)]}'
{"specs/approved/bmc-settings-management.rst":[{"author":{"_account_id":10239,"name":"Dmitry Tantsur","email":"dtantsur@protonmail.com","username":"dtantsur"},"change_message_id":"a5c7b039f34e3d8d365aa5c4012cf184148e6858","unresolved":true,"context_lines":[{"line_number":18,"context_line":"registry. We don\u0027t expose them today, and we should, but not by pretending"},{"line_number":19,"context_line":"they are BIOS settings. They belong to a different part of the machine and,"},{"line_number":20,"context_line":"more importantly, to a different set of people. This spec proposes managing"},{"line_number":21,"context_line":"BMC settings as their own resource, sitting behind the ``management``"},{"line_number":22,"context_line":"interface, with an access policy that keeps them away from the lessee."},{"line_number":23,"context_line":""},{"line_number":24,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":2,"id":"74e6a879_0e1ae04e","line":21,"updated":"2026-09-11 10:44:48.000000000","message":"I wonder if it\u0027s a good point to add a new interface. Management is already overloaded with everything. It does have BMC reset on iDRAC, but it also has boot settings for a System.\n\nI have a somewhat related use case of managing credentials and networking: https://bugs.launchpad.net/ironic/+bug/2162051. These could go as steps on the new interface.\n\nBy the way, do you think networking parameters will be possible to manage through the API you propose?","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"},{"author":{"_account_id":11655,"name":"Julia Kreger","email":"juliaashleykreger@gmail.com","username":"jkreger","status":"Flying to the moon with a Jetpack!"},"change_message_id":"c92d11088d680bd6e8f08e832a7e0d044ebfd37d","unresolved":true,"context_lines":[{"line_number":18,"context_line":"registry. We don\u0027t expose them today, and we should, but not by pretending"},{"line_number":19,"context_line":"they are BIOS settings. They belong to a different part of the machine and,"},{"line_number":20,"context_line":"more importantly, to a different set of people. This spec proposes managing"},{"line_number":21,"context_line":"BMC settings as their own resource, sitting behind the ``management``"},{"line_number":22,"context_line":"interface, with an access policy that keeps them away from the lessee."},{"line_number":23,"context_line":""},{"line_number":24,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":2,"id":"5207a766_eaa14a28","line":21,"in_reply_to":"74e6a879_0e1ae04e","updated":"2026-09-25 19:08:50.000000000","message":"I view management as the management interface, so I\u0027m semi -1 to cutting yet another interface.\n\nRegarding networking settings, We\u0027ve seen both formats and some overlap of settings in to the BMC bios settings themselves which get squirreled away under Oem resources. :(","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"},{"author":{"_account_id":11655,"name":"Julia Kreger","email":"juliaashleykreger@gmail.com","username":"jkreger","status":"Flying to the moon with a Jetpack!"},"change_message_id":"c92d11088d680bd6e8f08e832a7e0d044ebfd37d","unresolved":true,"context_lines":[{"line_number":41,"context_line":"settings configure the management controller itself -- its networking, its"},{"line_number":42,"context_line":"IPMI-over-LAN and Redfish access, its users and authentication policy, its"},{"line_number":43,"context_line":"serial console redirection. Folding the two together would make it impossible"},{"line_number":44,"context_line":"to talk about, audit, or reason about either one on its own."},{"line_number":45,"context_line":""},{"line_number":46,"context_line":"Second, and this is the part that actually matters, they have a different"},{"line_number":47,"context_line":"audience. In a multi-tenant bare metal deployment we routinely let the lessee"}],"source_content_type":"text/x-rst","patch_set":2,"id":"e9c0a42b_1c30d8bc","line":44,"updated":"2026-09-25 19:08:50.000000000","message":"As an aside, this feels like we\u0027re on the direct elevator to a separate DB table and API resource.","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"},{"author":{"_account_id":10239,"name":"Dmitry Tantsur","email":"dtantsur@protonmail.com","username":"dtantsur"},"change_message_id":"a5c7b039f34e3d8d365aa5c4012cf184148e6858","unresolved":true,"context_lines":[{"line_number":51,"context_line":"weaken or lock us out of our own management path. So BMC settings need to be"},{"line_number":52,"context_line":"restricted to the system and to the node\u0027s *owner* administrator, and the"},{"line_number":53,"context_line":"lessee should not see them at all."