This is a summary of work items discussed at the PTG.
Assisted-by: Claude code
Change-Id: I5fe43622c6f03c3caacc1cc6b169f7354b32523e
Signed-off-by: Jay Faulkner <jay@jvf.cc>
Follow-up from Oct 2025 PTG: template for workitems that includes a
specific checklist we can reference each meeting.
Signed-off-by: Jay Faulkner <jay@jvf.cc>
Change-Id: Icdcdcdf6d337e08f1eda0fd8d805da60d06a8be3
This updates the standalone networking spec to addresses issues related
to the initial starting conditions. Specifically, there is a 'chicken-
and-egg' scenario that exists in that LLDP information is required
in order to know which switch ports need to be configured for a node
but if those switch ports are disabled by default there is no way to
know which switch ports to configure.
This updates the document to state that switch ports must be manually
configured ahead of time and must be setup on the inspection VLAN to
facilitate collecting LLDP information.
Related-Bug: 2113769
Change-Id: I9a5854ce81b0a1478af34604b4fe093ecdac6e01
Signed-off-by: Allain Legacy <alegacy@redhat.com>
Describes a potential design that adds the ability to configure TOR
switch equipment, without involvement of Neutron, when baremetal nodes
are managed. This is a continuing of the ideas put forward by project
Mercury.
Related-Bug: 2113769
Change-Id: I8888bea37e5e4ca4bfa1d5d7bafb7c12a0b71ae3
Signed-off-by: Allain Legacy <alegacy@redhat.com>
Ironic's network mapping is extremely basic and many aspects of it are
static -- such as the physical_network a port maps to, or the creation
of portgroups being done as a part of the node enrollment process.
This means we need to have a way to hint Ironic, at deploy time, how to
map ports. Luckily, we already have this mechanism via custom traits.
The change proposes a new configuration grammar which would allow
operators to specify logic around how to manage port/portgroup/network
mappings keyed to traits configured during deployment.
Generated-By: claude-code was used to help draft the expression grammar
Change-Id: Ie040f8972802991cace044f830d600951aa0fe6c
A number of specs were still referenced as not-implemented but actually
have been.
Change-Id: I2d07e03d77e0e1a16a4f8aa5f893e8cd8f37ca20
Signed-off-by: Doug Goldstein <cardoe@cardoe.com>
Notes about work items discussed at the PTG and expected to be worked on
during the Flamingo cycle.
Change-Id: Ic0c8cbb0b8deeeab8fc4a188a6d8a5346f7a9d0a
Hardware vendors are adding support for remote graphical consoles via
non-standardised Redfish interfaces. End users would like to have a
consistent way of accessing these consoles via Ironic using
command-line, the Nova driver, or the Horizon web interface.
Related-Bug: 2086715
Change-Id: Ieb69edd7f9e559438fc340fb4ef934d6ab8365e0
Adds a specifcation document related to the use of bootable
containers utilizing bootc and as a result the ostree strucutral
model.
Related-Bug: 2085801
Change-Id: I815f2403d0ae75f2c03a99207629c845cc5aeaaa
This specification document, which someone should take over,
aims to support the retrieval of disk image file artifacts
from a container registry, to allow operators in
container native environments to leverage a simplified model
instead of having to deploy their own webserver which has been
viewed as burdomsome.
Related-Bug: 2085565
Change-Id: Ia7fbc514ed69d1f3e445a710ed1d93ce8e5b07fa
Cleaned up the priorities and specs for any sphinx-lint issues.
Change-Id: I5adc43161078e5fb6301d7222cfa92bbf9723922
Signed-off-by: Doug Goldstein <cardoe@cardoe.com>
Consensus from discussion in IRC represents a forward direction
on the graphical console modeling.
A few distinct aspects which drives this consensus:
1) The existing specification defines a fairly heavy weight/design
intensive modeling which was based on an attempt to create a
specific delineation, between "console" and "graphical console",
where console was explicitly "serial console".
2) Except, drivers, including out of tree patches could always just
respond to the console creation request with a VNC url. Julia
Kreger has indicated she was aware of a HPE patch from her time
at HPE where Helion OpenStack did exactly this to integrate with
Nova's framework.
3) Serial consoles are sort of on their way out of vouge. What we
mean by this the dual challenge of evolution of architecture,
combined with a lack of evolution due to low level designs.
UEFI specifically is modeled around graphical interactions,
which largely started from the hardware vendors wanting logos
and "graphical" boot splash screens. At the same time, serial
consoles continue to leveral low level interrupt which we've
since observed issues with serial console interaction resulting
in detrimental performance impact due to the interrupts being
triggered with general console output. Comparing interaction
with a graphics card, the OS kernel generally just writes to
an IO address without triggering interrupts.
4) The base interaction with nova and users *really* doesn't *need*
to change, and the existing serial console functionality, while
needing a *ton* of work does work for some operators, be it a
port to connect to, or a webpage to load from the console
activation interaction.
Where this leaves Ironic, is we likely need to figure out
*exactly* what we need to do on the lower level for *redfish*
and unlock ourselves to do so. This action sort of allows us to
do that, focus on the lower level challenge instead of focusing
on creating an entirely new design that is not really needed *aside*
from resolving the ambiguity we've created for ourselves. Partly
which this change does help work towards resolving as a first step.
Also see: https://meetings.opendev.org/irclogs/%23openstack-ironic/%23openstack-ironic.2024-11-01.log.html
Change-Id: Idb86ac9f7e27b4fe4b73f6163da81b743517adef
Change-Id: I6527458dbd28d8de6b86a406392b219a1a7c31d1
Added running codespell via pre-commit. Corrected all the spelling
mistakes as well. It seemed better to correct them than to leave them
incorrect in past specs and priorities since it looks more professional.
Change-Id: If96abb56726a4d8164ebdd5de4dc5ba09bc734a5
Signed-off-by: Doug Goldstein <cardoe@cardoe.com>
Switch to using pre-commit for linters. This consolidates the doc8 run
into pre-commit. To ensure that CI matches what users get and to not
require users to do anything different, call pre-commit from tox. Added
the same pre-commit hooks from the ironic repo where they pass cleanly.
Change-Id: Ic49b06d8ef1621646bf5455b705af1e0db830ab1
Signed-off-by: Doug Goldstein <cardoe@cardoe.com>
This appears to be the first time we've done this since just before the
xena release.
This includes a fix for firmware-interface being a file in
not-implemented/ instead of in approved/ and symlinked in.
Change-Id: Iad890a729b219946d510fd28ae3685e133d8b77b
Spec ilo-uefi-iscsi-boot.rst which was proposed and
approved causes Gerrit Zuul CI linter failure in
newly proposed spec.
Change-Id: If0044f325184185acc54fd5d5c73ddab98b88004
Fixing a few typos and make some changes which I noted in the
specification as it was being reviewed.
Nothing major, just clean-up.
Change-Id: Ic8b3b371488d97396f413168dd1720d4ec47c73c
