| opendevreview | Jacob Anders proposed openstack/sushy-tools master: Replace hardcoded BiosVersion with an updatable field https://review.opendev.org/c/openstack/sushy-tools/+/909487 | 01:17 |
|---|---|---|
| opendevreview | Jacob Anders proposed openstack/sushy-tools master: [WIP] Add support for BIOS update emulation https://review.opendev.org/c/openstack/sushy-tools/+/909500 | 06:14 |
| opendevreview | Merged openstack/ironic-ui master: Update python classifier in setup.cfg https://review.opendev.org/c/openstack/ironic-ui/+/905815 | 06:20 |
| dtantsur | TheJulia, hjensas, JayF, FYI https://review.opendev.org/c/x/ironic-staging-drivers/+/909506 | 07:58 |
| rpittau | good morning ironic! o/ | 08:23 |
| opendevreview | Merged openstack/bifrost master: Wire in node auto-discovery https://review.opendev.org/c/openstack/bifrost/+/906635 | 10:38 |
| *** dmellado74522 is now known as dmellado7452 | 11:06 | |
| opendevreview | Sven Kieske proposed openstack/networking-baremetal master: don't force amqp_auto_delete for quorum queues https://review.opendev.org/c/openstack/networking-baremetal/+/903995 | 11:45 |
| *** dmellado74522 is now known as dmellado7452 | 12:00 | |
| frickler | so it looks like lextudio admitted their error. of not doing a major version bump, which they have done now. don't think that helps us much | 12:55 |
| frickler | to lighten things up, I also found testing failures when upgrading jsonschema, if someone might be looking for a different task https://review.opendev.org/c/openstack/requirements/+/909452 | 12:57 |
| rpittau | frickler: there were major changes in jsonschema between 4.17 and 4.18 after the change to RPDS, I'm actually surprised to see 4.19 in upper-constraints, I wonder if the changes are hitting now but the errors were masked before for some reason | 13:06 |
| rpittau | anyway, I'll try to have a look today and see if hopefully it's something trivial | 13:06 |
| TheJulia | good morning | 13:59 |
| TheJulia | dtantsur: reality | 14:00 |
| frickler | rpittau: well the update is not yet committed into reqs exactly because of the test failures, what I linked to is the patch that would bump it, as the bot tries to do this every week | 14:03 |
| rpittau | frickler: yeah, I know :) | 14:04 |
| rpittau | I'm just surprised to see that 4.19 worked well so far :) | 14:04 |
| TheJulia | Regarding lecture-piano, 3.0.35 and above is getting yanked. | 14:04 |
| TheJulia | Err autocorrect | 14:04 |
| rpittau | good | 14:04 |
| TheJulia | Pysnmp | 14:04 |
| frickler | lol, that's a nice correctexpansion for lextudio :) | 14:07 |
| rpittau | mmm interesting looks like they changed schema for some errors :D | 14:27 |
| rpittau | frickler: for the jsonschema change that is a breaking change and need to find a way to handle old and new error messages | 14:40 |
| rpittau | or exclude the failing unit tests until the new jsonschema version gets in :/ | 14:45 |
| dtantsur | Gertty seems completely broken with Python 3.12 :( | 15:11 |
| dtantsur | or with new sqlalchemy. or whatever. | 15:12 |
| opendevreview | Riccardo Pittau proposed openstack/ironic master: Handle jsonschema empty error message update https://review.opendev.org/c/openstack/ironic/+/909592 | 15:25 |
| rpittau | frickler: would it be possible to retest the jsonschema update with this ^ ? | 15:26 |
| opendevreview | Riccardo Pittau proposed openstack/ironic master: Handle jsonschema empty error message update https://review.opendev.org/c/openstack/ironic/+/909592 | 15:28 |
| frickler | rpittau: you can either add a depends-on from your patch to https://review.opendev.org/c/openstack/requirements/+/909452 or the other way around, your choice. or add a dnm patch with the depends-on on top of yours, if you want to see results with old and new in parallel | 15:37 |
| rpittau | I'll change the patch on requirements side as we need to land the ironic one first | 15:38 |
| opendevreview | Riccardo Pittau proposed openstack/ironic master: Handle jsonschema empty error message update https://review.opendev.org/c/openstack/ironic/+/909592 | 15:52 |
| *** dking is now known as Guest342 | 15:57 | |
| *** Guest342 is now known as dking | 15:57 | |
| dtantsur | An interesting failure in one of the bifrost jobs: it seems that ironic and IPA simply stop reaching each other: https://637b02491f0435a9a86b-ccec73fd7dde7a9826f6a9aeb49ab878.ssl.cf5.rackcdn.com/905192/2/check/bifrost-integration-tinyipa-debian-bullseye/64bfd6f/logs/testvm1_console.log | 15:57 |
| opendevreview | Riccardo Pittau proposed openstack/ironic master: Handle jsonschema empty error message update https://review.opendev.org/c/openstack/ironic/+/909592 | 15:59 |
| rpittau | bye everyone, see you tomorrow! o/ | 16:01 |
| TheJulia | dtantsur: impressive?! | 16:41 |
| opendevreview | Verification of a change to openstack/ironic stable/wallaby failed: [iRMC] Fix parse_driver_info bug enforcing SNMP v3 under FIPS mode https://review.opendev.org/c/openstack/ironic/+/885246 | 19:11 |
