| opendevreview | Jay Faulkner proposed openstack/ironic master: Support Automatic Lessee from instance metadata https://review.opendev.org/c/openstack/ironic/+/924887 | 01:15 |
|---|---|---|
| opendevreview | Iury Gregory Melo Ferreira proposed openstack/ironic master: Firmware Update via Firmware Interface Docs https://review.opendev.org/c/openstack/ironic/+/926961 | 01:38 |
| opendevreview | Takashi Kajinami proposed openstack/ironic master: Remove skip check for Python 3.6 https://review.opendev.org/c/openstack/ironic/+/927780 | 07:06 |
| opendevreview | Adam Rozman proposed openstack/ironic-python-agent master: WIP - root device encryption https://review.opendev.org/c/openstack/ironic-python-agent/+/926425 | 11:24 |
| opendevreview | cid proposed openstack/ironic master: Upgrade pysnmp-lextudio https://review.opendev.org/c/openstack/ironic/+/927804 | 11:42 |
| opendevreview | cid proposed openstack/ironic master: Upgrade pysnmp-lextudio https://review.opendev.org/c/openstack/ironic/+/927804 | 11:48 |
| *** jlejeune_ is now known as jlejeune | 12:30 | |
| cardoe | Good morning ironic | 12:43 |
| TheJulia | Good morning! | 13:12 |
| opendevreview | cid proposed openstack/ironic master: Upgrade pysnmp-lextudio https://review.opendev.org/c/openstack/ironic/+/927804 | 13:38 |
| cardoe | I hope it's okay but I closed out a few launchpad issues that said "fix committed" but now that sushy 5.2.0 is released I moved them to "fix released" | 13:39 |
| cardoe | dtantsur: ping to release sushy-oem-idrac and potentially abandon https://review.opendev.org/c/openstack/ironic/+/902107 | 13:43 |
| dtantsur | thanks! | 13:43 |
| cardoe | Did you guys want to look at bringing sushy-oem-idrac into sushy proper in the future? | 13:44 |
| dtantsur | It was my idea, at least. I don't know what the others think | 13:44 |
| dtantsur | Before I do the release, is anyone available to review https://review.opendev.org/c/openstack/sushy-oem-idrac/+/917021 and https://review.opendev.org/c/openstack/sushy-oem-idrac/+/917020 ? | 13:44 |
| dtantsur | TheJulia or JayF maybe ^^ | 13:44 |
| * TheJulia looks up from vim | 13:45 | |
| cardoe | https://review.opendev.org/c/openstack/ironic/+/926899 should be good for a workflow review or a core reviewer. That would have given some of my folks a friendlier message. | 13:46 |
| * dtantsur does not object :) | 13:46 | |
| TheJulia | Any other quick reviews? | 13:47 |
| cardoe | https://review.opendev.org/q/project:openstack/sushy-oem-idrac all I see that's opened. | 13:48 |
| cardoe | I'm working on bringing our LLDP inspector code to sushy and sushy-oem-idrac | 13:48 |
| TheJulia | workflowed | 13:48 |
| cardoe | But that'll be another release. | 13:48 |
| dtantsur | TheJulia: thx! | 13:49 |
| opendevreview | Merged openstack/sushy-oem-idrac master: Add OEM DellAttributes to DellManagerExtension https://review.opendev.org/c/openstack/sushy-oem-idrac/+/917021 | 13:50 |
| opendevreview | Julia Kreger proposed openstack/ironic-python-agent master: WIP: Provide operator an idea why the deploy failed https://review.opendev.org/c/openstack/ironic-python-agent/+/927823 | 13:51 |
| TheJulia | So on the subject of 4k block devices: what do folks think about ^ | 13:51 |
| dtantsur | TheJulia: generally in favour, although: | 13:53 |
| dtantsur | $ sudo file /dev/nvme0n1p1 | 13:53 |
| dtantsur | /dev/nvme0n1p1: block special (259/1) | 13:53 |
| dtantsur | does not seem to match what you expect? | 13:53 |
| cardoe | Yeah I'm a +1 on better error messages / details. | 13:54 |
| opendevreview | Merged openstack/sushy-oem-idrac master: Added method for getting kvm session https://review.opendev.org/c/openstack/sushy-oem-idrac/+/917020 | 13:56 |
