| TheJulia | JayF: well, if it is tracked, such a way, then we have to actively edit the repo for such changes. *shrug* dunno | 00:14 |
|---|---|---|
| opendevreview | Vanou Ishii proposed openstack/ironic master: Change boot_interface order of iRMC driver https://review.opendev.org/c/openstack/ironic/+/862616 | 03:31 |
| opendevreview | Vanou Ishii proposed openstack/ironic master: Change boot_interface order of iRMC driver https://review.opendev.org/c/openstack/ironic/+/862616 | 05:38 |
| opendevreview | Vanou Ishii proposed openstack/ironic master: Align iRMC driver with Ironic's default boot_mode https://review.opendev.org/c/openstack/ironic/+/862552 | 06:17 |
| opendevreview | Vanou Ishii proposed openstack/ironic master: Change boot_interface order of iRMC driver https://review.opendev.org/c/openstack/ironic/+/862616 | 06:18 |
| arozman | Hello Ironic! | 09:12 |
| janders | good morning Ironic o/ | 10:00 |
| iurygregory | good morning Ironic | 11:28 |
| opendevreview | Merged openstack/sushy master: Improve resiliency of eTag handling https://review.opendev.org/c/openstack/sushy/+/856123 | 11:58 |
| mraineri | dtantsur: Sorry, was out last week and just noticed your ping; Redfish does support specifying an HTTP boot. You do this by setting "BootSourceOverrideTarget" to "UefiHttp" and setting "HttpBootUri" to the desired URI for booting. | 12:50 |
| mraineri | It's been in the standard since the 2019.3 bundle | 12:50 |
| dtantsur | mraineri: nice! do you know if it's commonly supported among vendors already? | 12:51 |
| mraineri | I would hope so at this point... | 12:51 |
| mraineri | Let me at least check some systems I have... | 12:51 |
| mraineri | Well that's distressing... The first system I poked shows UefiHttp in the allowable values... But it doesn't have the HttpBootUri property, which makes it useless... | 12:53 |
| dtantsur | :( | 12:54 |
| mraineri | Wow... This second system from another vendor is behaving the same... | 12:55 |
| mraineri | Third one doesn't support it at all | 12:55 |
| dtantsur | double sadness | 12:55 |
| dtantsur | this is bad news, but thanks for checking mraineri! | 12:56 |
| mraineri | Yeah... the others don't support it... | 12:56 |
| mraineri | At the very least I need to add a standard test to ensure services that say "UefiHttp" is allowable need to also support the HttpBootUri property... | 12:57 |
| mraineri | Sigh... You would think this is common sense... | 12:57 |
| dtantsur | mraineri: well, I guess, they could also provide the URI via DHCP? | 12:57 |
| dtantsur | maybe they imply that? | 12:57 |
| mraineri | I don't think so... The intent is the two play together | 12:58 |
| mraineri | They were added at the same time and have ties to each other | 12:58 |
| dtantsur | fair enough | 12:58 |
| mraineri | Actually... No they weren't... | 12:58 |
| mraineri | Maybe you're right... | 12:58 |
| mraineri | "UefiHttp" was added in 1.1 of ComputerSystem, and HttpBootUri was added later in 1.9 | 12:59 |
| mraineri | The description of HttpBootUri makes it sound like it's essential to use though: "The URI to boot from when BootSourceOverrideTarget is set to `UefiHttp`." | 13:00 |
| mraineri | We don't have any description text that says what to do when the property isn't set | 13:00 |
| mraineri | I'll ask around... | 13:00 |
| TheJulia | good morning | 13:00 |
| dtantsur | mraineri: thanks! | 13:11 |
| dtantsur | morning TheJulia | 13:11 |
| hjensas | good morning | 13:38 |
| *** knikolla[m] is now known as knikolla | 14:38 | |
| TheJulia | I feel like I need a very big cup of coffee | 15:09 |
| JayF | FYI: There are EPEL mirror issues in infra | 15:09 |
| JayF | I suspect that our CI depends on that; so just keep an eye out if you see weird failures on relevent distros | 15:09 |
| dtantsur | bifrost probably | 15:10 |
| TheJulia | JayF: commented on the shard key spec comments. I know the cycle time has not been as *much* of an issue, is tracking of nodes disappearing/reappearing elsewhere with hash ring changes and the disappearance/reappearance of nodes from the cache in the compute service | 15:17 |
| JayF | I think those responses mostly make sense; it wasn't clear to me we had agreement that a "move stuff" tool would be in the v1 | 15:18 |
| TheJulia | aiui, the node *can* move, if the appropriate fields are updated in nova, which makes it a nova-manage. If we just treated it as a rbac policy restricted field, with lots of appropriate documentation, then I think we should be perfectly fine | 15:20 |
| TheJulia | Ironic won't care at all in this model, its for the consumer of the API more than anything else | 15:21 |
| TheJulia | it will likely just say "no, your not a service, or the system, or an admin, nope, you cannot change it" | 15:21 |
