| rpittau | good morning ironic! o/ | 07:19 |
|---|---|---|
| rpittau | JayF TheJulia dtantsur iurygregory: good news! we've managed to get the ironic-core group in the sushy-oem-idrac-release group, that should give us more control on sushy-oem-idrac, and we're not too pressed about its future | 08:06 |
| kubajj | Good morning rpittau, and ironic o/ | 08:16 |
| opendevreview | Merged openstack/ironic master: Add unit test to ensure full node object is not logged https://review.opendev.org/c/openstack/ironic/+/926487 | 09:20 |
| opendevreview | cid proposed openstack/ironic master: [WIP] Add inspection rules https://review.opendev.org/c/openstack/ironic/+/918303 | 09:42 |
| opendevreview | cid proposed openstack/ironic master: [WIP] Add inspection rules https://review.opendev.org/c/openstack/ironic/+/918303 | 10:30 |
| iurygregory | rpittau, awesome! | 10:36 |
| iurygregory | good morning ironic | 10:36 |
| dtantsur | rpittau: \o/ (re sushy-oem-idrac) | 11:15 |
| opendevreview | cid proposed openstack/ironic master: [WIP] Add inspection rules https://review.opendev.org/c/openstack/ironic/+/918303 | 11:28 |
| * cid apologies for the noice | 11:28 | |
| opendevreview | Riccardo Pittau proposed openstack/ironic stable/2024.1: [CI][stable only] fix zuul config https://review.opendev.org/c/openstack/ironic/+/926636 | 13:07 |
| opendevreview | Riccardo Pittau proposed openstack/ironic stable/2023.2: [CI][stable only] fix zuul config https://review.opendev.org/c/openstack/ironic/+/926637 | 13:08 |
| opendevreview | Riccardo Pittau proposed openstack/ironic stable/2023.1: [CI][stable only] fix zuul config https://review.opendev.org/c/openstack/ironic/+/926638 | 13:08 |
| TheJulia | good morning | 13:23 |
| jamesdenton_alt | Good morning! Sorry to hit you with this so early, but question about virtual media-based deployments: is there any special element that needs to be included in the IPA ramdisk to support configdrive to allow for dhcp-less network configuration during deployment? Is simple-init enough? | 13:24 |
| *** jamesdenton is now known as Guest972 | 13:25 | |
| *** jamesdenton_alt is now known as jamesdenton | 13:25 | |
| dtantsur | https://docs.openstack.org/ironic/latest/admin/dhcp-less.html still suggests though, I'm not sure how much it has changed due to TheJulia's efforts | 13:26 |
| TheJulia | simple-init should be enough | 13:27 |
| jamesdenton | I've been following that one using a 14th gen Dell. Just wanted to make sure i wasn't missing something before going too far down the hole. | 13:27 |
| jamesdenton | I'm using 'neutron' as the network_interface, so i assume it will generate the needful? | 13:28 |
| TheJulia | Trying to recall, I think cloud-init could also be an option but I think by default it explicitly blocks that. I just don't remember anymore | 13:28 |
| dtantsur | IIRC the network interface is orthogonal to dhcp-less | 13:28 |
| TheJulia | yeah, it is, i can work with, or without it | 13:28 |
| TheJulia | s/i/it/ | 13:29 |
| jamesdenton | k | 13:29 |
| * TheJulia gets the corgi questioning "why are we not walking yet!?" | 13:30 | |
| dtantsur | A fair question on the corgi's side | 13:39 |
| dtantsur | My question on https://review.opendev.org/c/openstack/ironic/+/925129 is probably worth bringing to a wider audience than cid and JayF | 13:42 |
| dtantsur | namely: should we go against the OpenStack-wide guidance and start using upper-case "Ironic" to refer to our project? | 13:42 |
| dtantsur | to me, it has always been the most reasonable option, but we've been trying to stick with how OpenStack does it for a long time | 13:43 |
| dtantsur | if we change our course, I suggest we start doing it consistently | 13:43 |
| dtantsur | same situation with https://review.opendev.org/c/openstack/ironic/+/925691 | 13:43 |
| dtantsur | cc rpittau | 13:43 |
| rpittau | well at least so far we've been consistently talking about that :D | 13:44 |