| cardoe | your error message says "looing" and I'm thinking that's Freud coming out in how you feel about 4k blocks | 13:56 |
| * dtantsur learns new English words nearly daily, lol | 13:57 | |
| dtantsur | cardoe: here goes https://pypi.org/project/sushy-oem-idrac/6.0.0/ | 14:06 |
| cardoe | awesome. thank you | 14:06 |
| TheJulia | dtantsur: nvme is... special that way | 14:08 |
| TheJulia | I've not figured out what to do with that | 14:08 |
| cardoe | So I know we commit the fake redfish json data from their spec for testing. How would you folks feel if I scrapped a real box as well? | 14:08 |
| TheJulia | since it *should* return raw data on the target device | 14:09 |
| TheJulia | p1 is also the partition | 14:09 |
| dtantsur | cardoe: it's probably a copyright issue | 14:09 |
| dtantsur | janders actually maintains a repo with scrapped real boxes internally, but I doubt we can put something like that in an opensource repo | 14:09 |
| opendevreview | Derek Higgins proposed openstack/ironic master: Set node "alive" when inspection finished https://review.opendev.org/c/openstack/ironic/+/927828 | 14:09 |
| opendevreview | Dmitry Tantsur proposed openstack/sushy-oem-idrac master: Replace github links in README https://review.opendev.org/c/openstack/sushy-oem-idrac/+/927829 | 14:09 |
| dtantsur | a trivial (and sad) change to approve ^^^ | 14:09 |
| TheJulia | approved | 14:10 |
| TheJulia | cardoe: Freud might be coming out, but I was also messaged by folks while writing just that for dates on 4k devies to be "fully supported" | 14:11 |
| cardoe | Since we're talking sushy, I'm not sure if https://review.opendev.org/c/openstack/sushy/+/927670 is correct but it would fix the issue for that user. | 14:16 |
| dtantsur | cardoe: hardware vendors and specifications, right? :-/ | 14:17 |
| TheJulia | dtantsur: so I *could* just copy off the first megabyte of data from the nvme device using dd and file that | 14:17 |
| TheJulia | that works like a charm to detect a filesystem | 14:18 |
| dtantsur | cardoe: I don't mind this change, but maybe we should apply it consistently to all resources? | 14:18 |
| dtantsur | TheJulia: lol, ship it! | 14:18 |
| TheJulia | oh | 14:20 |
| TheJulia | add -s to the command and it should be good | 14:20 |
| TheJulia | yup, that works | 14:21 |
| TheJulia | "treat special files as ordinary files" | 14:21 |
| dtantsur | yep | 14:30 |
| dtantsur | btw this is how the UEFI partition looks: | 14:30 |
| dtantsur | $ sudo file -s /dev/nvme0n1p1 | 14:30 |
| dtantsur | /dev/nvme0n1p1: DOS/MBR boot sector, code offset 0x58+2, OEM-ID "mkfs.fat", sectors/cluster 8, Media descriptor 0xf8, sectors/track 32, heads 64, hidden sectors 2048, sectors 1048576 (volumes > 32 MB), FAT (32 bit), sectors/FAT 1024, reserved 0x1, serial number 0xb83a89ab, unlabeled | 14:30 |
| dtantsur | TheJulia: not sure if you handle this ^^ | 14:30 |
| TheJulia | that is your first partition | 14:30 |
| TheJulia | the check should be looking at the overall device | 14:30 |
| dtantsur | I mean, you're checking for ext4, xfs, btrfs and ntfs | 14:31 |
| dtantsur | do you want to check for FAT32 and ISO 9660? | 14:31 |
| TheJulia | Hmm... I guess I could add "mkfs.fat" to it | 14:31 |
| dtantsur | I can easily imagine someone accidentally writing an ISO image, for instance | 14:31 |
| TheJulia | vfat as in fat32, at least locally came back as "filesystem data" | 14:31 |
| TheJulia | I'll add on to it | 14:32 |
| dtantsur | wanna be fancy, also add Linux Logical Volume Manager | 14:33 |
| opendevreview | cid proposed openstack/ironic master: Upgrade pysnmp-lextudio https://review.opendev.org/c/openstack/ironic/+/927804 | 14:33 |
