| opendevreview | Steve Baker proposed openstack/ironic-python-agent master: Switch centos9 jobs to centos10 https://review.opendev.org/c/openstack/ironic-python-agent/+/992571 | 00:03 |
|---|---|---|
| opendevreview | Merged openstack/ironic bugfix/34.0: Set bound for failure reason from RFB https://review.opendev.org/c/openstack/ironic/+/991939 | 00:10 |
| opendevreview | Merged openstack/ironic stable/2026.1: Set bound for failure reason from RFB https://review.opendev.org/c/openstack/ironic/+/991938 | 00:33 |
| opendevreview | Merged openstack/ironic unmaintained/2024.1: ci: stable: fixes: mem in ironic-base, rm iso test https://review.opendev.org/c/openstack/ironic/+/991859 | 01:14 |
| opendevreview | Merged openstack/ironic stable/2025.2: Set bound for failure reason from RFB https://review.opendev.org/c/openstack/ironic/+/991941 | 01:15 |
| opendevreview | Dmitry Tantsur proposed openstack/ironic master: Modernize the Metal3 functional jobs https://review.opendev.org/c/openstack/ironic/+/992607 | 10:06 |
| dtantsur | Hopefully ^^ brings the BMO job back into operation | 10:06 |
| rpittau | dtantsur: should we remove Julia's comment then ? | 10:14 |
| rpittau | and I thin we need to modify the releae how page | 10:15 |
| rpittau | (I can't write today) | 10:15 |
| dtantsur | Julia's comment still applies to the metal3-integration job (while it exists) | 10:15 |
| dtantsur | good point on release how. let me wait for the CI results first though | 10:16 |
| dtantsur | (do you also feel fried today? I cannot wake up after sleeping nearly 9 hours) | 10:16 |
| rpittau | (yep, kind of, my neck is busted :D ) | 10:17 |
| dtantsur | ugh, sorry to hear, I hope you get better soon | 10:17 |
| rpittau | thanks :) | 10:17 |
| rpittau | dtantsur: oh ok, going back to the comment, that refers to the metal3-integration only, alright | 10:18 |
| opendevreview | Dmitry Tantsur proposed openstack/ironic master: Modernize the Metal3 functional jobs https://review.opendev.org/c/openstack/ironic/+/992607 | 11:14 |
| dtantsur | rpittau: with updated release docs ^^ | 11:14 |
| opendevreview | Harald Jensås proposed openstack/networking-generic-switch master: Refactor: extract OpenConfigModelMixin https://review.opendev.org/c/openstack/networking-generic-switch/+/992884 | 11:36 |
| opendevreview | Harald Jensås proposed openstack/networking-generic-switch master: Add to_restconf_dict() methods to OpenConfig model https://review.opendev.org/c/openstack/networking-generic-switch/+/992885 | 11:36 |
| opendevreview | Harald Jensås proposed openstack/networking-generic-switch master: restconf: add path helpers and resource_path util https://review.opendev.org/c/openstack/networking-generic-switch/+/992886 | 11:36 |
| opendevreview | Harald Jensås proposed openstack/networking-generic-switch master: Add RestconfSwitch base class for HTTP transport https://review.opendev.org/c/openstack/networking-generic-switch/+/992887 | 11:36 |
| opendevreview | Harald Jensås proposed openstack/networking-generic-switch master: Add RestconfOpenConfigSwitch driver https://review.opendev.org/c/openstack/networking-generic-switch/+/992888 | 11:36 |
| opendevreview | Harald Jensås proposed openstack/networking-generic-switch master: Add RESTCONF documentation https://review.opendev.org/c/openstack/networking-generic-switch/+/992889 | 11:36 |
| opendevreview | Mithun Krishnan Umesan proposed openstack/networking-baremetal master: Upgrade OVN SSL connections to TLS 1.3 with hybrid PQC key exchange https://review.opendev.org/c/openstack/networking-baremetal/+/992495 | 12:02 |
| TheJulia | good morning | 13:15 |
| TheJulia | dtantsur: personally, I feel like a zombie this morning | 13:16 |
| dtantsur | it's one of these mornings.. | 13:46 |
| opendevreview | Mithun Krishnan Umesan proposed openstack/networking-baremetal master: Upgrade OVN SSL connections to TLS 1.3 with hybrid PQC key exchange https://review.opendev.org/c/openstack/networking-baremetal/+/992495 | 14:50 |