| rpittau | jokes aside, I think we should go for upper-case "Ironic" | 13:44 |
| dtantsur | heh | 13:44 |
| dtantsur | cid: meanwhile, I've noticed in a couple of your patches: https://review.opendev.org/c/openstack/ironic/+/925691/2/doc/source/configuration/index.rst changes to "reference", but the previous file keeps Upper Case. Why is that? | 13:44 |
| TheJulia | I think we're making a proper name reference. The OpenStack, original guidance from the long *gone* docs folks was always "Bare Metal Service" which I think was wrong to begin with | 13:45 |
| rpittau | I think "ironic" should refer to the services/libraries, and "Ironic" to the entire project | 13:46 |
| dtantsur | rpittau: I like this approach | 13:46 |
| dtantsur | Another angle to look at it: literally nobody outside of our small bubble is using "the Bare Metal service" and very few opt for lower case "ironic" (mostly accidentally) | 13:47 |
| rpittau | right | 13:47 |
| iurygregory | ++ to go with upper-case Ironic =) | 13:55 |
| cid | dtantsur: it was mostly an attempt to follow the audit recommendation as closely, than anything else. | 14:00 |
| dtantsur | cid: sure, but we need to stay consistent at least within the changes we're making | 14:01 |
| dtantsur | what was the recommendation, only upper-case the first letter? | 14:02 |
| cid | Yeah! By the way, it's more of an example than a recommendation. So, whatever the larger community thinks is probably best. | 14:03 |
| cid | Should I update the patch? | 14:03 |
| opendevreview | Dmitry Tantsur proposed openstack/ironic-tempest-plugin master: A simple standalone test for in-band inspection https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/925981 | 14:06 |
| opendevreview | Dmitry Tantsur proposed openstack/ironic master: Enable the inspection test in the redfish standalone job https://review.opendev.org/c/openstack/ironic/+/926647 | 14:08 |
| opendevreview | Riccardo Pittau proposed openstack/ironic master: [docs] Add procedure to EoL and delete old bugfix branch https://review.opendev.org/c/openstack/ironic/+/926648 | 14:16 |
| dtantsur | cid: let's ask Reverbverbverb first? I have no clue what is considered standard for titles in English literature | 14:16 |
| dtantsur | (or any literature, I'm universally clueless!) | 14:17 |
| opendevreview | Riccardo Pittau proposed openstack/ironic master: [docs] Add procedure to EoL and delete old bugfix branch https://review.opendev.org/c/openstack/ironic/+/926648 | 14:17 |
| TheJulia | dtantsur: Yeah, lets stamp out "the Bare Metal service" in our docs. | 14:45 |
| * TheJulia feels the need to have a safety station with "more fire" and "less fire" tools | 14:47 | |
| dtantsur | more fire \o/ | 14:47 |
| opendevreview | Verification of a change to openstack/ironic master failed: Update configuration value in iRMC https://review.opendev.org/c/openstack/ironic/+/924349 | 14:49 |
| JayF | I am probably going to land somewhere more like "use ironic but the first time add (Bare Metal Service)" or something like that | 14:56 |
| JayF | just so people searching for the openstack-compliant term will be able to find it | 14:56 |
| dtantsur | do these people exist? | 14:57 |
| TheJulia | They do | 14:58 |
| dtantsur | JayF: a large part of this conversation is to do what basically everyone is doing: just use "Ironic" | 14:58 |
| dtantsur | wow, really? | 14:58 |
| JayF | there are basically two sets of ironic users: | 14:58 |
| TheJulia | Yeah, Turns out multitenant baremetal is actually a thing people need to do | 14:58 |
| JayF | people who use it as a cornerstone of their deployment (metal3 users, bare-metal-first users) | 14:58 |
| JayF | and people who use it was a plugin to their openstack cloud to add some BM capabilities | 14:58 |
| JayF | the way those personas think about Ironic, in relation to the rest of OpenStack, is radically different | 14:59 |