| TheJulia | oh, yes | 14:33 |
| TheJulia | if you have one handy! | 14:33 |
| dtantsur | $ sudo file -s /dev/nvme0n1p3 | 14:33 |
| dtantsur | /dev/nvme0n1p3: LVM2 PV (Linux Logical Volume Manager), UUID: hn3P2a-hksA-UlUQ-UFmW-6wuZ-WeXP-5MeETY, size: 479028314112 | 14:33 |
| cardoe | dtantsur: so in some places we use "identity" and in some places we use odata.id | 14:33 |
| dtantsur | consistency \o/ | 14:34 |
| cardoe | I was thinking odata.id is more correct | 14:34 |
| cardoe | So I was gonna update to that. | 14:34 |
| cardoe | I was talking to JayF about adding pre-commit from ironic (and looking at what some other projects do) | 14:34 |
| cardoe | Anyway, I'm happy to make it consistent and abandon that change. I was just proposing that for 1 user. | 14:35 |
| dtantsur | as I said, I don't mind working around a real world problem | 14:35 |
| dtantsur | just wanted to give consistency a thought | 14:35 |
| cardoe | Technically the "Id" and "Name" field are required on all Redfish models. But we don't require them on a lot and the links are suppose to be via odata.id so I don't see why "Id" and "Name" need to be required. | 14:35 |
| dtantsur | I don't see why Id is even needed.. | 14:36 |
| dtantsur | Redfish is wildly inconsistent in requiring or not requiring Name | 14:36 |
| dtantsur | I've literally fixed one instance in the RAID code recently | 14:36 |
| cardoe | Absolutely. So lemme do a bigger consistency clean up and we can abandon that one change (unless you wanna land it for that user with the bug). | 14:36 |
| dtantsur | Up to you | 14:37 |
| cardoe | So the spec says Name is required but their own test data misses Name a large percentage of the time. | 14:37 |
| TheJulia | so, I wonder how many bioses would actually boot an ISO9660 filesystem written to a disk | 14:37 |
| cardoe | I guess for the user's sake lets land it and then I'll work on the bigger clean up | 14:37 |
| cardoe | I have a convo later this week about Dell testing with the Ironic interop profile. | 14:38 |
| dtantsur | TheJulia: hmmmmmmmm. good point. | 14:38 |
| TheJulia | dtantsur: *today* that *just won't work* | 14:38 |
| TheJulia | so, I could see checking and erroring, for now | 14:38 |
| TheJulia | but I wouldn't object to a slightly smarter path | 14:39 |
| TheJulia | then again.... | 14:39 |
| TheJulia | there is no guarentee the operator's ISO and onboard logic will work properly in that sort of case as well | 14:39 |
| opendevreview | Doug Goldstein proposed openstack/sushy master: WIP: add Port resource https://review.opendev.org/c/openstack/sushy/+/927838 | 14:40 |
| cardoe | So no tests for that but just to get an idea of what I was aiming for. | 14:41 |
| cardoe | There's LLDP info in those hills. So want to use that for the redfish inspector. | 14:41 |
| dtantsur | It's even present in actual real hardware? | 14:42 |
| cardoe | Unfortunately Dell doesn't use the standardized fields and uses their own OEM section. So I'll have to add that into sushy-oem-idrac | 14:42 |
| dtantsur | ahhh, of course | 14:42 |
| cardoe | oh yeah my non-Dell stuff returns LLDP in those fields. | 14:42 |
| dtantsur | \o/ | 14:42 |
| dtantsur | This is really nice, I'd never expect it | 14:42 |
| cardoe | Part of my convo this week is to waggle a PO in Dell's face if they fix some of this stuff. | 14:42 |
| dtantsur | If we get it working, this will significantly reduce the need for in-band inspection | 14:43 |
| dtantsur | cardoe++ | 14:43 |
| TheJulia | wheeee, by default file only detects ExFAT if you add -k to it | 14:43 |
| TheJulia | otherwise it just goes "this is an dos mbr" | 14:43 |
| cardoe | So our goal is to make the out of band inspector just as feature full as the in band inspector. | 14:43 |