| JayF | fyi CI is failing on some bugfix branch security fixes: https://review.opendev.org/c/openstack/ironic/+/987920 + https://review.opendev.org/c/openstack/ironic/+/987921 (for the jinja rendering) | 15:22 |
| JayF | I'm working to get the stable/* and unmaintained/* patches all merged and those branches in a good state | 15:22 |
| JayF | https://review.opendev.org/c/openstack/ironic/+/991384 is for unmaintained/2024.1 -- backported by me, I need someone else to +2A please | 15:23 |
| JayF | same for https://review.opendev.org/c/openstack/ironic/+/991385 which deps on the previous one | 15:24 |
| dtantsur | JayF: do you see any logs for the failures? I don't see anything there. | 15:25 |
| JayF | wow, I didn't even notice how old they were | 15:25 |
| JayF | good call I shoulda just rechecked to get the data | 15:26 |
| dtantsur | and maybe even rebase because I'm quite sure I've fixed teh bugfix branches since then | 15:26 |
| JayF | just a recheck will pull in the stuff | 15:26 |
| JayF | (right?) | 15:26 |
| JayF | I'm 99% sure that's how it works but I can't articulate why so now I'm nervous lol | 15:27 |
| dtantsur | Now that you say it, I do remember that Gerrit/Zuul rebase everything internally before running tests | 15:29 |
| JayF | https://review.opendev.org/c/openstack/ironic/+/992717 (and friends) need a review/land to move forward | 15:29 |
| dtantsur | (fun story: we just got into a situation in BMO that Zuul would have prevented: two correct PRs produced incorrect results after merging together) | 15:29 |
| JayF | that's a backport I'm doing for hardening that I'd like to get in | 15:29 |
| JayF | alegacy: I'm looking at https://docs.openstack.org/ironic/latest/install/standalone/networking.html and it's not clear to me how IP addresses get assigned, can you help? e.g. if I provision node X, how does node X know what IP to use? | 15:56 |
| opendevreview | Merged openstack/ironic unmaintained/2024.1: Fix kernel parameter parsing for quoted values and whitespace https://review.opendev.org/c/openstack/ironic/+/992659 | 16:34 |
| opendevreview | Merged openstack/ironic bugfix/34.0: security: Use sandbox rendering for jinja2 https://review.opendev.org/c/openstack/ironic/+/987920 | 18:06 |
| opendevreview | Merged openstack/ironic unmaintained/2024.1: security: move file url validation up into deploy_utils main path https://review.opendev.org/c/openstack/ironic/+/988359 | 18:08 |
| JayF | https://review.opendev.org/c/openstack/ironic/+/991384 and https://review.opendev.org/c/openstack/ironic/+/991385 need approvals | 18:14 |
| TheJulia | Anyone know of ilo folks to talk to regarding nvmeof/tcp booting of nodes | 18:23 |
| opendevreview | Julia Kreger proposed openstack/ironic-specs master: Crazy idea: RCOE Enabled BMaaS https://review.opendev.org/c/openstack/ironic-specs/+/990413 | 18:28 |
| opendevreview | Merged openstack/ironic unmaintained/2024.1: security: Use sandbox rendering for jinja2 https://review.opendev.org/c/openstack/ironic/+/987777 | 18:41 |
| opendevreview | Jay Faulkner proposed openstack/ironic unmaintained/2023.1: ci: stable: fixes: mem in ironic-base, rm iso test https://review.opendev.org/c/openstack/ironic/+/993014 | 19:07 |
| opendevreview | Merged openstack/ironic unmaintained/2024.1: security: validate molds url against swift in keystone catalog https://review.opendev.org/c/openstack/ironic/+/986816 | 19:23 |
| opendevreview | Clif Houck proposed openstack/ironic-python-agent stable/2026.1: Add a flag to disable installing bootloaders https://review.opendev.org/c/openstack/ironic-python-agent/+/993016 | 19:36 |
| opendevreview | Verification of a change to openstack/ironic bugfix/34.0 failed: Add operator-configurable step disallow lists https://review.opendev.org/c/openstack/ironic/+/992717 | 19:36 |
| alegacy | JayF: sorry, not sure I follow your question. Which part of the doc are you asking about? | 19:36 |