| dtantsur | I know, I meet this people regularly still. I haven't seen a single one use "the Bare Metal service". Definitely nobody was unaware of "Ironic". | 14:59 |
| dtantsur | Now, a part of my experience is biased: while "Ironic" is easy to say in non-English, "the Bare Metal service" is not translatable | 14:59 |
| TheJulia | oh, well, I was more thinking in the context, not the english word use | 14:59 |
| JayF | ah, that's an interesting point | 14:59 |
| * TheJulia thinks more fire should be applied to the name "the Bare Metal service" | 15:00 | |
| JayF | but either way, it's still what openstack docs would say to look for | 15:00 |
| JayF | so I'd just like to ensure it's on the page somewhere when we also say Ironic | 15:00 |
| JayF | maybe a little bit of the journalism-style left in me to want to define the term the first time :D | 15:00 |
| * TheJulia gets out the plasma cutter | 15:00 | |
| JayF | I will note that consistency between Ironic/ironic/Bare Metal service was one of the things pointed out in the audit | 15:01 |
| dtantsur | JayF: would it be enough to define all the terms on the landing page? (both pages: docs.o.o/ironic and the site) | 15:01 |
| dtantsur | yeah | 15:01 |
| JayF | if someone searches "bare metal openstack rescue" | 15:01 |
| JayF | I want the ironic rescue docs to come up | 15:01 |
| JayF | is the simplest way I know how to express it | 15:01 |
| dtantsur | it's probably enough if we use "bare-metal node" often enough. One mention of "bare metal service" is unlikely to trick google | 15:02 |
| dtantsur | anyway, I have nothing against the expansion of the term in the intro of every feature page | 15:02 |
| dtantsur | I'm just tired of typing "the Bare Metal service" all the time and pretending it makes sense to me :D | 15:02 |
| JayF | I am +1/+2 to making all references to ironic/Ironic/Bare Metal Service "Ironic" or "ironic" as long as we do something like Ironic (Bare Metal Service) the first time we mention it lol | 15:03 |
| JayF | dtantsur: well, and the thing is, they aren't the same | 15:03 |
| JayF | Ironic is openstack/ironic | 15:03 |
| JayF | bare metal service is all 24-ish repos | 15:03 |
| dtantsur | ironicclient is not bare metal service | 15:03 |
| dtantsur | nor is metalsmith | 15:03 |
| JayF | well, they are part of bare metal project | 15:04 |
| JayF | lolsob | 15:04 |
| * JayF goes and fetches his naming-decoder-ring from the bottom drawer | 15:04 | |
| dtantsur | 100% I've met would say "ironic project". Even I would say that :D | 15:05 |
| JayF | yeah but like, projects.yaml says we're Bare Metal Project, for instance | 15:05 |
| rpittau | they're part of Ironic (the project), but not of ironic (the service) | 15:05 |
| JayF | lolsob-harder :P | 15:05 |
| rpittau | :D | 15:06 |
| dtantsur | JayF: yeah, I"m talking about the world around our small bubble where projects.yaml is a thing :) | 15:06 |
| dtantsur | lolsob++ | 15:06 |
| JayF | My point is, there are some of our users inside that bubble too | 15:07 |
| JayF | even if we don't like it | 15:07 |
| JayF | (from a "using the long name" standpoint) | 15:07 |
| dtantsur | how many of them are unaware of "Ironic" though? | 15:07 |
| dtantsur | I claim that some of our users may not recognize "the Bare Metal service", at least without the "OpenStack" part. Like, in the CNCF context, is it Ironic or is it Metal3? Or TinkerBell? | 15:08 |
| opendevreview | Riccardo Pittau proposed openstack/ironic master: [docs] Add procedure to EoL and delete old bugfix branch https://review.opendev.org/c/openstack/ironic/+/926648 | 15:08 |
| rpittau | need to split, see you tomorrow! o/ | 15:09 |
| JayF | I wonder if Reverbverbverb has any insight given he did user interviews | 15:09 |
| JayF | because essentailly this argument is "our users are X" v "our users are Y" which are both just a combination of conjecture and personal experience | 15:10 |