| * TheJulia suddenly wants gifs with API errors | 15:22 | |
| JayF | I'm talking about the *nova side* tool to move stuff around | 15:22 |
| TheJulia | it will be needed, in my opinion, for *any* migration | 15:22 |
| JayF | I guess either way I don't have a better option even for the MVP than your suggestion | 15:22 |
| JayF | oh; that's true | 15:22 |
| JayF | going from existing->new will require ~the exact same tool | 15:23 |
| TheJulia | mainly because instance/computenode records will need to be "fixed" | 15:23 |
| TheJulia | and *having* that tooling there allows it to "fix" any outstanding issues as well, which is another operator complaint | 15:23 |
| JayF | aight, cool, thanks :) | 15:24 |
| TheJulia | But I think the overall simplification and removing the hashring from the mix should at least make a lot of the nova-compute behavior self reconcilable as well | 15:25 |
| JayF | TheJulia: johnthetubaguy and I were going to have a spec writing session on the Nova side, probably Weds (in 6 days). You wanna be in on it? | 15:26 |
| TheJulia | sure | 15:27 |
| JayF | ack; I'll make sure you're on the invite | 15:28 |
| TheJulia | cool, thanks | 15:29 |
| TheJulia | The main thing I want to avoid is nova-compute trying to take on a responsibility of doing the initial shard key setting since it would involve it having to still go through the painful download everything path | 15:30 |
| JayF | Oh; I never even considered that as a possible path. | 15:31 |
| JayF | It's not nova's business and we're going to have more consumers of that data than just nova | 15:31 |
| JayF | In some of my previous jobs, we probably would've used shard key to help with performing bulk, scripted operations in parallel | 15:32 |
| TheJulia | Yeah, and the networking-baremetal case makes a lot of sense to just handle as well | 15:33 |
| JayF | In fact, I could see us documenting the ideal size of a shard as "the amount of nodes that can be read and processed from the API by a python client in [time]" | 15:34 |
| JayF | which makes it not-nova-centric for standalone users, too, who might want some sharding | 15:34 |
| TheJulia | eh, that would be a really small number and the nova-compute translation could quickly be an hour with >500-1000 nodes. | 15:35 |
| TheJulia | they were saying 100-200 per nova-compute on the ptg call | 15:36 |
| JayF | Yeah, that's what I was thinking too | 15:36 |
| TheJulia | multi-process could make that much more sane | 15:36 |
| JayF | per nova-compute *service* or *process* | 15:36 |
| TheJulia | service == process currently | 15:36 |
| JayF | Most places I've worked ran N nova-computes for N cores on a single server | 15:36 |
| JayF | via devops style configuration | 15:36 |
| JayF | (you just need to set CONF.host differently for each) | 15:36 |
| TheJulia | yup | 15:37 |
| kubajj | dtantsur: If I just execute the test setup script, does it create the database? | 16:20 |
| TheJulia | Do we have any ironic-cores that can take a look at https://review.opendev.org/c/openstack/ironic/+/862506 ? Dmitry and I have both worked on it, and I, Steve both have a +2 on it. | 17:04 |
| JayF | lookin | 17:05 |
| JayF | oh, I didn't know that got up, good stuff | 17:05 |
| JayF | does that fix the sqlite ? | 17:05 |
| JayF | TheJulia: I think we wrote a method at $job[-3] that would strip the sqlalchemy special attrs off the model objects | 17:10 |
| JayF | TheJulia: to completely punt on that issue where the object had to be garbage collected for the txns to complete | 17:11 |
| TheJulia | our db object to object model method only copies fields in the object model | 17:11 |
| TheJulia | so the db result handlers do get dropped on the floor of the object model | 17:11 |
| JayF | this was much, much older sqla though so semantics may have changed | 17:11 |
| TheJulia | hjensas: great job on https://review.opendev.org/c/openstack/ironic/+/862292! | 17:19 |
| TheJulia | hjensas: any chance I can get you to eliminate https://github.com/openstack/ironic/blob/master/ironic/api/controllers/v1/port.py#L128 with a join ? :) Same for portgroup :) | 17:22 |
| JayF | TheJulia: that patch is great; and it reminds me of another fix we did at that same job which helped a TON with deadlocking in our use of sqlite | 17:27 |
| JayF | TheJulia: it's like, the more magic you all remove, the more I can understand what is actually happening \o/ | 17:27 |
| TheJulia | Oh yes. Very yes. :) | 17:28 |
| TheJulia | I remember a PTG once where I basically said magic is bad. | 17:28 |
| JayF | I don't think our usage pattern makes sense for actually doing this; but this is also the makings of being able to split write (primary) and r/o (replica) clusters | 17:29 |
| TheJulia | so technically, I think enginefacade will allow it if the reader is allowed to be an async reader | 17:35 |