| opendevreview | Clif Houck proposed openstack/ironic-python-agent stable/2025.2: Add a flag to disable installing bootloaders https://review.opendev.org/c/openstack/ironic-python-agent/+/993020 | 19:38 |
| JayF | alegacy: For a user of ironic-networking, how do they assign IP addresses to nodes? | 19:43 |
| alegacy | JayF: that's not really something that the feature does. It strictly configures switch ports. | 19:44 |
| JayF | Oh wow. That's like, half of the feature that folks would need to run this | 19:44 |
| JayF | I'm assuming there's a strong assumption of DHCP + Metal3-only use cases? | 19:44 |
| JayF | What do we put in network_data.json? | 19:44 |
| alegacy | the network data would be handled just as it would normally/without ironic-networking | 19:45 |
| alegacy | anyone running ironic in a standalone fashion today (w/out neutron) would have to configure network data, and manually configure the switch ports before booting the node. ironic-networking removes the "manually configure the switch port" part, but the rest is the same. | 19:47 |
| JayF | manually configure network data meaning what? | 19:47 |
| opendevreview | Clif Houck proposed openstack/ironic-python-agent stable/2025.1: Add a flag to disable installing bootloaders https://review.opendev.org/c/openstack/ironic-python-agent/+/993024 | 19:47 |
| JayF | I'm specifically talking about the file that cloud-init reads to setup networks; network_data.json | 19:47 |
| JayF | I had some downstream interest in this, but unless we have some way for folks to assign IP addresses in Ironic and pass thru network data, this feature is not generically useful | 19:49 |
| alegacy | ordinarily that's provided when the node is enrolled is it not -- if static ip addressing is required? | 19:49 |
| JayF | In the integrated case, flat network interface or not, the IP address information comes over from Neutron. Neutron acts as IPAM. | 19:50 |
| JayF | So one of the somewhat-obvious questions in a neutron->ironic-networking migration is "Who assigns my IP addresses" and/or "how do I integrate an external IPAM" | 19:50 |
| JayF | Nova puts that into a configdrive and passes it over to Ironic. | 19:50 |
| JayF | So I'm thinking: "As a user, I want to provision a machine using Ironic and Ironic Networking, and have cloud-init configure networking on startup of the machine" as a user story | 19:52 |
| JayF | if we don't have that, we implicitly are limited to non-LAG, dhcp-all-interfaces style configurations only, which means ironic-networking is more-or-less suitable for metal3-shaped problems only | 19:52 |
| opendevreview | Clif Houck proposed openstack/ironic-python-agent unmaintained/2024.1: Add a flag to disable installing bootloaders https://review.opendev.org/c/openstack/ironic-python-agent/+/993025 | 19:54 |
| JayF | (To be clear; I'm trying to get the borders so I can draw up what's needed to get this usable for the use cases I'm interested in; what we have is cool just less than I thought we did) | 19:55 |
| clif | JayF: do you have a list of bugfix branches I need to backport to? or just all of them? | 19:55 |
| JayF | oh, for IPA? | 19:56 |
| JayF | oh | 19:56 |
| clif | yep | 19:56 |
| JayF | https://etherpad.opendev.org/p/IronicWhiteBoard line 86 | 19:56 |
| JayF | that looks correct to me | 19:56 |
| JayF | (this is where the bitly link in /topic goes to) | 19:56 |
| clif | so just 11.6 and 11.4 | 19:57 |
| clif | what about the named unmaintained releases? | 19:57 |
| JayF | From an Ironic-only perspective: we only owe for bugfix/ and stable/ | 19:57 |
| opendevreview | Clif Houck proposed openstack/ironic-python-agent unmaintained/2023.1: Add a flag to disable installing bootloaders https://review.opendev.org/c/openstack/ironic-python-agent/+/993026 | 19:57 |
| JayF | from a GR-OSS perspective: we need at least unmaintained/2024.1 | 19:57 |