| JayF | we don't have the right perspective to make this call | 15:10 |
| JayF | (in the same way that a doc written 10 years ago saying to use "Bare Metal Service" probably doesn't have the right perspective either) | 15:10 |
| opendevreview | Dmitry Tantsur proposed openstack/ironic-specs master: Support for hardware that cannot be powered off https://review.opendev.org/c/openstack/ironic-specs/+/926654 | 15:11 |
| dtantsur | ta-dam, the first draft ^^^ cc TheJulia | 15:11 |
| TheJulia | The key is linking word proximity for search engines, so on the main splash page makes tons of sense to build that keying association. | 15:11 |
| TheJulia | JayF: projects.yaml hasn't been updated in ages and is still framed from the early view of the project universe. | 15:12 |
| TheJulia | In other words, it has not been updated terminology wise to align with the times and focuses on "openstack as a whole" with a product focus | 15:13 |
| JayF | I actually completely agree with you | 15:13 |
| TheJulia | reading your very last line, seems like convergence was occuring | 15:13 |
| JayF | I think where we split is that it's more valuable for Ironic to at least align with those docs in some ways until they are updated | 15:13 |
| JayF | as long as there's a taped "X" marking our spot on the floor, we should stand on it until the X is moved :) | 15:14 |
| TheJulia | Is that really valuable to the day to day operation of the project to create churn? | 15:14 |
| JayF | but I am sometimes too much of a rule follower | 15:14 |
| * TheJulia suspects changing to align to unalign later is just an antipattern | 15:14 | |
| JayF | Well I'm saying more if we care about it, lets fix it from the source on through | 15:15 |
| JayF | if we don't care about it lets talk about some fun hardware instead ;) | 15:15 |
| opendevreview | Dmitry Tantsur proposed openstack/ironic-specs master: Support for hardware that cannot be powered off https://review.opendev.org/c/openstack/ironic-specs/+/926654 | 15:15 |
| TheJulia | I like that idea | 15:15 |
| * TheJulia goes back to wordsmithing | 15:15 | |
| * JayF has to go put up his green screen and record a podcast | 15:15 | |
| dtantsur | The thing is: if we accept certain cid's patches, we're essentially breaking the established rule | 15:16 |
| JayF | (which are aligned with the audit) | 15:16 |
| JayF | I'll note I +2'd many of those patches | 15:16 |
| dtantsur | they use "Ironic", which is something we have avoided before | 15:16 |
| dtantsur | exactly | 15:16 |
| dtantsur | so you de facto agree with us using "Ironic" from now on? | 15:16 |
| JayF | I already said I think it's a good idea *as long as* we also ensure "bare metal" is somewhere on most/all of our pages as well | 15:17 |
| TheJulia | Jay, when will you be done with your podcast | 15:17 |
| TheJulia | ? | 15:17 |
| JayF | e.g. my comment of "openstack bare metal [thing]" should have a chance of bringing up docs on ironic [thing] | 15:17 |
| JayF | TheJulia: meeting is over at 10am PT, usually done earlier than that | 15:17 |
| dtantsur | got it | 15:17 |
| JayF | we book 90 minutes but usually take <1h | 15:17 |
| TheJulia | dtantsur: established rule enforced by culture convention. This entire discussion highlights how powerful that was, and now we seek to change it as a groupd, which means it is *fine* to do so if we the people with the "skin in the game" agree | 15:18 |
| TheJulia | JayF: ack ack | 15:19 |
| TheJulia | JayF: have fun! | 15:19 |
| dtantsur | ironic-standalone-redfish is not voting, ugh | 16:09 |
| iurygregory | woot?! | 16:15 |
| iurygregory | =I | 16:15 |
| opendevreview | Dmitry Tantsur proposed openstack/ironic-tempest-plugin master: A simple standalone test for in-band inspection https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/925981 | 16:16 |
| dtantsur | JayF: do you know if we're still affected by https://review.opendev.org/c/openstack/ironic/+/910548? | 16:23 |