},{"line_number":54,"context_line":""},{"line_number":55,"context_line":"Use cases:"},{"line_number":56,"context_line":""},{"line_number":57,"context_line":"* As an owner administrator, I want to audit the current BMC configuration of"}],"source_content_type":"text/x-rst","patch_set":2,"id":"05a96e4c_e7964451","line":54,"updated":"2026-09-11 10:44:48.000000000","message":"Third, a single BMC may manage several nodes. It\u0027s not explicit in your proposed API (we should discuss it too), but at least you highlight that BMC settings are not System settings.","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"},{"author":{"_account_id":11655,"name":"Julia Kreger","email":"juliaashleykreger@gmail.com","username":"jkreger","status":"Flying to the moon with a Jetpack!"},"change_message_id":"c92d11088d680bd6e8f08e832a7e0d044ebfd37d","unresolved":true,"context_lines":[{"line_number":51,"context_line":"weaken or lock us out of our own management path. So BMC settings need to be"},{"line_number":52,"context_line":"restricted to the system and to the node\u0027s *owner* administrator, and the"},{"line_number":53,"context_line":"lessee should not see them at all."},{"line_number":54,"context_line":""},{"line_number":55,"context_line":"Use cases:"},{"line_number":56,"context_line":""},{"line_number":57,"context_line":"* As an owner administrator, I want to audit the current BMC configuration of"}],"source_content_type":"text/x-rst","patch_set":2,"id":"216b6888_6c74b61f","line":54,"in_reply_to":"05a96e4c_e7964451","updated":"2026-09-25 19:08:50.000000000","message":"I concur, I\u0027m not sure we can really model that short of Chassis....\n\n*ducks*\n\nSeriously though, maybe the key is to build a separate table, link the data, and just refresh occassionally. That may mean in those cases that the bmc settings evolve, but we don\u0027t surface it to the end users in terms of manageability.\n\nActually, lessee scoepd readers should likely be blocked from even seeing the resource.","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"},{"author":{"_account_id":10239,"name":"Dmitry Tantsur","email":"dtantsur@protonmail.com","username":"dtantsur"},"change_message_id":"a5c7b039f34e3d8d365aa5c4012cf184148e6858","unresolved":true,"context_lines":[{"line_number":80,"context_line":"hardware interface. Managing the BMC is part of managing the node, and every"},{"line_number":81,"context_line":"driver that can talk to a BMC already has a ``management`` interface, so we"},{"line_number":82,"context_line":"extend that with optional methods rather than forcing every hardware type and"},{"line_number":83,"context_line":"out-of-tree driver to grow a new interface it would mostly no-op."},{"line_number":84,"context_line":""},{"line_number":85,"context_line":"Reading and writing are split the same way they are for BIOS. Reading is done"},{"line_number":86,"context_line":"over the REST API with a ``GET``; the values are served from the database,"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3afb9f04_15fb28b9","line":83,"updated":"2026-09-11 10:44:48.000000000","message":"Not sure (see above). We already put too much into Management. Also, the new interface will be mostly no-op for IPMI. I\u0027d expect many downstream drivers to inherit Redfish nowadays.\n\nNot a blocking concern, but something to think about.","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"},{"author":{"_account_id":11655,"name":"Julia Kreger","email":"juliaashleykreger@gmail.com","username":"jkreger","status":"Flying to the moon with a Jetpack!"},"change_message_id":"c92d11088d680bd6e8f08e832a7e0d044ebfd37d","unresolved":true,"context_lines":[{"line_number":89,"context_line":"(``apply_bmc_settings``, and in a follow-on ``factory_reset_bmc``) through the"},{"line_number":90,"context_line":"normal cleaning and deploy path, exactly as BIOS configuration is applied"},{"line_number":91,"context_line":"through clean steps today. The stricter read policy then makes sure the"},{"line_number":92,"context_line":"resulting values are only visible to the system and the owner administrator."