Rework nodes registration and configuration
* Convert some notes to warnings, move important text out of notes. * Move enroll state documentation into a newly written section (mostly copy-pasted from one of my recent emails to openstack-dev). * Stop mentioning API version 1.11, as we already use it by default. * Explain the 'baremetal configure boot' is no longer mandatory starting with Newton. * Write a section on automatically setting the root device hints. * Move power state sync note to troubleshooting, as people are unlikely to be interested in or even understand this part. * Fix small mistakes. Change-Id: I5f078bafd64d4fbcc5e83557f7761226cca93fd6
This commit is contained in:
Dmitry Tantsur
parent
02869b84d3
commit
6b060fc283
@ -6,6 +6,7 @@ In this chapter you will find advanced deployment of various |project| areas.
|
||||
|
||||
.. toctree::
|
||||
|
||||
node_states
|
||||
root_device
|
||||
build_single_image
|
||||
introspect_single_node
|
||||
|
||||
@ -5,7 +5,8 @@ In addition to bulk introspection, you can also introspect nodes one by one.
|
||||
When doing so, you must take care to set the correct node states manually.
|
||||
Use ``ironic node-show UUID`` command to figure out whether nodes are in
|
||||
``manageable`` or ``available`` state. For all nodes in ``available`` state,
|
||||
start with putting a node to ``manageable`` state::
|
||||
start with putting a node to ``manageable`` state (see
|
||||
:doc:`../advanced_deployment/node_states` for details)::
|
||||
|
||||
ironic node-set-provision-state UUID manage
|
||||
|
||||
|
||||
58
doc/source/advanced_deployment/node_states.rst
Normal file
58
doc/source/advanced_deployment/node_states.rst
Normal file
@ -0,0 +1,58 @@
|
||||
Bare Metal Node States
|
||||
======================
|
||||
|
||||
This document provides a brief explanation of the bare metal node states that
|
||||
TripleO uses or might use. Please refer to `the Ironic documentation
|
||||
<http://docs.openstack.org/developer/ironic/>`_ for more details.
|
||||
|
||||
enroll
|
||||
------
|
||||
|
||||
With recent versions of the bare metal API (starting with 1.11), nodes begin
|
||||
their life in a state called ``enroll``. Nodes in this state are not available
|
||||
for deployment, nor for most of other actions. Ironic does not touch such nodes
|
||||
in any way.
|
||||
|
||||
Starting with the Newton release, TripleO node registration command allows
|
||||
to import nodes in this state instead of the default ``available``.
|
||||
To do so pass the option ``--initial-state=enroll``::
|
||||
|
||||
openstack baremetal import --initial-state=enroll instackenv.json
|
||||
|
||||
Then move the nodes to manageable_ state and eventually to available_.
|
||||
|
||||
manageable
|
||||
----------
|
||||
|
||||
To make nodes alive an operator uses ``manage`` provisioning action to move
|
||||
nodes to ``manageable`` state. During this transition the power and management
|
||||
credentials (IPMI, SSH, etc) are validated to ensure that nodes in
|
||||
``manageable`` state are actually manageable by Ironic. This state is still not
|
||||
available for deployment. With nodes in this state an operator can execute
|
||||
various pre-deployment actions, such as introspection, RAID configuration, etc.
|
||||
So to sum it up, nodes in ``manageable`` state are being configured before
|
||||
exposing them into the cloud.
|
||||
|
||||
Nodes get into ``manageable`` state automatically. The ``manage`` action
|
||||
can be used to bring nodes already moved to available_ state back to
|
||||
``manageable`` for configuration::
|
||||
|
||||
ironic node-set-provision-state <NAME OR UUID> manage
|
||||
|
||||
|
||||
available
|
||||
---------
|
||||
|
||||
The last step before the deployment is to make nodes ``available`` using the
|
||||
``provide`` provisioning action. Such nodes are exposed to nova, and can be
|
||||
deployed to at any moment. No long-running configuration actions should be run
|
||||
in this state.
|
||||
|
||||
.. note::
|
||||
The TripleO introspection command ``openstack baremetal introspection bulk
|
||||
start`` moves ``available`` nodes to manageable_ state automatically
|
||||
before and moves them back after a successful introspection. However, nodes
|
||||
which failed introspection stay in ``manageable`` state and must be
|
||||
reintrospected or made ``available`` manually::
|
||||
|
||||
ironic node-set-provision-state <NAME OR UUID> provide
|
||||
@ -27,6 +27,32 @@ Where the new value is calculated as a real disk size in GiB minus 1 GiB to
|
||||
account for partitioning (the introspection process does this calculation
|
||||
automatically).
|
||||
|
||||
Setting root device hints automatically
|
||||
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
||||
|
||||
Starting with the Newton release it is possible to autogenerate root device
|
||||
hints for all nodes instead of setting them one by one. Pass the
|
||||
``--root-device`` argument to the ``openstack baremetal configure boot``
|
||||
**after a successful introspection**. This argument can accept a device list
|
||||
in the order of preference, for example::
|
||||
|
||||
openstack baremetal configure boot --root-device=sdb,sdc,vda
|
||||
|
||||
It can also accept one of two strategies: ``smallest`` will pick the smallest
|
||||
device, ``largest`` will pick the largest one. By default only disk devices
|
||||
larger than 4 GiB are considered at all, set the ``--root-device-minimum-size``
|
||||
argument to change.
|
||||
|
||||
.. note::
|
||||
Subsequent runs of this command on the same set of nodes does nothing,
|
||||
as root device hints are already recorded on nodes and are not overwritten.
|
||||
If you want to change existing root device hints, first remove them manually
|
||||
as described above.
|
||||
|
||||
.. note::
|
||||
This command relies on introspection data, so if you change disk devices on
|
||||
the machines, introspection must be rerun before rerunning this command.