| JayF | yes | 16:23 |
| dtantsur | I recall us working around dnsmasq a lot.. having the standalone job disabled does look problematic | 16:23 |
| JayF | we still downgrade dnsmasq to the much less crashy version | 16:23 |
| JayF | but it's still crashy | 16:23 |
| JayF | basically: you use current DNSMasq, it flaps like a mad process and doesn't even segfault | 16:23 |
| JayF | you use the version we downgrade to, it only flaps *sometimes* and does segfault | 16:24 |
| JayF | but we've heard zero reports of this happening in the wild, and given it's basically racey, that may just be in part because hardware is slow | 16:24 |
| JayF | but really the state of dhcp server stuff in ironic, with dnsmasq being so broken and their upstream having no interest in fixing it ... it's going to be a big problem quickly | 16:24 |
| dtantsur | do you envision any path out of non-voting? the standalone job covers a lot | 16:25 |
| JayF | Julia knows more than I do about *that specific job* | 16:25 |
| JayF | but getting rid of dnsmasq is the true fix for all this BS | 16:25 |
| dtantsur | TheJulia: when you have a minute ^^ | 16:25 |
| dtantsur | yeah, sure, not happening soon though | 16:25 |
| dtantsur | is Neutron the problem? I wonder how much we lose if we kick Neutron/OVN out of the picture on the standalone job | 16:29 |
| TheJulia | OVN doesn't use dnsmasq | 16:33 |
| TheJulia | we should just move to OVN on all jobs, realistically | 16:33 |
| TheJulia | but then we're saying the dnsmasq path is not advisable, which is likely for th ebest | 16:34 |
| JayF | TheJulia: dtantsur: I think I should be able to justify pointing some GR-OSS resources at improving our DHCP server story (both in Ironic and in neutron-dhcp-agent), the biggest problem is ... what alternatives exist? | 16:36 |
| JayF | cid: ^ I may be volunteering you for a thing :D | 16:36 |
| JayF | (next cycle / to discuss at PTG timescale) | 16:36 |
| TheJulia | I'm suspecting your unaware OVN has it's own dhcp server | 16:39 |
| TheJulia | That being said, likely needs explicit support added for httpboot into ovn | 16:40 |
| * cid nods, ++, no problems | 16:40 | |
| dtantsur | hmm, so is the path forward just to use OVN in the standalone job? | 16:42 |
| * dtantsur needs to drop, sorry | 16:42 | |
| TheJulia | likely, however I suspect httpboot just won't work | 16:43 |
| TheJulia | so work in OVN+neutron will likely be needed for it to be a thing | 16:43 |
| JayF | TheJulia: I thought it was broken on baremetal? | 16:43 |
| TheJulia | JayF: what was broken on baremetal? | 16:43 |
| JayF | OVN DHCP? | 16:43 |
| TheJulia | no, it works, there are just some cases like... use of TFTP across a router, which are problematic | 16:44 |
| JayF | I have a hunch integrating a new DHCP server is likely a lower bar to clear than enabling those various bits in OVN, but I have little specific knowledge about | 16:44 |
| JayF | it | 16:45 |
| TheJulia | ... quite likely | 16:46 |
| TheJulia | OVN seems *very* slow to adopt/change underlying dhcp server interactions | 16:46 |
| TheJulia | a plus side with httpboot, we don't need the whole ipxe dance | 16:47 |
| JayF | I appreciate that, but I do think we should have a better answer for people who are still using a dhcp-based path. | 16:47 |
| * JayF curious if you need kea premium hooks to do any kind of dynamic config | 16:47 | |
| JayF | if the OSS bits of kea were enough, might be worth plugging it in, idk | 16:47 |
| jamesdenton | it seems that if one were to rely on config-drive exclusively (no dhcp), and a node has a partition on the hard disk labeled 'config-2', and you initiate clean steps that involve booting from virtual media that is labeled 'config-2'... then you likely won't get the right one mounted to /mnt/config | 18:51 |
| jamesdenton | and then Glean can't do its thing, and you sit there infinitely | 18:51 |
| jamesdenton | just an observation from a noob | 18:51 |