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"Where the values come from is a driver detail, and Redfish does not put them"},{"line_number":95,"context_line":"in one standard place. The generic ``redfish`` ``management`` implementation"}],"source_content_type":"text/x-rst","patch_set":2,"id":"4c1752a2_7cff0396","line":92,"updated":"2026-09-25 19:08:50.000000000","message":"FWIW, as an aside, I kind of like this model.","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"},{"author":{"_account_id":10239,"name":"Dmitry Tantsur","email":"dtantsur@protonmail.com","username":"dtantsur"},"change_message_id":"a5c7b039f34e3d8d365aa5c4012cf184148e6858","unresolved":true,"context_lines":[{"line_number":95,"context_line":"in one standard place. The generic ``redfish`` ``management`` implementation"},{"line_number":96,"context_line":"reads the attributes carried on the Manager resource itself, via sushy\u0027s"},{"line_number":97,"context_line":"``Manager.bmc``. The ``idrac`` hardware type -- whose management interface"},{"line_number":98,"context_line":"already subclasses the ``redfish`` one -- overrides where it looks and reads"},{"line_number":99,"context_line":"the Dell OEM ``DellAttributes`` resource instead, since that is where iDRAC"},{"line_number":100,"context_line":"keeps them. Either way the sync into ``bmc_settings`` and the"},{"line_number":101,"context_line":"``{\"Attributes\": {...}}`` update used to apply them are the same code, the"}],"source_content_type":"text/x-rst","patch_set":2,"id":"836aa027_6a69079b","line":98,"updated":"2026-09-11 10:44:48.000000000","message":"Aside: I want to kill the iDRAC driver. We already have lots of vendor-specific code in the Redfish one. Dell does not maintain their driver. It\u0027s just a burden on us.","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"},{"author":{"_account_id":11655,"name":"Julia Kreger","email":"juliaashleykreger@gmail.com","username":"jkreger","status":"Flying to the moon with a Jetpack!"},"change_message_id":"c92d11088d680bd6e8f08e832a7e0d044ebfd37d","unresolved":true,"context_lines":[{"line_number":95,"context_line":"in one standard place. The generic ``redfish`` ``management`` implementation"},{"line_number":96,"context_line":"reads the attributes carried on the Manager resource itself, via sushy\u0027s"},{"line_number":97,"context_line":"``Manager.bmc``. The ``idrac`` hardware type -- whose management interface"},{"line_number":98,"context_line":"already subclasses the ``redfish`` one -- overrides where it looks and reads"},{"line_number":99,"context_line":"the Dell OEM ``DellAttributes`` resource instead, since that is where iDRAC"},{"line_number":100,"context_line":"keeps them. Either way the sync into ``bmc_settings`` and the"},{"line_number":101,"context_line":"``{\"Attributes\": {...}}`` update used to apply them are the same code, the"}],"source_content_type":"text/x-rst","patch_set":2,"id":"1a929ce6_ffac8e99","line":98,"in_reply_to":"836aa027_6a69079b","updated":"2026-09-25 19:08:50.000000000","message":"++","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"},{"author":{"_account_id":10239,"name":"Dmitry Tantsur","email":"dtantsur@protonmail.com","username":"dtantsur"},"change_message_id":"a5c7b039f34e3d8d365aa5c4012cf184148e6858","unresolved":true,"context_lines":[{"line_number":112,"context_line":"We could add a brand new ``bmc`` hardware interface, mirroring ``bios``. We"},{"line_number":113,"context_line":"rejected this as too much surface area for what it buys. Every hardware type"},{"line_number":114,"context_line":"and every out-of-tree driver would have to declare and usually no-op-implement"},{"line_number":115,"context_line":"the new interface, and we would need new configuration defaults and"},{"line_number":116,"context_line":"enabled-interface lists to go with it. BMC configuration is a natural part of"},{"line_number":117,"context_line":"node management, so a handful of optional methods on ``ManagementInterface``"},{"line_number":118,"context_line":"is the lighter, backwards-compatible path."}],"source_content_type":"text/x-rst","patch_set":2,"id":"a21fd15c_0c1aff60","line":115,"updated":"2026-09-11 10:44:48.000000000","message":"I don\u0027t think this statement is true. A lot of hardware types will (or at least should) inherit from Redfish nowadays, so they\u0027ll get the implementation for free and automatically.","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"},{"author":{"_account_id":11655,"name":"Julia Kreger","email":"juliaashleykreger@gmail.com","username":"jkreger","status":"Flying to the moon with a Jetpack!"