| TheJulia | at least, out of the box. I've not digged at that | 17:35 |
| opendevreview | Julia Kreger proposed openstack/ironic master: Replace more instances of model_query https://review.opendev.org/c/openstack/ironic/+/862506 | 17:36 |
| opendevreview | Julia Kreger proposed openstack/ironic master: Remove autocommit, again. https://review.opendev.org/c/openstack/ironic/+/862832 | 17:36 |
| TheJulia | rebase ^ | 17:36 |
| JayF | TheJulia: did we add a job to validate sqlite is working before re-enabling autocommit? | 17:37 |
| JayF | I sorta feel like we should, but don't have a good concept of how difficult that might be | 17:37 |
| TheJulia | we do not have one, afaik. I think the only way we could do it is using bifrost | 17:37 |
| JayF | I mean, that would capture the obviously-broken cases we've seen so far, right? | 17:39 |
| TheJulia | most likely | 17:42 |
| TheJulia | dtantsur: thoughts on a bifrost job? | 17:43 |
| TheJulia | ebbex: wow, you've been busy on bifrost | 18:01 |
| TheJulia | stevebaker[m]: rebase needed on https://review.opendev.org/c/openstack/bifrost/+/851682 :) | 18:06 |
| opendevreview | Julia Kreger proposed openstack/bifrost master: Move kpartx to dib_host_required_packages https://review.opendev.org/c/openstack/bifrost/+/861539 | 18:08 |
| opendevreview | Julia Kreger proposed openstack/ironic-inspector master: Remove unicode literal from code https://review.opendev.org/c/openstack/ironic-inspector/+/844709 | 18:31 |
| opendevreview | Merged openstack/ironic-inspector master: Remove unicode literal from code https://review.opendev.org/c/openstack/ironic-inspector/+/844709 | 18:53 |
| opendevreview | Merged openstack/bifrost stable/yoga: Do not install grub2 and shim on the host system https://review.opendev.org/c/openstack/bifrost/+/858391 | 19:25 |
| opendevreview | Steve Baker proposed openstack/bifrost master: Switch bifrost to use the dnsmasq DHCP driver https://review.opendev.org/c/openstack/bifrost/+/851682 | 19:39 |
| stevebaker[m] | TheJulia: Done! | 19:40 |
| opendevreview | Steve Baker proposed openstack/bifrost master: Copy shim and grub into tftp and http directories https://review.opendev.org/c/openstack/bifrost/+/849247 | 19:51 |
| opendevreview | Steve Baker proposed openstack/bifrost master: Remove enable_uefi_ipxe https://review.opendev.org/c/openstack/bifrost/+/849248 | 19:51 |
| opendevreview | Steve Baker proposed openstack/bifrost master: Support PXE network boot with grub https://review.opendev.org/c/openstack/bifrost/+/807220 | 19:51 |
| opendevreview | Steve Baker proposed openstack/ironic-python-agent stable/train: DNM test previous change https://review.opendev.org/c/openstack/ironic-python-agent/+/862840 | 20:02 |
| opendevreview | Merged openstack/ironic master: Replace more instances of model_query https://review.opendev.org/c/openstack/ironic/+/862506 | 20:22 |
| hjensas | TheJulia: regarding eliminating populate_node_uuid by using a join. | 20:34 |
| hjensas | TheJulia: not quite sure how to do that, would we add a 'node_uuid' field to object.Port, and then in sqlalchemy api do some magic joining data from Port and Node from the db? | 20:37 |
| hjensas | TheJulia: It looks like chassi_uuid on nodes have the same ineffective pattern. | 20:40 |
| ashinclouds[m] | Yeah, basically that is what we would need to do, fill the field using an this as that name thing. I’ve done it in sql, but not sqlalchemy yet | 20:42 |
| JayF | Is there anyone around who understands this "Cleanup client RAID" work stream for this cycle better than me? | 20:42 |
| JayF | I'm trying to write it up for the workstreams doc and I am not doing well :) | 20:43 |
| TheJulia | JayF: I can look in like 1.5 hours. Just sat down for lunch and groceries | 20:46 |
| hjensas | ok, I'll go read sqlalchemy docs and try to figure it out. | 20:47 |
| JayF | TheJulia: ack; I'm going to basically get the rest of the doc done, feel free to put a suggested blurb in the comments or just push a revision there | 20:47 |
| opendevreview | Jay Faulkner proposed openstack/ironic-specs master: Add Ironic work items for 2023.1 https://review.opendev.org/c/openstack/ironic-specs/+/862538 | 20:53 |
| opendevreview | Jay Faulkner proposed openstack/ironic-specs master: Add Ironic work items for 2023.1 https://review.opendev.org/c/openstack/ironic-specs/+/862538 | 20:55 |
| JayF | I am doing a major revision to the whiteboard right now, please ignore it as usual for a bit lol | 21:03 |
| JayF | RFR the whiteboard updates I just did | 21:22 |
| opendevreview | Jay Faulkner proposed openstack/ironic-specs master: Add Ironic work items for 2023.1 https://review.opendev.org/c/openstack/ironic-specs/+/862538 | 21:30 |
| janders | good morning Ironic o/ | 23:24 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!