| * TheJulia will ponder on leaf_id vs physnet with sleep | 01:17 | |
| TheJulia | cardoe: wwiw, I was thinking if a port is in a storage state, it becomes disqualified for ip traffic | 01:18 |
|---|---|---|
| cardoe | We'd have to delete the port or add a flag of some sort. | 01:19 |
| cardoe | I've also got annoying OpenStack users like James (who isn't on IRC otherwise I'd poke him) who is building an OpenStack deployment on top of my OpenStack deployment. | 01:20 |
| cardoe | And I've got to provision him a SVM on the NetApp and let him run Nova with os-brick talking to that NetApp. | 01:21 |
| cardoe | insert spiderman meme | 01:22 |
| cardoe | or turtles jokes | 01:22 |
| cardoe | So while yes I'm 100% focused on the Ironic attaches at the BMC. I do know I'll have to drive another fabric. | 01:23 |
| cardoe | The other part of a "leaf ID" is that you would keep that value the same for both switches that are providing that port group. | 01:25 |
| cardoe | So n-g-s would know about that pair. | 01:26 |
| opendevreview | David Nwosu proposed openstack/ironic master: Add Getting Started guide https://review.opendev.org/c/openstack/ironic/+/965189 | 06:48 |
| rpittau | good morning ironic! o/ | 06:52 |
| queensly[m] | Good morning o/ | 08:44 |
| Ayo[m] | Good morning ironic :) | 08:45 |
| Ayo[m] | Hello queensly :) | 08:46 |
| opendevreview | Verification of a change to openstack/ironic stable/2025.2 failed: Add upgrade check for iLO/iLO5 driver deprecation https://review.opendev.org/c/openstack/ironic/+/965158 | 10:38 |
| TheJulia | cardoe: Thinking about it, the leaf delineation does sort of work in the vxlan case. But it just confuses the physical vlan delineation case because your building a layer2 fabric in one case where as your sharing a layer2 fabric in another. I think if we just document and treat ":" as the delineation character, so cross-matching can still sort of work where appropriate, then that could be the happiest path overall | 12:18 |
| opendevreview | Doug Szumski proposed openstack/networking-generic-switch master: Add Dell Enterprise Sonic-CLI driver https://review.opendev.org/c/openstack/networking-generic-switch/+/965373 | 12:52 |
| opendevreview | Julia Kreger proposed openstack/ironic-tempest-plugin master: Add stable/2025.2 job https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/961390 | 12:53 |
| opendevreview | Julia Kreger proposed openstack/ironic-tempest-plugin master: Clean up CI jobs https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/965374 | 12:53 |
| iurygregory_ | PTG time | 12:57 |
| *** iurygregory_ is now known as iurygregory | 12:57 | |
| TheJulia | I will be a minute or two late, my coffee maker decided to be a jerk | 12:59 |
| JayF | I'm taking my time this morning, I don't have much interest or input in the first two topics. | 13:01 |
| opendevreview | Doug Szumski proposed openstack/networking-generic-switch master: Add Dell Enterprise SONiC CLI driver https://review.opendev.org/c/openstack/networking-generic-switch/+/965373 | 13:07 |
| JayF | I am having a chat with some folks this afternoon about connecting dpus up to ovn. Sounds like based on our chat yesterday, that we have a general idea of the shape of what that feature would be? | 13:11 |
| JayF | I wouldn't mind if someone can spare 15 or 20 minutes after the ptg is over to help me get this pitch right. If I'm able to make the case of this meeting, I almost certainly will be pointing a lot of gross at this problem. | 13:12 |
| TheJulia | JayF: yeah, I figured out an idea during the summit but haven't written it up | 13:17 |
| JayF | If you wanna give me 15 minutes to get the high points into an etherpad, I may just be able to run with it (depending on my knowledge + outcome of meeting) | 13:17 |
| JayF | I think during that chat with Doug yesterday the idea popped out, at least I think it did | 13:17 |
| JayF | mainly wanting to make sure my understandings are sane | 13:17 |
| JayF | clif: ^ if you can be around for that chat too it'd be awesome | 13:34 |
| JayF | or at least mea nd you work out a plan if it's just us :D | 13:35 |
| JayF | cardoe: I'm just warning folks to gather pitchforks and torches /s | 13:38 |