| JayF | TheJulia has done major work on getting this sane in our DIB image | 18:51 |
| TheJulia | jamesdenton: I think we have fixes to head that in | 18:51 |
| TheJulia | s/in/off/ | 18:51 |
| jamesdenton | you can't see me now, but i am doing a Wayne's World bow your way | 18:52 |
| TheJulia | lol :) | 18:53 |
| jamesdenton | are there any reviews that need to be looked at for that? | 19:03 |
| TheJulia | I think it all merged except a direct lockout in glean if memory serves | 19:04 |
| TheJulia | which they rejected | 19:04 |
| clarkb | right the suggestion was to disable the glean unit or something similar as part of the glean installation | 19:06 |
| clarkb | I thought we came to a reasonable compromise but I can't recall exactly what it was | 19:06 |
| TheJulia | which I think is what i ended up doing | 19:06 |
| TheJulia | it was so long ago | 19:07 |
| JayF | yeah that's why I specifically mentioned DIB in the callout to jamesdenton | 19:09 |
| jamesdenton | ahh ok, i installed ironic-python-agent-builder from pip rather that git, so i got 5.3.0 | 19:09 |
| JayF | ah, good | 19:09 |
| jamesdenton | and it doesn't include those recent commits | 19:09 |
| TheJulia | did we not release ironic-python-agent-builder recently? | 19:11 |
| jamesdenton | the latest changes improved my situation, thanks. One thing I've noticed, though, is that network_data.json is missing the 'services' bits, so there's no DNS and the API is unreachable until i munge /etc/resolv.conf. However, the provisioning subnet does have a dns nameserver in place. Not sure what's going on there | 20:29 |
| TheJulia | but no dhcp for them to be discovered through? | 20:30 |
| jamesdenton | right, dhcp is disabled on the subnet. ifcfg-eno2 is being configured with a static IP based on network_data.json (good) but there's no DNS info there | 20:34 |
| jamesdenton | however, the network_data.json that's generated and available in the *provisioned* node has services; it's seems to be limited to what's there during cleaning and deployment phases | 20:36 |
| jamesdenton | fwiw, dhcp is disabled on my provisioning network and the tenant network. | 20:37 |
| TheJulia | jamesdenton: hmm... I guess the thing to check is what is the data getting generated from the provisioning network | 20:42 |
| jamesdenton | it looks good, except all i have is a 'links' and a 'networks' key, no 'services' | 20:43 |
| jamesdenton | i'd share but it's a screenshot from console | 20:43 |
| TheJulia | ..... hmmmm no worries | 20:44 |
| TheJulia | I guess we should check to ensure Ironic is grabbing services | 20:44 |
| TheJulia | it wouldn't surprise me if it is getting dropped on the floor | 20:44 |
| jamesdenton | https://i.postimg.cc/X7XD6ggw/Screenshot-2024-08-20-at-3-27-05-PM.png | 20:44 |
| JayF | pretty hard to blame glean/cloud-init for not adding a DNS address if we didn't put one in /o\ | 21:02 |
| jamesdenton | if i knew where to look i'd happily do so :D | 21:02 |
| JayF | I know where it's generated in the nova driver case; I'm unsure where it is in Ironic | 21:03 |
| TheJulia | uhhh | 21:12 |
| TheJulia | ironic/common/neutron I think | 21:12 |
| TheJulia | https://github.com/openstack/ironic/blob/master/ironic/common/neutron.py#L499 | 21:14 |
| jamesdenton | looks that way! | 21:14 |
| JayF | I would consider a fix to that backportable in general terms | 21:14 |
| TheJulia | agree 110% | 21:18 |
| jamesdenton | i'll try something locally and push a change if it works | 21:19 |
| opendevreview | James Denton proposed openstack/ironic master: Update network_data.json to include dns nameservers https://review.opendev.org/c/openstack/ironic/+/926685 | 22:22 |
| cid | o/ | 22:42 |
| JayF | \o | 22:45 |
| opendevreview | cid proposed openstack/ironic master: [WIP] Add inspection rules https://review.opendev.org/c/openstack/ironic/+/918303 | 23:46 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