| cardoe | I'm just not sure the best way to upstream this. | 14:43 |
| cardoe | Because I don't know if upstream really wants to see "idrac-redfish" inspector interface make a comeback. | 14:44 |
| cardoe | Or if we merge sushy-oem-idrac into sushy proper | 14:44 |
| dtantsur | The way things work now, you'll need to make idrac-redfish come back | 14:44 |
| dtantsur | How we want to move forward, is a great PTG discussion IMO | 14:45 |
| cardoe | If we just make interfaces like today we've got https://opendev.org/openstack/sushy/src/commit/033287f8c2d254cd3a51dd2ba2e414e04311eb16/sushy/resources/system/ethernet_interface.py#L59 | 14:45 |
| opendevreview | Julia Kreger proposed openstack/ironic master: docs: Add context around asynchronous device initialization https://review.opendev.org/c/openstack/ironic/+/927518 | 14:45 |
| dtantsur | We already have quite a few workarounds for Supermicro, Lenovo and Huawei right in sushy | 14:45 |
| cardoe | Okay. Next PTG I'll add it as an item. | 14:46 |
| dtantsur | ++ | 14:46 |
| cardoe | I'm happy to figure out a way to donate hardware to test against as well. | 14:46 |
| cardoe | Hence why the ask around real data. | 14:46 |
| dtantsur | an upstream 3rd party CI could be awesome | 14:47 |
| TheJulia | we, unfortunately, have also seen some vendors, dell included, changing stuff between releases where they didn't think anyone would notice | 14:55 |
| cardoe | Yep. I plan to hold their feet to the fire. | 14:58 |
| cardoe | I had a decently sized PO in my hands. | 14:58 |
| opendevreview | Merged openstack/ironic master: Better handle missing inspection_network https://review.opendev.org/c/openstack/ironic/+/926899 | 14:58 |
| TheJulia | hehe :) | 14:58 |
| cardoe | They've made a number of engineering resources available to me and setup some calls. | 14:59 |
| cardoe | They're at least going to have to listen to me give them a disappointed dad lecture. | 15:00 |
| opendevreview | Merged openstack/sushy-oem-idrac master: Replace github links in README https://review.opendev.org/c/openstack/sushy-oem-idrac/+/927829 | 15:02 |
| dtantsur | sounds fun :D | 15:03 |
| cardoe | If you guys are on a merge kick... https://review.opendev.org/c/openstack/sushy/+/927444 https://review.opendev.org/c/openstack/sushy/+/927598 are two small ones. | 15:07 |
| cardoe | and https://review.opendev.org/c/openstack/sushy/+/926995 | 15:07 |
| dtantsur | Any chances for a 2nd +2 on https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/925981 and https://review.opendev.org/c/openstack/ironic/+/926647? | 16:36 |
| groundnuty | Hello, I'm trying to create two software raids using ironic-python-agent. This is a target_raid_config that is applied: {'logical_disks': [{'controller': 'software', 'physical_disks': [{'serial': 's== S664NN0W200258'}, {'serial': 's== S664NN0W200257'}], 'raid_level': '1', 'size_gb': 447}, {'controller': 'software', 'physical_disks': [{'serial': 's== | 16:52 |
| groundnuty | BTYJ2296059C1P9CGN'},{'serial': 's== BTYJ22960A7H1P9CGN'}], 'raid_level': '1', 'size_gb': 'MAX'}]} | 16:52 |
| groundnuty | I would expect two mdadm raids to be created each having 2 disks. While the first one is created, the second one for some reason uses wrong index for partitions: | 16:54 |
| groundnuty | mdadm --create /dev/md1 --force --run --metadata=1 --level 1 --name md1 --raid-devices 2 /dev/sdh2 /dev/sdg2 | 16:54 |
| groundnuty | there is no /dev/sdh2 but I assume there is /dev/sdh1 | 16:54 |
| groundnuty | the guilty party is this line of code that for reasons I cannot understand assumes that each new raid will have a higher partition number: https://github.com/openstack/ironic-python-agent/blob/master/ironic_python_agent/raid_utils.py#L222 | 16:55 |