| dking | When we do root device hints, is there a way to use `dev/disk/by-path/pci-*` entries? | 19:11 |
| dking | I'm still poking around in the code because I know that `list_all_block_devices` does something with them, but when we tried using the path, it looks like it was comparing it to the `/dev/*` name and so rejecting the hints. | 19:16 |
| TheJulia | I think we've never codified that in large part because of randomized bus initialization ordering which is the same base problem, really, just presents differently | 19:29 |
| dking | Am I understanding that you're saying that the feature wasn't added because it wasn't believed that it would actually solve any typical problem? | 19:32 |
| TheJulia | device ordering inside systems is sort of unrealiable | 19:32 |
| dking | So far, it seems that in our system, using Supermicro, so far, the PCI slots seem to boot with the same names at least more consistent than we might expect if a there were a trouble with a disk, for instance. | 19:32 |
| TheJulia | because your dealing with initialization ordering, and by pci path is likely the most as long as nothing else gets added | 19:33 |
| TheJulia | Yeah, the disk names are just an abstraction that applies bus device *and* driver load ordering | 19:33 |
| TheJulia | more than anything, i suspect is is lack of demand, most folks just do serial number if they know the device that well | 19:34 |
| dking | I can see how it wouldn't be a 100% solution, but I think that some people might still like the idea of using PCI path rather than just the device name. | 19:36 |
| TheJulia | yeah, device name is highly unreliable on some hardware | 19:37 |
| dking | I'm thinking that I'm probably going to create a custom deploy step that translates the path to to a device name, but I wanted to check that it wasn't already an option somewhere before I did that. | 19:37 |
| TheJulia | I don't see why we would reject a patch just to add it to the filters | 19:37 |
| dking | I wondered if it would be worth the discussion about whether to allow them. It would be opt-in by its nature since nobody has to use the /by-path form in their hints unless they specifically are wanting to do that, and it's probably better to either do what they ask (perhaps by translating the symlink) or perhaps it should fail with an error to indicate that the format is invalid. | 19:39 |
| dking | I was talking with our storage guy and it made me wonder what other places the device name was being used. I know that it's used a lot during cleaning disks and for the location to install the boot image. But I was wondering if it is used for other things as well, including things like setting up Grub. | 19:41 |
| TheJulia | Well, it wouldn't need to translate it and safe it, just identify it I guess | 19:41 |
| TheJulia | grub setup is based upon where it writes the image | 19:41 |
| TheJulia | so after the selection match occurs | 19:41 |
| dking | That would be the fastest way. It's a separate, but related conversation about whether it might be good to default to by-path names. But that's a lot of work. | 19:42 |
| TheJulia | internally it gets set to a variable in memory for the ramdisk to complete with but the ramdisk also supplies hints back to ironic which it sends for boot setup | 19:42 |
| TheJulia | All of the block devices get consulted hunting for the UUID/PartID values | 19:42 |
| dking | Oh, nice. So, then really nothing that stays on the device after deploy is going to use the device names, anyway, if they're already converting to UUID/PartID, and if it's just during cleaning, then apart from the device hint, any name for a drive would be just as good as long as it's consistent. | 19:44 |
| TheJulia | Yeah, the static device naming is just for human convenience at the end of the day. It helps inform a decision when we write artifacts, but that is about it | 19:47 |
| opendevreview | Merged openstack/bifrost master: Deprecate ironic-inspector support https://review.opendev.org/c/openstack/bifrost/+/905192 | 19:53 |
| dking | Okay. I'm taking a look at ironic_lib.utils.find_devices_by_hints. So, perhaps the team wouldn't mind if there were something in there which would allow devices to be returned if the hint is a by-path symlink which resolves to that device? | 19:55 |
| TheJulia | I don't see why we would, off hand, but I've not glanced at the code recently | 19:56 |
| dking | Okay. Now, I just need to figure out the code. Interestingly, as long a I've been writing Python, I think today is the first time I've seen a for...else. | 19:59 |
| TheJulia | try... except... else.... finally ;) | 19:59 |
| opendevreview | Merged openstack/ironic stable/wallaby: [iRMC] Fix parse_driver_info bug enforcing SNMP v3 under FIPS mode https://review.opendev.org/c/openstack/ironic/+/885246 | 22:39 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!