|
||||
|
||||
Using introspection data to find the root device
|
||||
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
||||
|
||||
|
||||
@ -175,24 +175,42 @@ Load the images into the undercloud Glance::
|
||||
Register Nodes
|
||||
--------------
|
||||
|
||||
Register nodes for your deployment with Ironic::
|
||||
Register and configure nodes for your deployment with Ironic::
|
||||
|
||||
openstack baremetal import instackenv.json
|
||||
|
||||
.. note::
|
||||
The file to be imported may be either JSON, YAML or CSV format, and
|
||||
the type is detected via the file extension (json, yaml, csv).
|
||||
The file to be imported may be either JSON, YAML or CSV format, and
|
||||
the type is detected via the file extension (json, yaml, csv).
|
||||
The file format is documented in :ref:`instackenv`.
|
||||
|
||||
.. admonition:: Stable Branch
|
||||
:class: stable
|
||||
|
||||
If you're running an old (stable/liberty or older) tripleoclient version
|
||||
however then you must specify --json explicitly and only use that format.
|
||||
For TripleO release Liberty and older an explicit ``--json`` or ``--csv``
|
||||
flag is required for JSON and CSV formats accordingly.
|
||||
|
||||
.. note::
|
||||
The file format for node import is documented, see :ref:`instackenv`.
|
||||
.. admonition:: Stable Branch
|
||||
:class: stable
|
||||
|
||||
.. note::
|
||||
For TripleO release Mitaka and older the following command must be run
|
||||
after registration to assign the deployment kernel and ramdisk to all nodes:
|
||||
|
||||
openstack baremetal configure boot
|
||||
|
||||
Starting with the Newton release you can take advantage of the ``enroll``
|
||||
provisioning state - see :doc:`../advanced_deployment/node_states` for details.
|
||||
|
||||
If your hardware has several hard drives, it's highly recommended that you
|
||||
specify the exact device to be used during introspection and deployment
|
||||
as a root device. Please see :ref:`root_device` for details.
|
||||
|
||||
.. warning::
|
||||
If you don't specify the root device explicitly, any device may be picked.
|
||||
Also the device chosen automatically is **NOT** guaranteed to be the same
|
||||
across rebuilds. Make sure to wipe the previous installation before
|
||||
rebuilding in this case.
|
||||
|
||||
.. warning::
|
||||
It's not recommended to delete nodes and/or rerun this command after
|
||||
you have proceeded to the next steps. Particularly, if you start introspection
|
||||
and then re-register nodes, you won't be able to retry introspection until
|
||||
@ -200,41 +218,6 @@ Register nodes for your deployment with Ironic::
|
||||
with nodes after registration, please follow
|
||||
:ref:`node_registration_problems`.
|
||||
|
||||
.. note::
|
||||
By default Ironic will not sync the power state of the nodes,
|
||||
because in our HA (high availability) model Pacemaker is the
|
||||
one responsible for controlling the power state of the nodes
|
||||
when fencing. If you are using a non-HA setup and want Ironic
|
||||
to take care of the power state of the nodes please change the
|
||||
value of the "force_power_state_during_sync" configuration option
|
||||
in the /etc/ironic/ironic.conf file to "True" and restart the
|
||||
openstack-ironic-conductor service.
|
||||
|
||||
Also, note that if ``openstack undercloud install`` is re-run the value of
|
||||
the "force_power_state_during_sync" configuration option will be set back to
|
||||
the default, which is "False".
|
||||
|
||||
.. note::
|
||||
If you want to take advantage of the ENROLL provisioning state introduced in
|
||||
OpenStack Liberty, you must enable an Ironic API version >= 1.11 (for
|
||||
instance by setting the OS_BAREMETAL_API_VERSION environment variable), and
|
||||
pass the option ``--initial-state=enroll`` to ``openstack baremetal
|
||||
import``.
|
||||
|
||||
Assign kernel and ramdisk to nodes::
|
||||
|
||||
openstack baremetal configure boot
|
||||
|
||||
.. note::
|
||||
If your hardware has several hard drives, it's highly recommended that you
|
||||
specify the exact device to be used during introspection and deployment
|
||||
as a root device. Please see :ref:`root_device` for details.
|
||||
|
||||
If you don't specify the root device explicitly, any device may be picked.
|
||||
Also the device chosen automatically is NOT guaranteed to be the same
|
||||
across rebuilds. Make sure to wipe the previous installation before
|
||||
rebuilding in this case.
|
||||
|
||||
.. _introspection:
|
||||
|
||||
Introspect Nodes
|
||||
|
||||
@ -42,6 +42,22 @@ A Wrong IPMI address can be fixed with the following command::
|
||||
|
||||
ironic node-update <NODE UUID> replace driver_info/ipmi_address=<NEW IPMI ADDRESS>
|
||||
|
||||
Node power state is not enforced by Ironic
|
||||
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
||||
|
||||
By default Ironic will not forcibly sync the power state of the nodes,
|
||||
because in our HA (high availability) model Pacemaker is the
|
||||
one responsible for controlling the power state of the nodes
|
||||
when fencing. If you are using a non-HA setup and want Ironic
|
||||
to take care of the power state of the nodes please change the
|
||||
value of the ``force_power_state_during_sync`` configuration option
|
||||
in the ``/etc/ironic/ironic.conf`` file to ``True`` and restart the
|
||||
openstack-ironic-conductor service.
|
||||
|
||||
Also, note that if ``openstack undercloud install`` is re-run the value of
|
||||
the ``force_power_state_during_sync`` configuration option will be set back to
|
||||
the default, which is ``False``.
|
||||
|
||||
|
||||
.. _introspection_problems:
|
||||
|
||||
|
||||
Loading…
Reference in New Issue
Block a user