| groundnuty | please advice, if it's me who maybe does not understand the API or is it a bug and my usecase is not supported | 16:56 |
| TheJulia | hmm | 16:57 |
| TheJulia | So I *know* there are specific restrictions around software raid support | 16:57 |
| TheJulia | I | 16:58 |
| TheJulia | I'm not sure two mirror sets are in that matrix of what it will do out of the box | 16:58 |
| TheJulia | But, it has been a long time since I've looked at that code. Largely CERN folks did that code to support their use cases and I've had a couple customers leverage it as well, but not as a mirror set for a second volume. | 16:59 |
| TheJulia | hmm, I'm not sure I understand exactly what the raid_utils is trying to do off hand | 17:00 |
| * TheJulia also has a migraine brewing | 17:00 | |
| opendevreview | Derek Higgins proposed openstack/ironic master: Set node "alive" when inspection finished https://review.opendev.org/c/openstack/ironic/+/927828 | 17:00 |
| groundnuty | I come come from https://github.com/metal3-io/baremetal-operator usecase and we intend to run k8s clusters using metal3, our servers came with 2 sets of disks - one for boot and second for the rest. Hence the need for two raids. | 17:02 |
| groundnuty | I know from the longs that for the first raid when size_gb is set to 447 it runs: parted /dev/sdg -s -a optimal -- mkpart primary 551MiB 480GiB | 17:03 |
| groundnuty | so the /dev/sdg1 is created and the first raid is also created | 17:03 |
| TheJulia | groundnuty: I think the coded intent was the second set was RAID5, fwiw | 17:10 |
| TheJulia | that does sound correct, fwiw | 17:11 |
| JayF | I'd also suggest generally, if I had your problem, I'd have Ironic setup the root disk raid set and provision the second raid set in cloud init | 17:22 |
| JayF | dtantsur: I dealt with your concerns on https://review.opendev.org/c/openstack/ironic/+/924887 if you want to revisit; TheJulia also had a +2 on the previous version. | 17:24 |
| JayF | We've landed the Nova side of this so I'd be keen on getting the feature fully landed | 17:24 |
| TheJulia | I'm pretty sure we documented the weird edges on the software raid stuffs, but maybe we didn't?! | 17:25 |
| dtantsur | On my queue for tomorrow | 17:25 |
| JayF | ack, ty | 17:25 |
| opendevreview | Merged openstack/ironic-tempest-plugin master: A simple standalone test for in-band inspection https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/925981 | 17:49 |
| opendevreview | Merged openstack/ironic master: Enable the inspection test in the redfish standalone job https://review.opendev.org/c/openstack/ironic/+/926647 | 17:49 |
| opendevreview | Julia Kreger proposed openstack/ironic-python-agent master: WIP: Provide operator an idea why the deploy failed https://review.opendev.org/c/openstack/ironic-python-agent/+/927823 | 17:56 |
| TheJulia | dtantsur: ^ for you to take a look at tomorrow | 17:56 |
| JayF | I like that change | 17:59 |
| JayF | will poke it with a sharp stick after the TC meeting | 17:59 |
| JayF | I suspect I can create some extra cases for us to catch :D | 18:00 |
| TheJulia | I'm just sitting here trying to figure out how to make it easier to deal with those pesky 4k devices | 18:01 |
| TheJulia | since apparently in my customer's case, there is just no way to reset the raid controller to anything else | 18:01 |
| JayF | someone go tell the hdd manufacturers to put it in metadata someplace standard :/ | 18:01 |
| TheJulia | which.. sucks | 18:01 |
| TheJulia | eh, in this case, go back in time | 18:01 |
| TheJulia | and *fix* it so we start GPT at byte 512 | 18:02 |
