| JayF | Just observing that only 1 spot out of the 5 possible for TC have been nominated-for so far this cycle; we have multiple people in the Ironic community who would be good candidates if they wanted to run :). | 02:07 |
|---|---|---|
| arne_wiebalck | Good morning, Ironic! | 07:14 |
| rpittau | good morning ironic! o/ | 07:41 |
| espenfl | Hi there. First, thanks a lot for the efforts around Ironic and OpenStack in general. I am a beginner trying first to use Bifrost to configure a few nodes. And in the process we hit a barrier and would like to ask for pointers where to find information or what is possible or not. On the node hosting Bifrost and Ironic we reach IPMI on an eth interface. So we would like IPMI | 08:24 |
| espenfl | traffice to go on that. But, PXE (on internal dhcp/tftp) needs preferably to happen on an ib interface (old HPC nodes and we would prefer not to manually change network boot order on all nodes). Is there are way to approach this from Bifrost? Or is this something that have to be fixed after Bifrost is deployed in the Ironic configs? I am a bit worried about this since we | 08:24 |
| espenfl | will later move to Kayobe to provision OpenStack and would like Bifrost to work in this setup. Any advice or pointers would be greatly appreciated. I know of the variable `network_interface`, but that seems to set also the IPMI traffic and I have not yet found another variable. I do notice in Ironic that there is an interface name for most properties, like dnsmasq, power, | 08:24 |
| espenfl | etc. | 08:24 |
| dtantsur | espenfl: hi. The network_interface thing defines where various services *listen*. I'm not sure why Ironic would not reach IPMI through any other interface, provided that routing is configured correctly on the machine. | 08:48 |
| dtantsur | also, network_interface in bifrost, unfortunately, clashes with *_interface on Ironic nodes. Legacy... | 08:48 |
| espenfl | Thanks. Yes, one option would be to route the IPMI traffic to/from the ib interface to the eth interface and then just use the ib interface as `network_interface`. Which can be added to any playbook. | 09:00 |
| dtantsur | espenfl: I'm still not sure why you're talking about routing traffic. Why does it matter, which interface Ironic listens on? | 09:04 |
| espenfl | dtantsur: Sorry for not being able to explain it clearly. Or maybe there is a knowledge gap for me here. Maybe both. So in Bifrost I have to give the interface responsible for communication. And it seems this have to be the same for IPMI and dnsmasq/tftp. Our nodes are a bit difficult to reconfigure and there are many. ib card is set first among the network boot devices so | 09:26 |
| espenfl | that is where PXE goes on boot. So we would like to answer requests for PXE on the ib device, while still running IPMI over the eth device on the Bifrost/Ironic node. But default is maybe that dnsmasq and tftp listens on all devices? I am testing now. | 09:26 |
| dtantsur | espenfl: the interface is only for incoming PXE/DHCP traffic, that's where dnsmasq and tftp listen. | 09:36 |
| dtantsur | It should not anyhow affect IPMI traffic, which is outgoing. | 09:36 |
| espenfl | dtantsur: Right, thanks a lot for clearing up that part. | 09:43 |
| Sandzwerg[m] | On our Openstack ironic setup the management interface is also not related to the deployment interface | 09:57 |
| Sandzwerg[m] | Something else: is anyone running nodes with openbmc? I saw an older spec for an openbmc driver but I'm not sure if that's still needed with redfish | 09:59 |
| iurygregory | good morning Ironic | 11:25 |
| TheJulia | good morning | 15:04 |
| TheJulia | Sandzwerg[m]: Last I heard, or the topic was discussed, they anticipated people would just use the redfish interface | 15:06 |
| rpittau | bye everyone, see you tomorrow! o/ | 15:32 |
| Sandzwerg[m] | TheJulia: I suspect that should work but still good to hear, thanks | 15:42 |
| TheJulia | Sandzwerg[m]: yeah, There was something with openbmc's design where just redfish was far less weight than trying to implement it in the ipmi code stack. | 15:43 |
| Sandzwerg[m] | We wouldn't use ipmi anyways. Probably generic redfish or something tailored for openbmc if that exists. It's all very far away if we get that but it seems we might test it so I'd thought I'd ask for experiences from others | 16:19 |
| TheJulia | ++ | 16:21 |
| opendevreview | Julia Kreger proposed openstack/ironic-specs master: Enumerate retired specs https://review.opendev.org/c/openstack/ironic-specs/+/873308 | 17:50 |
| opendevreview | Julia Kreger proposed openstack/ironic-specs master: Enumerate retired specs https://review.opendev.org/c/openstack/ironic-specs/+/873308 | 17:52 |
| TheJulia | so, in response to Sandzwerg[m]'s question, I've proposed a change to the specs repository to hopefully provide more clarity. | 17:52 |
| Sandzwerg[m] | Me and my questions cause more work 😅 | 17:54 |
| TheJulia | eh, it needed to happen! | 17:55 |
| TheJulia | stevebaker[m]: when you get a minute, a glance at https://review.opendev.org/c/openstack/metalsmith/+/872966 would be awesome | 19:45 |
| JayF | Also my sharding changes stand complete and lonely | 19:50 |
| TheJulia | stevebaker[m]: pretty pretty please? :) | 19:51 |
| stevebaker[m] | TheJulia: Done! | 19:56 |
| stevebaker[m] | JayF: and I'll take a proper look today | 19:56 |
| JayF | thank you very much | 19:57 |
| JayF | just remember they should merge more-or-less atomically | 19:57 |
| JayF | so don't +A any of them until you're OK with them all | 19:57 |
| opendevreview | Steve Baker proposed openstack/metalsmith master: test_provision: don't assert provision_node call order https://review.opendev.org/c/openstack/metalsmith/+/873035 | 21:16 |
| opendevreview | Merged openstack/metalsmith master: Use a network cache in Instance https://review.opendev.org/c/openstack/metalsmith/+/872963 | 21:18 |
| opendevreview | Merged openstack/metalsmith master: Get ports by 'binding:host_id' query filter https://review.opendev.org/c/openstack/metalsmith/+/872966 | 21:19 |
| * TheJulia wonders how many bare metal bear fans will align behind Otters.... | 21:40 | |
| opendevreview | Merged openstack/metalsmith master: list_instances - cache allocations https://review.opendev.org/c/openstack/metalsmith/+/873048 | 21:53 |
| TheJulia | https://openinfrafoundation.formstack.com/forms/vancouver_mascot <-- otters++ | 22:12 |
| opendevreview | Harald Jensås proposed openstack/metalsmith stable/zed: Use a network cache in Instance https://review.opendev.org/c/openstack/metalsmith/+/873276 | 22:15 |
| NobodyCam | Good afternoon Ironic folks | 22:17 |
| JayF | o/ | 23:30 |
| janders | I'm all for a Hummingbird in general. I've had their Australian relatives, Sunbirds, nest around the house for years and they are awesome. I got into bird photography because of them. BUT - they are definitely not social things. People less familiar with these birds get genuinely shocked how badass and aggressive they can be chasing off other birds | 23:32 |
| janders | (including their own kind). | 23:32 |
| janders | That anti-social-ness isn't exactly the best match for a conference mascot :/ | 23:33 |
| TheJulia | I just love how otters build things and swim and recognize humans | 23:43 |
| JayF | I like that it sounds like another Alanis Morrisette song. | 23:43 |
| JayF | YOU OTTER KNOWWWWWW | 23:43 |
| * TheJulia spent time with sea otters at one point with a camera | 23:43 | |
| JayF | I'm here to remind you of the PUN YOU GAVE ME | 23:43 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!