| JayF | I've been doing all the way back to um/2023.1 which is about the oldest reasonable one someone would be running | 19:58 |
| alegacy | JayF: sorry, just trying to find an example of what I'm aware of. to make sure i'm understanding correctly... will get back here as soon as I can, but I need to stop soon so might be tomorrow morning. | 20:01 |
| JayF | there is zero urgency to the ask :) had a leader at my downstream read the docs and started asking questions and I ran outta answers :D | 20:01 |
| JayF | this is the sorta thing where I suspect it's 2-5 minor "glue" style features (+LAG) to get it to a point where it's useful for my use cases | 20:02 |
| alegacy | k, will get back to you. possible I'm in the weeds, but 75% sure I'm not. | 20:02 |
| JayF | If we can answer the question "where does the source data for network_data.json generation in ironic-networking come from", I'll have 75% of my answer :) | 20:03 |
| JayF | if that answer is "configdrive + ironic-networking? Nope" then I'll start to actually be sad :D | 20:03 |
| alegacy | this is what I was thinking of: https://docs.openstack.org/ironic/2026.1/admin/dhcp-less.html | 20:06 |
| alegacy | aside from the metal3 type usecase where that's baked into the virtual iso being mounted/booted | 20:07 |
| JayF | Ooof. OK. | 20:08 |
| JayF | That means the people/automation pushing the buttons to deploy also get to assign whatever IPs they want | 20:09 |
| JayF | which I guess is OK since we control VLANs and can prevent breakage | 20:09 |
| TheJulia | Yeah, I think the world it is built for is one where your just feeding the data, or having a higher level orchustration system doing some of that data generation and push out | 20:09 |
| JayF | but makes it not suitable for multitenant | 20:09 |
| JayF | I will say, I'll have a hard time +2'ing a design for LAG that will boil down to an operator properly writing their own network_data.json and Ironic configuring the switch | 20:10 |
| alegacy | I think IP address management is a different concern than the switch port management. Not saying there isn't a use case for both, but I view them as separate things. | 20:11 |
| alegacy | so tying LAG to it isn't really respecting that they are seperate things. | 20:12 |
| opendevreview | Merged openstack/ironic unmaintained/2023.1: Fix kernel parameter parsing for quoted values and whitespace https://review.opendev.org/c/openstack/ironic/+/992661 | 20:12 |
| JayF | I agree, but at the point you start doing LAGs, you are requiring the network config on the node cooperate with that -- almost all configurations of LAG that work under load require client/OS side awareness | 20:12 |
| JayF | So I'm not saying, for instance, that Ironic needs to handle IPAM. I'm saying that asking an operator to statically setup OS bond configs and ALSO statically setup the ports to be bonded is maybe a bridge too far | 20:13 |
| alegacy | take a look at get_node_network_data in ironic/drivers/modules/network/ironic_networking.py. it is capable of providing network data (minus ip addressing) which in theory should allow the bond setup to work at least. | 20:15 |
| JayF | (I would assume even if we found a way to make this work better, we'd still be requiring node-specific IP/Netmask/DNS -- it's less about Ironic being the IPAM, and more about having a abstracted-from-hardware way to indicate what IP you want) | 20:15 |
| JayF | OK yes exactly this is shaped like I hoped it would be | 20:16 |
| JayF | if we just added a way for someone to indicate the desired IP/netmask/DNS on the port as well (port{.group}.extra.ip_address/netmask/dns_servers), it'd be a trivial add-on feature to populate them into the network_data | 20:17 |
| JayF | this is awesome | 20:17 |
| JayF | I'm back to "I think we need 3-5 minor-sized UI features to make this awesome for standalone users" | 20:17 |
| TheJulia | Yeah, there is a whole allocation/orchestration standpoint which is a trade-off | 20:23 |