| TheJulia | *not* block 1 | 18:02 |
| JayF | I mean, at a minimum, documentation around which controllers behave that way would be a start at all | 18:02 |
| JayF | I'm assuming that doesn't work on !4k | 18:02 |
| JayF | the 512 byte trick, I mean | 18:02 |
| TheJulia | yeah, no | 18:03 |
| TheJulia | and by fix it | 18:03 |
| TheJulia | I mean, fix the standard | 18:03 |
| JayF | yes | 18:03 |
| TheJulia | so, go back and time and whack people upside the head with clue-by-fours from the future | 18:03 |
| JayF | I do have a bit of a platform | 18:03 |
| JayF | could make a video about it or something | 18:03 |
| TheJulia | *this* *will* *cause* *pain* | 18:03 |
| JayF | "GR-OSS moment in hardware" | 18:04 |
| * TheJulia takes a short break | 18:05 | |
| JayF | https://news.ycombinator.com/item?id=41431244 I think most folks here would find this interesting | 18:33 |
| JayF | tl;dr HN discovered IPMI exists | 18:33 |
| opendevreview | cid proposed openstack/ironic master: Upgrade pysnmp-lextudio https://review.opendev.org/c/openstack/ironic/+/927804 | 19:18 |
| mnaser | i'm trying to find docs for spine-leaf (l2 per rack) with ironic + ovn | 19:38 |
| mnaser | i cant find much, is it even possible, my concern here is how to dhcp would work | 19:39 |
| JayF | I assume you've seen the big Ironic OVN doc? | 19:39 |
| JayF | https://docs.openstack.org/ironic/latest/admin/ovn-networking.html | 19:40 |
| mnaser | yeah, i was on that page, i can successfully use it over a l2 network | 19:40 |
| mnaser | leads to https://docs.openstack.org/neutron/latest/admin/ovn/baremetal.html | 19:41 |
| JayF | Let me put it this way: all I know about OVN is in that document, which Julia wrote :D | 19:41 |
| mnaser | lool | 19:41 |
| JayF | so I just wanted to make sure you had seen it, but I can't be helpful beyond that | 19:41 |
| JayF | Maybe she'll come around and have an idea? | 19:42 |
| JayF | but I suspect you may be trailblazing | 19:42 |
| mnaser | now im wondering if i cant use ovn if i am doing relay dhcp and might do neutron-dhcp-agent | 19:42 |
| JayF | I think there was some bug around OVN and baremetal dhcp | 19:42 |
| JayF | but I don't remember what it was | 19:42 |
| JayF | > As of the 2023.2 Release of Ironic, The Ironic project can confirm that OVN’s DHCP server does work for PXE and iPXE operations when using IPv4, OVS version 3.11, and OVN version 23.06.0. | 19:42 |
| JayF | so I guess i t kinda works at least | 19:42 |
| JayF | not with v6 though, I think that might have been the limitation I'm remembering | 19:42 |
| mnaser | yeah no i can confirm it works fine directly with ovn in the same l2 network | 19:45 |
| mnaser | but unsure on how we can make it work over another network | 19:45 |
| mnaser | i am wondering about using the `dnsmasq` driver though | 19:46 |
| mnaser | ftr i think this is more of a neutron issue than ironic for what its worth | 19:51 |
| TheJulia | Possibly semi-trail blazing | 20:00 |
| TheJulia | IPv6, I believe landed in ovn later, I think the question is if it supports dhcp relays at all | 20:05 |
| TheJulia | hjensas might know | 20:05 |
| *** ubuntu is now known as Guest2399 | 20:18 | |
| Guest2399 | JayF: that seems like a resonable wokradound, tho the resulting configuration will be much less clear ;-( | 20:25 |
| JayF | Guest2399: you lost your name, friend :) | 20:26 |
| JayF | Guest2399: I always took the approach of "OpenStack provides raw material of an OS and a root device" and then any config on top was done after the fact... mainly because the array of things people want to do with "storage disks" is MUCH larger than the array of things people want to do with "root disks" | 20:26 |