| opendevreview | David Nwosu proposed openstack/ironic master: Add getting started guide for Ironic https://review.opendev.org/c/openstack/ironic/+/965189 | 13:46 |
| opendevreview | Riccardo Pittau proposed openstack/ironic master: Handle HTTP 400 and 409 race condition in Redfish power operations https://review.opendev.org/c/openstack/ironic/+/965248 | 13:54 |
| clif | cardoe: I'm not sure if this is what you're looking for but I feel like the new 'category' fields on ports and portgroups could be used as a differentiation field pretty easily | 14:05 |
| clif | trying to ingest more scrollback: are you trying to store vxlan IDs inside `physical_network`? | 14:10 |
| cardoe | Would that conflict with your intended use case? | 14:10 |
| cardoe | No. Segment name | 14:10 |
| clif | afaict `port{group}.category` (referred to as `class` in the spec...) doesn't have a strict use case other than being another way for trait based networking to, well, categorize port{groups} | 14:11 |
| clif | so to me... I have no problem with an operator using that field however is helpful | 14:13 |
| clif | although I did not write the spec so maybe JayF would disagree | 14:13 |
| JayF | the choice of fields in that spec was carefully thought out^W^W^W a jumble of fields that seemed useful at the time | 14:14 |
| clif | :D | 14:14 |
| JayF | I think the success case for this would lead to an expansion of information there on both sides of the match | 14:15 |
| JayF | e.g. less-neutron coded stuff for standalone or more port/portgroup fields as those get more complex | 14:15 |
| JayF | this is more of a "lets pour the foundation and see what houses people wanna build" feature | 14:16 |
| opendevreview | Julia Kreger proposed openstack/ironic-python-agent master: Remove PReP support https://review.opendev.org/c/openstack/ironic-python-agent/+/965390 | 14:16 |
| JayF | except we already knew we wanted 4 walls and a roof | 14:16 |
| JayF | when cardoe goes asking for a solarium we'll be ready ;) | 14:16 |
| JayF | that 512MB limitation on dracs, that's VMedia size, yeah? | 14:16 |
| JayF | **idrac10 | 14:17 |
| TheJulia | looks like it applies to network as well | 14:18 |
| TheJulia | Its the amount of data which a bootloader is permitted to work with | 14:18 |
| JayF | that fits with what I thought, thanks for telling me | 14:18 |
| JayF | and THANKS DELL | 14:18 |
| TheJulia | so iso -> grub -> 500MB of ramdisk | 14:19 |
| * JayF flips table | 14:19 | |
| TheJulia | or PXE -> grub|ipxe -> 500MN | 14:19 |
| JayF | so basically for real hardware which might need a firmware or two | 14:19 |
| JayF | you MUST use a two-stage initrd | 14:19 |
| TheJulia | we've not been able to verify PXE since the gear is not in a lab | 14:19 |
| TheJulia | well, a lab where we can control PXE | 14:19 |
| JayF | boot minimal for network -> fetch second stage full FS? | 14:19 |
| TheJulia | It felt like, when we discussed doing that yesterday that the juice might not be worth the squeeze | 14:20 |
| JayF | how do you fit IPA into 512mb? | 14:20 |
| TheJulia | at least for upstream, if folks need to double stage then so be it and we need to figure out a way | 14:20 |
| JayF | when you need like, DPU firmwares | 14:20 |
| TheJulia | Well, it is only 512MB *when* your in UEFI runtime | 14:20 |
| JayF | I don't know what the alternative is | 14:20 |
| JayF | ooooh | 14:20 |
| TheJulia | as soon as the kernel inits | 14:20 |
| JayF | so you have to be <512MB compressed at ipxe time | 14:21 |
| TheJulia | you get all the system memory, you just need to get your starting bits in that small footprint | 14:21 |
| TheJulia | yas | 14:21 |
| JayF | so we can outsource some % of the problem to zstd | 14:21 |
| TheJulia | sort of yeah | 14:21 |
| TheJulia | but even *then* its a loosing battle | 14:22 |
| TheJulia | see: centos10 ramdisk sizes | 14:22 |
| JayF | I've been told even with zstd the compressed ramdisk we pass around is ~1G | 14:22 |
| TheJulia | I've yet to tear it apart yet, fwiw | 14:22 |
| TheJulia | huh? | 14:22 |
| TheJulia | 16? | 14:22 |
| JayF | at my downstream, the compressed initrd, already compressed with zstd, is >1GB | 14:23 |