},"change_message_id":"c92d11088d680bd6e8f08e832a7e0d044ebfd37d","unresolved":true,"context_lines":[{"line_number":112,"context_line":"We could add a brand new ``bmc`` hardware interface, mirroring ``bios``. We"},{"line_number":113,"context_line":"rejected this as too much surface area for what it buys. Every hardware type"},{"line_number":114,"context_line":"and every out-of-tree driver would have to declare and usually no-op-implement"},{"line_number":115,"context_line":"the new interface, and we would need new configuration defaults and"},{"line_number":116,"context_line":"enabled-interface lists to go with it. BMC configuration is a natural part of"},{"line_number":117,"context_line":"node management, so a handful of optional methods on ``ManagementInterface``"},{"line_number":118,"context_line":"is the lighter, backwards-compatible path."}],"source_content_type":"text/x-rst","patch_set":2,"id":"7c97e316_3e518f1e","line":115,"in_reply_to":"a21fd15c_0c1aff60","updated":"2026-09-25 19:08:50.000000000","message":"I guess the quesiton more so, becomes do we \"hide the collection/management\", or do we make it selectable. I guess under the existing model, if we base logic based upon the bmc vendor, we can identify the path, otherwise we have to build bmc handling into the driver model which I\u0027m not really a fan of because it sort of feels like we\u0027re creating \"the_other_management_interface\"","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"},{"author":{"_account_id":11655,"name":"Julia Kreger","email":"juliaashleykreger@gmail.com","username":"jkreger","status":"Flying to the moon with a Jetpack!"},"change_message_id":"c92d11088d680bd6e8f08e832a7e0d044ebfd37d","unresolved":true,"context_lines":[{"line_number":120,"context_line":"We considered making BMC settings writable directly over REST. We rejected it"},{"line_number":121,"context_line":"to stay consistent with BIOS: applying BMC or firmware configuration often"},{"line_number":122,"context_line":"needs a controlled reboot and asynchronous verification, which the clean and"},{"line_number":123,"context_line":"deploy step machinery already knows how to do."},{"line_number":124,"context_line":""},{"line_number":125,"context_line":"Data model impact"},{"line_number":126,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"93dbc148_ac183200","line":123,"updated":"2026-09-25 19:08:50.000000000","message":"So the big difference here is we don\u0027t really need reboots for most BMC actions. They are immediate or self-handled disjointed from the workload.","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"},{"author":{"_account_id":11655,"name":"Julia Kreger","email":"juliaashleykreger@gmail.com","username":"jkreger","status":"Flying to the moon with a Jetpack!"},"change_message_id":"c92d11088d680bd6e8f08e832a7e0d044ebfd37d","unresolved":true,"context_lines":[{"line_number":277,"context_line":"Nova driver impact"},{"line_number":278,"context_line":"------------------"},{"line_number":279,"context_line":""},{"line_number":280,"context_line":"None. BMC settings are not surfaced to Nova and do not affect scheduling or"},{"line_number":281,"context_line":"the ``nova.virt.ironic`` driver."},{"line_number":282,"context_line":""},{"line_number":283,"context_line":"Ramdisk impact"}],"source_content_type":"text/x-rst","patch_set":2,"id":"c3e05ee5_7ad011d7","line":280,"updated":"2026-09-25 19:08:50.000000000","message":"nit: This can just be \"None\"","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"},{"author":{"_account_id":11655,"name":"Julia Kreger","email":"juliaashleykreger@gmail.com","username":"jkreger","status":"Flying to the moon with a Jetpack!"},"change_message_id":"c92d11088d680bd6e8f08e832a7e0d044ebfd37d","unresolved":true,"context_lines":[{"line_number":283,"context_line":"Ramdisk impact"},{"line_number":284,"context_line":"--------------"},{"line_number":285,"context_line":""},{"line_number":286,"context_line":"None. BMC settings are read and applied out of band via the BMC; the"},{"line_number":287,"context_line":"ironic-python-agent ramdisk is not involved."},{"line_number":288,"context_line":""},{"line_number":289,"context_line":"Security impact"}],"source_content_type":"text/x-rst","patch_set":2,"id":"a7b33148_24b5595b","line":286,"updated":"2026-09-25 19:08:50.000000000","message":"Same as above.","commit_id":"527244e430166e321b10f3ab29fba326507f2a74"}]}