| *** Guest2399 is now known as groundnuty | 20:30 | |
| JayF | dtantsur: cardoe: I'm punting on enum-ifying stuff in Ironic. Given behavior changes between e.g. BootDevices.LOL and boot_device.LOL and BootDevices.LOL.value ... I just think it'll be a lot of churn for dubious value ... but it's also possible I'm overthinking this :) | 20:31 |
| cardoe | Yeah punt away. | 20:32 |
| JayF | I was working on a refactoring change to do it, on top of my lessee change, just a lot of code churn to make it look actively worse | 20:32 |
| cardoe | So some of the differences are cause you might want to use StrEnum instead of regular Enum. | 20:33 |
| cardoe | I know Sean brings up a point about frozen enums but a lot of that is mitigated via static checkers like mypy and even flake8. | 20:34 |
| JayF | well, freezing it is trivial | 20:36 |
| JayF | I have that solved | 20:36 |
| JayF | but StrEnum sounds interesting | 20:36 |
| JayF | or even RefrEnum() | 20:37 |
| JayF | https://docs.python.org/3/library/enum.html#enum.ReprEnum | 20:37 |
| JayF | thank you for the pointer! | 20:37 |
| * cid 's EOD. o/ | 20:41 | |
| JayF | \o | 20:42 |
| JayF | cardoe: This seems to be happier for our purposes. For the few times we need the list of items in an enum, we can call Blah.list() https://www.irccloud.com/pastebin/LDlP3VmS/ | 21:06 |
| JayF | I know I could reduce that down into one class def, but I think the way it's written makes it more clear what's happening | 21:07 |
| TheJulia | Anyone seen anything like this where Tianocore says "nope" | 21:09 |
| TheJulia | https://usercontent.irccloud-cdn.com/file/VxBBU0Wb/centos9-not-secureboot.png | 21:09 |
| JayF | That's a new one on me | 21:09 |
| JayF | might wanna ask in -nova? | 21:10 |
| JayF | They might have a better idea but idk | 21:10 |
| TheJulia | Oh, I think i grok it now | 21:13 |
| JayF | ? | 21:18 |
| groundnuty | JayF: I was hoping to find the world of infrastructure automation ready for declarative gitops where you can setup your server with single yamls :-) | 21:51 |
| groundnuty | JayF: as for cloud-init it does not support mdadm natively https://github.com/canonical/cloud-init/blob/5a74d19d066e932c6d8925c858acf92b1406654b/doc/examples/cloud-config-disk-setup.txt#L121 | 21:51 |
| JayF | groundnuty: I just saw a great clip from my next podcast episode, where Dejan (a guy who works on K8s) gave this great quote: Devops is just a big box of yaml :D | 21:51 |
| groundnuty | so I guess to do anything one would have to write custom scripts/one-liners ;-( | 21:51 |
| JayF | groundnuty: When doing this in the past, I used bash | 21:51 |
| JayF | https://github.com/jayofdoom/onmetal-fast-docker | 21:52 |
| JayF | The timestamps on that repo will tell you how dusty my knowledge in this specific space is LOL | 21:52 |
| JayF | groundnuty: seriously though, it's just a matter of where the lines are drawn: OpenStack is more about getting you a platform to put things on, and then using config management, image management, etc for config | 21:53 |
| JayF | In other solutions, they draw the lines in different places | 21:53 |
| JayF | For instance, my downstream uses Ironic via Nova via Terraform :) | 21:53 |
| JayF | so their users get the "write yaml -> recieve infra" you just need another layer (also rip Terraform, long live opentf) | 21:53 |
| groundnuty | JayF: in my case baremetal-operator will be the yaml layer, embeding bash in yaml while doable is not perfered :-) | 21:55 |
| JayF | Ahh. I wonder if dtantsur (author of said operator) would have any ideas that might be more k8s-flavored :) | 21:55 |
| groundnuty | I looked at the pull request that introduced raid - it's very bare bones and supports small portion of ironic features ;-( | 21:56 |
