[ironic] Summit PTG Summary
Hey all, As indicated before the summit, the Ironic PTG was dedicated to in-person collaboration, hacking, and design chats but were not intended for any final decision making due to not all of our community being present. Full, contemporaneous notes taken are at https://etherpad.opendev.org/p/ironic-openinfra-2023 -- this is meant to be a high level summary. For the summit, there were several extremely well-attended Ironic talks. Thank you to all of those who gave talks! In addition, I, and other members of the Ironic community, were able to connect with many people using Ironic quietly in production with a large amount of success. As always, the Ironic community strongly encourages people with success stories to loudly communicate about them :D. Ironic did have a single forum session, where we met with several operators, answering questions and in some places providing solutions to people struggling with a problem. The full notes from the session, primarily consisting of a census of Ironic installs, are accessible from the above linked etherpad. As for the PTG sessions, there were a few topics. These are extremely rough outlines of what was discussed; but again, no specific decisions were made. - networking-generic-switch -- Several contributors, including Baptise Jonglez and John Garbutt, discussed options for scaling NGS further up in the future, including enhancing it to support other protocols, such as VXLAN. Some of these discussions have already moved to the list; and I encourage folks to engage with Baptise to make our network tooling scale even more. - Future of Ironic -- We spoke for a while about the Ironic vision document created in Rocky, targeting approximately now: https://docs.openstack.org/ironic/latest/contributor/vision.html -- we've accomplished many of the items on the list, but what's next? -- Possibilities brainstormed included: --- Enhanced network interfaces that use SDN or DPU orchestration to configure baremetal networks --- more distinct support for composible hardware --- expanding Ironic standalone use cases --- getting more directly connected with communities like Metal3 integrating into Ironic --- scaling down Ironic into a tool useful at smaller scale (that a tool like cobbler has a strong hold on today) --- terraform driver designed to call Ironic directly -- I think we should update the vision document with some of these ideas so we can use it as a measuring stick in 5-6 years, like we were able to use the Rocky vision document this time. Finally, we closed up the summit with an Ironic dinner, with 16 attendees from various companies, use cases, and backgrounds with one thing in common: we all need bare metal servers :). If you're wondering who the faces of Ironic are, here are some of us https://twitter.com/jayofdoom/status/1669531671937384449 :). I'll say, on a personal note, it was extremely nice to get to see many of my old and new friends in the OpenInfra community face to face for the first time in years. The absence of those unable to travel was felt deeply as well, and I hope we'll be able to reconnect in person in the future. Thanks, Jay Faulkner Ironic PTL
On Wed, Jun 21, 2023 at 9:29 AM Jay Faulkner <jay@gr-oss.io> wrote:
Hey all,
As indicated before the summit, the Ironic PTG was dedicated to in-person collaboration, hacking, and design chats but were not intended for any final decision making due to not all of our community being present.
Full, contemporaneous notes taken are at https://etherpad.opendev.org/p/ironic-openinfra-2023 -- this is meant to be a high level summary.
For the summit, there were several extremely well-attended Ironic talks. Thank you to all of those who gave talks! In addition, I, and other members of the Ironic community, were able to connect with many people using Ironic quietly in production with a large amount of success. As always, the Ironic community strongly encourages people with success stories to loudly communicate about them :D.
Ironic did have a single forum session, where we met with several operators, answering questions and in some places providing solutions to people struggling with a problem. The full notes from the session, primarily consisting of a census of Ironic installs, are accessible from the above linked etherpad.
As for the PTG sessions, there were a few topics. These are extremely rough outlines of what was discussed; but again, no specific decisions were made. - networking-generic-switch -- Several contributors, including Baptise Jonglez and John Garbutt, discussed options for scaling NGS further up in the future, including enhancing it to support other protocols, such as VXLAN. Some of these discussions have already moved to the list; and I encourage folks to engage with Baptise to make our network tooling scale even more.
- Future of Ironic -- We spoke for a while about the Ironic vision document created in Rocky, targeting approximately now: https://docs.openstack.org/ironic/latest/contributor/vision.html -- we've accomplished many of the items on the list, but what's next?
I agree it is time to add a new entry for a new date to our vision document.
-- Possibilities brainstormed included: --- Enhanced network interfaces that use SDN or DPU orchestration to configure baremetal networks
I also had an interesting conversation about also allowing NGS to be invoked directly. We've sort of mused with the idea before, but I think maybe someone hacking on it might be the next logical step. Also, a number of different interests in use of DPUs, from "hiding" LACP from the workload to wanting to do VXLAN or geneve tunnel termination. Given I'm leading work to try and get us into a place to better support things such as this, I expect it will be a topic we'll be revisiting in the next few months.
--- more distinct support for composible hardware --- expanding Ironic standalone use cases --- getting more directly connected with communities like Metal3 integrating into Ironic --- scaling down Ironic into a tool useful at smaller scale (that a tool like cobbler has a strong hold on today)
Would a one shot command deploy thing be of value?
--- terraform driver designed to call Ironic directly
Consider this a solid +1! I was *really* surprised by the number of conversations I had where someone had a large BMaaS sort of cloud, and where they were presently using Terraform to drive Nova to ask for nodes from ironic. Flavor sprawl was an aspect that also came up because of a need for more or less matching down to represent required physical architectures for multi-machine workloads. I think it would be wise of us to try and push that capability forward so operators could leverage terraform with a whole deployed cloud or an Ironic only BMaaS cloud. At the same time, I would encourage us to also think of other use integrations and generally theme to make it easier to tie into processes and tools where operators presently take multiple steps.
