the unrescue verb.
  • available (through deleting) by setting the node's provision state using the deleted verb.