| groundnuty | in that baremetal-operator reapo I mean | 21:56 |
| JayF | yeah, the operator is extremely new | 22:06 |
| JayF | Dmitry started on it a few months back iirc | 22:06 |
| JayF | I'm sure he'd love patches if you can find a way to map in more of the interface | 22:06 |
| opendevreview | Julia Kreger proposed openstack/ironic-python-agent master: WIP: Provide operator an idea why the deploy failed https://review.opendev.org/c/openstack/ironic-python-agent/+/927823 | 22:07 |
| cardoe | That’s a good solution JayF. | 22:33 |
| JayF | if you get one second | 22:34 |
| JayF | I'm almost done with boot_devices and boot_modes | 22:34 |
| JayF | was going to stop there and push a WIP/Prototype | 22:34 |
| cardoe | See and I’m using GitOps to ensure we have an Ironic, Neutron, Nova that is in a good state. | 22:34 |
| JayF | before mechanical-turking the rest of it | 22:34 |
| JayF | Devops is a big box of yaml. | 22:34 |
| cardoe | And 20 years ago life was a big box of XML. | 22:36 |
| JayF | no, I lived in the jsoniverse | 22:41 |
| JayF | the great schism | 22:41 |
| mnaser | i found a thing | 22:44 |
| mnaser | https://bugs.launchpad.net/neutron/+bug/1692486 | 22:44 |
| mnaser | this could be very handy for ironic folks :) | 22:44 |
| opendevreview | Jay Faulkner proposed openstack/ironic master: WIP/PoC: Move constants to enum, constants module https://review.opendev.org/c/openstack/ironic/+/927879 | 22:45 |
| JayF | cardoe: dtantsur: ^ let me know how you all feel about this direction | 22:45 |
| cardoe | mnaser: agreed. Almost wanna make a k8s friendly dhcp server. I’m the one that suggested the OSH operator that Vlad pinged you about btw. | 22:51 |
| mnaser | cardoe: oh nice! yeah, well ironic already has a dnsmasq dhcp driver fwiw :P | 22:52 |
| cardoe | Yeah but dnsmasq is well… dnsmasq. | 22:53 |
| mnaser | haha.. yep | 22:53 |
| mnaser | cardoe: i have started and decided not to write an OSH operator about 15 times now because ENOTIME | 22:54 |
| cardoe | That’s how I feel. | 22:54 |
| mnaser | i did end up with https://github.com/vexxhost/openstack-service-operator and learned about https://github.com/k-orc/openstack-resource-controller after | 22:54 |
| mnaser | to get rid of those pesky jobs at least :) | 22:56 |
| cardoe | I saw yours. That’s roughly what I wanted to do. Just get rid of the bootstrap scripts basically. | 22:56 |
| cardoe | Yeah bootstrap and jobs. | 22:56 |
| mnaser | funny how same frustrations develop over time aross operators :P | 22:56 |
| cardoe | Absolutely | 22:57 |
| cardoe | When I get some cycles I’m going to make the secrets in OSH work by ref. | 22:58 |
| mnaser | yeah fitbeard did do some cool work around that with the idea of potentially using vault and dynamic secrets, that would be cool as heck | 22:59 |
| cardoe | I wish keystone wasn’t so painful with OIDC as well. | 23:05 |
| TheJulia | mnaser: Ahh yes, except that was dhcp-agent, not ovn | 23:09 |
| TheJulia | (referring to the launchpad link earlier) | 23:09 |
| mnaser | cardoe: we've mostly solved that with keycloak native + keystoneauth-websso | 23:11 |
| mnaser | TheJulia: yeah, i dont think there's actually a way with non-ovn because there's no actual ip to send that traffic too | 23:11 |
| TheJulia | That is... saddening | 23:11 |
| opendevreview | Merged openstack/ironic-python-agent master: Fix series in release notes https://review.opendev.org/c/openstack/ironic-python-agent/+/926992 | 23:18 |
| opendevreview | Merged openstack/sushy master: Fix versions in release notes https://review.opendev.org/c/openstack/sushy/+/926995 | 23:24 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