| JayF | so I'm going to have this problem | 14:23 |
| TheJulia | yeeeeeeee ouch | 14:23 |
| TheJulia | So maybe there is juice to squeeze here in a way to have a two stage pattern? | 14:24 |
| JayF | likely, I'll talk with my downstream to ensure that we use upstream tooling before I get into it | 14:27 |
| JayF | yeah, DIB+IPA-B, so I'll add it to my list | 14:27 |
| TheJulia | I think there may be a viable "here is the IPA container" pattern, and "here is the ramdisk to boot" pattern, but it feels like that slice might not be much better | 14:28 |
| TheJulia | cardoe: we miss you! | 14:48 |
| cardoe | TheJulia: sorry! I'm here. | 14:53 |
| JayF | TheJulia: ironic be https://www.youtube.com/watch?v=5DDZwqnQKbM ? :D | 15:12 |
| rpittau | lol | 15:13 |
| janders | ++ | 15:14 |
| cardoe | JayF: I've had this argument a bajillion times. | 15:49 |
| cardoe | cloud-init supports everything and ponies and extensions | 15:49 |
| cardoe | I think its possible to configure IP over Avian Carrier with it. | 15:50 |
| cardoe | It's the datasource reader that needs to support each of the features and extensions. | 15:50 |
| cardoe | The datasource reader for OpenStack says that it will only support spec conforming fields. | 15:50 |
| cardoe | And the nova folks own the spec. | 15:50 |
| iurygregory | we got our first answer in the sushy thread | 16:05 |
| * TheJulia files a perma-negative-1 to Metal-via-Avian-Carriers | 16:05 | |
| iurygregory | but I think the person thought it was sushy-tools... | 16:05 |
| iurygregory | cardoe, can you join again? | 16:21 |
| iurygregory | maybe you had thoughts/ use case | 16:21 |
| JayF | team photo https://usercontent.irccloud-cdn.com/file/Sqyp1B2P/IRONIC_PTG_2026.1.png | 16:49 |
| JayF | TheJulia: .o(we have a spec for this at all?) | 17:15 |
| TheJulia | Vxlan on the ironic side more as a mental scratch pad in writing | 17:16 |
| JayF | ack | 17:16 |
| JayF | this is cardoe's vxlan plan + a little NGS code, it seems | 17:17 |
| JayF | and if not vxlan, just a little NGS code | 17:17 |
| JayF | which means with hardware, and assuming docs are right (lol) it shouldn't be a big lift | 17:17 |
| cardoe | Sorry. Just juggling moving around. Need me to join? | 17:20 |
| JayF | we're all done | 17:20 |
| cardoe | Car was being naughty. Still being naughty unfortunately so no success. | 17:21 |
| cardoe | Hardware is naughty no matter the form factor basically. | 17:22 |
| cardoe | I’ll push some docs for that VXLAN stuff in a second. | 17:23 |
| opendevreview | David Nwosu proposed openstack/ironic master: Add getting started guide for Ironic https://review.opendev.org/c/openstack/ironic/+/965189 | 17:25 |
| opendevreview | David Nwosu proposed openstack/ironic master: Add getting started guide for Ironic https://review.opendev.org/c/openstack/ironic/+/965189 | 17:30 |
| cardoe | TheJulia, JayF, clif, alegacy, janders: (i think thats everyone that expressed interest in the VXLAN) I've made https://review.opendev.org/c/openstack/neutron/+/965415 which attempts to start documenting the use cases and how they interact. No doubt I'll have missed the mark a little bit. But happy to evolve it. | 17:33 |
| cardoe | Please weigh in. | 17:35 |
| opendevreview | Doug Szumski proposed openstack/networking-generic-switch master: Add Dell Enterprise SONiC CLI driver https://review.opendev.org/c/openstack/networking-generic-switch/+/965373 | 17:35 |
| TheJulia | cardoe: ack, fwiw, you also missed a related discussion, I've got a todo to update some stuff as well | 17:42 |
| TheJulia | I wonder how many mechanics are now scratching their head at software race conditions...\ | 17:43 |
| TheJulia | cardoe: wow that is a small book :) | 18:53 |
| cardoe | And I cut it down. But I wanted to make it clear. | 18:58 |
| opendevreview | Merged openstack/ironic-tempest-plugin master: Clean up CI jobs https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/965374 | 19:16 |
| opendevreview | Merged openstack/ironic-tempest-plugin master: Add stable/2025.2 job https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/961390 | 19:16 |