-- I think we should update the vision document with some of these ideas so we can use it as a measuring stick in 5-6 years, like we were able to use the Rocky vision document this time.
Finally, we closed up the summit with an Ironic dinner, with 16 attendees from various companies, use cases, and backgrounds with one thing in common: we all need bare metal servers :). If you're wondering who the faces of Ironic are, here are some of us https://twitter.com/jayofdoom/status/1669531671937384449 :).
I'll say, on a personal note, it was extremely nice to get to see many of my old and new friends in the OpenInfra community face to face for the first time in years. The absence of those unable to travel was felt deeply as well, and I hope we'll be able to reconnect in person in the future.
Thanks, Jay Faulkner Ironic PTL
Hi! On 6/21/23 18:25, Jay Faulkner wrote:
Hey all,
As indicated before the summit, the Ironic PTG was dedicated to in-person collaboration, hacking, and design chats but were not intended for any final decision making due to not all of our community being present.
Full, contemporaneous notes taken are at https://etherpad.opendev.org/p/ironic-openinfra-2023 <https://etherpad.opendev.org/p/ironic-openinfra-2023> -- this is meant to be a high level summary.
For the summit, there were several extremely well-attended Ironic talks. Thank you to all of those who gave talks! In addition, I, and other members of the Ironic community, were able to connect with many people using Ironic quietly in production with a large amount of success. As always, the Ironic community strongly encourages people with success stories to loudly communicate about them :D.
Ironic did have a single forum session, where we met with several operators, answering questions and in some places providing solutions to people struggling with a problem. The full notes from the session, primarily consisting of a census of Ironic installs, are accessible from the above linked etherpad.
As for the PTG sessions, there were a few topics. These are extremely rough outlines of what was discussed; but again, no specific decisions were made. - networking-generic-switch -- Several contributors, including Baptise Jonglez and John Garbutt, discussed options for scaling NGS further up in the future, including enhancing it to support other protocols, such as VXLAN. Some of these discussions have already moved to the list; and I encourage folks to engage with Baptise to make our network tooling scale even more.
- Future of Ironic -- We spoke for a while about the Ironic vision document created in Rocky, targeting approximately now: https://docs.openstack.org/ironic/latest/contributor/vision.html <https://docs.openstack.org/ironic/latest/contributor/vision.html> -- we've accomplished many of the items on the list, but what's next? -- Possibilities brainstormed included: --- Enhanced network interfaces that use SDN or DPU orchestration to configure baremetal networks
I can haz standalone switch management? :)
--- more distinct support for composible hardware --- expanding Ironic standalone use cases --- getting more directly connected with communities like Metal3 integrating into Ironic --- scaling down Ironic into a tool useful at smaller scale (that a tool like cobbler has a strong hold on today)
Not clear from the etherpad: is it a question of scale or UX? I'm not sure a Bifrost installation is much larger than MaaS, etc (although we can always optimize it further, e.g. add sqlite as an option). The etherpad mentions "Easy enough", and ease of use is definitely the field where we Ironic leaves a lot to be desired. Possibilities that come to mind: 1) My "deployment API" proposal, 2) Adding a strict schema to our API and clean up confusing JSON fields, 3) An image building service, 4) The already mentioned network management, 5) More high-level API actions/workflows.
--- terraform driver designed to call Ironic directly
We have something like this, admittedly, limited to our case: https://github.com/openshift-metal3/terraform-provider-ironic/ Dmitry
-- I think we should update the vision document with some of these ideas so we can use it as a measuring stick in 5-6 years, like we were able to use the Rocky vision document this time.
Finally, we closed up the summit with an Ironic dinner, with 16 attendees from various companies, use cases, and backgrounds with one thing in common: we all need bare metal servers :). If you're wondering who the faces of Ironic are, here are some of us https://twitter.com/jayofdoom/status/1669531671937384449 <https://twitter.com/jayofdoom/status/1669531671937384449> :).
I'll say, on a personal note, it was extremely nice to get to see many of my old and new friends in the OpenInfra community face to face for the first time in years. The absence of those unable to travel was felt deeply as well, and I hope we'll be able to reconnect in person in the future.
Thanks, Jay Faulkner Ironic PTL
--- scaling down Ironic into a tool useful at smaller scale (that a tool
like cobbler has a strong hold on today)
Not clear from the etherpad: is it a question of scale or UX? I'm not sure a Bifrost installation is much larger than MaaS, etc (although we can always optimize it further, e.g. add sqlite as an option).
Yes to all? It's more that this is a space that we have very little market penetration into, and I'd like to solve it. I think there's going to need to be a nexus of documentation, simplification, and fighting back against false perceptions to make progress here, but it's worthwhile because of the size of the potential growth.
The etherpad mentions "Easy enough", and ease of use is definitely the field where we Ironic leaves a lot to be desired. Possibilities that come to mind: 1) My "deployment API" proposal, 2) Adding a strict schema to our API and clean up confusing JSON fields, 3) An image building service, 4) The already mentioned network management, 5) More high-level API actions/workflows.
These are all really good ideas that go in the direction I was thinking :). - Jay Faulkner Ironic PTL
participants (3)
-
Dmitry Tantsur
-
Jay Faulkner
-
Julia Kreger