| JayF | Honestly, I just see it being a ... magnet for failures if we expect folks to automate network_data to match how the ports are configured, just a little bit of help from us can reduce the complexity of what would be driving this considerably | 20:25 |
| JayF | especially if we routed it through ports -- then you're essentially provisioning the IP along with the VLANs and other networking information -- as an "administrative" or "managerial" task that's separate from "deploy me a server" | 20:25 |
| JayF | alegacy: thank you :) this is good food for thought and I think we can whip up an RFE or three to set the table for LAG support \o/. Maybe have it figured out so it's ready for you once you come back to implement it after getting metal3 hooked up | 20:28 |
| TheJulia | I guess part of the base challenge is everyone's model is different and what one operator may do and mean by a word or meanting may be entirely different for another's use case and context | 20:33 |
| JayF | Yep. I'm working under the assumption I'm the first person looking at it from a usability perspective with the use case of "use with integrated openstack without neutron" | 20:34 |
| JayF | as opposed to what it was designed for: metal3 running the show :D | 20:35 |
| TheJulia | Yeah, pretty much | 20:52 |
| TheJulia | The other issue is if you have anything requiring dhcp, your going to have to manage it along with vlan numbering/allocation/etc. | 21:10 |
| TheJulia | and... depending on the what, in the end, that may be okay | 21:10 |
| opendevreview | Clif Houck proposed openstack/ironic-python-agent unmaintained/2023.1: Set tox basepython to python3.11 https://review.opendev.org/c/openstack/ironic-python-agent/+/993039 | 21:22 |
| clif | what's the right way to pin back python versions so tests will pass on past releases? | 21:24 |
| clif | set basepython in tox.ini? something else? | 21:24 |
| clif | for instance the failures on this look like python mismatch errors: https://review.opendev.org/c/openstack/ironic-python-agent/+/993024 | 21:24 |
| clif | ModuleNotFoundError: No module named 'pkg_resources' | 21:24 |
| alegacy | JayF: looking forward to discussing more when I've succeeded in integrating what we've got so far up the stack a bit. | 21:44 |
| opendevreview | Steve Baker proposed openstack/ironic-python-agent-builder master: Mask dnf-makecache and fwupd-refresh services in ramdisk https://review.opendev.org/c/openstack/ironic-python-agent-builder/+/993049 | 22:11 |
| opendevreview | Steve Baker proposed openstack/ironic-python-agent-builder master: Uninstall unnecessary packages to reduce ramdisk size https://review.opendev.org/c/openstack/ironic-python-agent-builder/+/993050 | 22:11 |
| opendevreview | Steve Baker proposed openstack/ironic-python-agent-builder master: Prune additional paths from ramdisk to reduce initramfs size https://review.opendev.org/c/openstack/ironic-python-agent-builder/+/993051 | 22:11 |
| opendevreview | Steve Baker proposed openstack/ironic-python-agent-builder master: Remove GPU, WiFi and vendor laptop firmware from ramdisk https://review.opendev.org/c/openstack/ironic-python-agent-builder/+/993052 | 22:11 |
| JayF | clif: there are changes in Ironic which pin setuptools in tox.ini to fix that shape of issue, dig for the commit in one of the older stable branches and it should be clear | 22:17 |
| opendevreview | Merged openstack/ironic unmaintained/2023.1: ci: stable: fixes: mem in ironic-base, rm iso test https://review.opendev.org/c/openstack/ironic/+/993014 | 22:37 |
| JayF | clif: https://opendev.org/openstack/ironic/commit/42a413b615fa0a72de78f7513b96c53650412d27 | 22:39 |
| opendevreview | Pierre Riteau proposed openstack/tenks master: Add openstack.cloud collection to requirements https://review.opendev.org/c/openstack/tenks/+/993060 | 23:07 |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!