| opendevreview | Jay Faulkner proposed openstack/ironic-tempest-plugin master: Change Portgroup minimum microversion to 1.26 https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/955799 | 19:42 |
| cardoe | ^ rebase by me which should free up clif's changes to land | 19:43 |
| clif | I think we're still blocked by this not getting merged yet: https://review.opendev.org/c/openstack/ironic/+/964810 | 19:44 |
| cardoe | did you see why it failed and give it a recheck if it was temp? | 19:49 |
| TheJulia | with the jobs cleaned up, should be happier | 19:50 |
| cardoe | looks like keystone failed in all of those 2 days ago. | 19:50 |
| cardoe | I'd venture a guess that devstack has been fixed in those 2 days. | 19:50 |
| TheJulia | cardoe: finally finished your doc, a couple comments but otherwise looks great to me | 19:55 |
| cardoe | Thank you. I'm actually reading it and seeing I got lazy on the VLANs and changing the right side to 105 | 19:55 |
| cardoe | ha most of your comment areas are the last areas I worked on and I sort of petered out in steam. | 19:59 |
| TheJulia | heh | 20:01 |
| JayF | alegacy: you are my hero today and don't even know it. Thank you for the standalone networking work. :D | 20:24 |
| alegacy | JayF: :-) | 20:25 |
| JayF | do you have it working end to end with NGS yet in a lab? | 20:26 |
| JayF | if so I may have it working end to end with NGS in a lab soon too ;) | 20:26 |
| alegacy | yes, I even have it prototyped into metal3 and working in openshift. Happy to give you a demo in the near future if you'd like. | 20:26 |
| JayF | I need to review all of it and get the patches together | 20:29 |
| JayF | we're going to do that + the DPU-configure-NGS switch driver | 20:29 |
| JayF | to try and get the magic cloud sdn working :D | 20:29 |
| *** darmach9 is now known as darmach | 20:33 | |
| alegacy | JayF: I'm interested in learning a bit more about what you're trying to accomplish. Maybe we could have a quick chat next week if you've got a bit of time? | 20:35 |
| JayF | the basic is this: standalone ironic w/networking support via NGS, combined with a new NGS plugin to treat DPUs as a switch | 20:35 |
| JayF | basically just your code + an NGS plugin | 20:36 |
| alegacy | cool | 20:36 |
| JayF | with some minor additional details | 20:36 |
| TheJulia | I just realized: The PTG is the cliffs of insanity. | 20:39 |
| clif | I'm still sane I think | 20:41 |
| JayF | yeah but you're only an "fs" away from the cliffs | 20:41 |
| JayF | just if anyone tries to tell you to run ceph, run away or else you'll get your insane fs | 20:43 |
| JayF | :D | 20:43 |
| JayF | l;ol | 20:43 |
| TheJulia | clif: that is a good accomplishment then, because I'm totally crazy now | 20:55 |
| JayF | I don't blame the PTG | 20:55 |
| JayF | I blame 6.5 hours of video meetings in an 8 hours workday | 20:55 |
| JayF | and the caff required to survive them | 20:55 |
| TheJulia | ++++ | 20:57 |
| JayF | but today was about as good as such a day can be | 20:57 |
| TheJulia | Indeed | 20:57 |
| TheJulia | wow nvidia has drastically changed their bluefield docs | 21:29 |
| cardoe | dtantsur: just wanted to follow up on https://review.opendev.org/c/openstack/sushy/+/962680 You +1'd it with some nits that prevented you from a +2. Nidhi addressed all those nits in a follow on. | 21:40 |
| cardoe | And https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/955799 is now green if that gets reviews and +W'd then clif's already +W'd stuff will land | 21:47 |
| TheJulia | Regarding the DPU idea: https://bugs.launchpad.net/networking-generic-switch/+bug/2130384 | 22:04 |
| TheJulia | super duper hand-wavey | 22:04 |
| TheJulia | and we would likely need to sort out vendor specific | 22:04 |
| TheJulia | details | 22:04 |
| opendevreview | Takashi Kajinami proposed openstack/ironic-tempest-plugin master: Clean up unused stable branch jobs https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/961391 | 22:41 |
| opendevreview | Merged openstack/ironic stable/2024.2: More RAM for Anaconda job https://review.opendev.org/c/openstack/ironic/+/964810 | 22:52 |
Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
