| rpittau | good morning ironic! o/ | 06:45 |
|---|---|---|
| opendevreview | Riccardo Pittau proposed openstack/bifrost master: [WIP] Test 2 major ansible versions https://review.opendev.org/c/openstack/bifrost/+/925563 | 07:46 |
| opendevreview | Riccardo Pittau proposed openstack/bifrost master: [WIP] Test 2 major ansible versions https://review.opendev.org/c/openstack/bifrost/+/925563 | 07:49 |
| masghar | Good morning! | 08:05 |
| adam-metal3 | good day Ironic! | 08:23 |
| janders | iurygregory should we consider enabling redfish firmware_interface by default in bifrost? (or alternatively add a flag allowing enabling it during install)? (CC dtantsur) | 08:34 |
| opendevreview | Maryna Savchenko proposed openstack/ironic master: Add community images to always available. https://review.opendev.org/c/openstack/ironic/+/903179 | 09:47 |
| dtantsur | janders: I have nothing against enabling it whenever redfish is enabled | 10:08 |
| janders | dtantsur that may be a slick way to do it | 10:22 |
| janders | good idea | 10:22 |
| opendevreview | cid proposed openstack/ironic master: Link to configuration options https://review.opendev.org/c/openstack/ironic/+/925756 | 10:43 |
| opendevreview | cid proposed openstack/ironic-inspector master: Link to configuration options https://review.opendev.org/c/openstack/ironic-inspector/+/925757 | 10:44 |
| opendevreview | cid proposed openstack/ironic-python-agent master: Link to configuration options https://review.opendev.org/c/openstack/ironic-python-agent/+/925758 | 10:45 |
| opendevreview | cid proposed openstack/python-ironic-inspector-client master: Link to configuration options https://review.opendev.org/c/openstack/python-ironic-inspector-client/+/925759 | 10:46 |
| opendevreview | cid proposed openstack/ironic-inspector master: Link to configuration options https://review.opendev.org/c/openstack/ironic-inspector/+/925757 | 10:47 |
| opendevreview | cid proposed openstack/ironic-python-agent master: Link to configuration options https://review.opendev.org/c/openstack/ironic-python-agent/+/925758 | 10:48 |
| opendevreview | cid proposed openstack/python-ironic-inspector-client master: Link to configuration options https://review.opendev.org/c/openstack/python-ironic-inspector-client/+/925759 | 10:48 |
| opendevreview | Jacob Anders proposed openstack/bifrost master: Enable configuration of firmware_interfaces during install https://review.opendev.org/c/openstack/bifrost/+/925761 | 11:23 |
| iurygregory | good morning o/ | 11:28 |
| opendevreview | Maryna Savchenko proposed openstack/ironic master: Fixes config drive base64 encoding in kickstart file https://review.opendev.org/c/openstack/ironic/+/925569 | 11:39 |
| opendevreview | Maryna Savchenko proposed openstack/ironic master: Fixes config drive base64 encoding in kickstart file https://review.opendev.org/c/openstack/ironic/+/925569 | 12:06 |
| opendevreview | Maryna Savchenko proposed openstack/ironic master: Add community images to always available. https://review.opendev.org/c/openstack/ironic/+/903179 | 12:27 |
| janders | cd .. | 12:50 |
| janders | oops :) wrong window fail | 12:50 |
| janders | dtantsur iurygregory https://review.opendev.org/c/openstack/bifrost/+/925761 seems to work, put my test setup details in a comment | 13:17 |
| iurygregory | janders, ack o/ | 13:17 |
| janders | with this patch in place I am able to test servicing end-to-end without any post-install hacks or node updates | 13:17 |
| janders | (only thing is changing boot_interface in vars but I think this is OK) | 13:17 |
| janders | I don't think it's a good goal to have a bifrost-cli option for every inventory variable so this one may just need to be an in-inventory adjustment | 13:18 |
| dtantsur | JayF: maybe a question for you: do you remember why we don't use soft power off via BMC in https://opendev.org/openstack/ironic/src/branch/master/ironic/drivers/modules/agent_base.py#L1283-L1339? Was it simply not a thing back in the days? | 13:26 |
| opendevreview | Riccardo Pittau proposed openstack/bifrost master: Test 2 major ansible versions https://review.opendev.org/c/openstack/bifrost/+/925563 | 13:27 |
| opendevreview | Riccardo Pittau proposed openstack/bifrost master: Test 2 major ansible versions https://review.opendev.org/c/openstack/bifrost/+/925563 | 13:28 |
| TheJulia | dtantsur: if we use soft power off, at least with ipmi, we would have had to wait and continue to poll which was likely a crash risk with awful ipmi bmcs | 14:47 |
| TheJulia | redfish that is likely less of a problem but some vendors are already horrible about updating power status in their bmcs today too | 14:47 |
| * TheJulia remembers when ironic could crash ilo3 and some early ilo4 bmc firmware | 14:48 | |
| TheJulia | s/could/would/ | 14:48 |
| TheJulia | o/ arne_wiebalck | 14:52 |
| arne_wiebalck | Good morning Ironic and TheJulia o/ | 14:53 |
| JayF | dtantsur: TheJulia: it also required the ipmi interface to have the modules loaded/driver side assistance | 15:04 |
| JayF | dtantsur: TheJulia: so that impacted our desire to use it heavily; it would, on much ipmi hardware, just break/timeout | 15:04 |
| TheJulia | Isent' there a whole thing about the OS has to handle the soft power off | 15:12 |
| dtantsur | TheJulia: fwiw we already wait and poll in that code | 15:13 |
| * dtantsur has another iteration of the NC SI discussion at RH... | 15:13 | |
| TheJulia | well, that code asks the agent to power off the node first | 15:13 |
| TheJulia | so a self power off | 15:13 |
| dtantsur | Yeah, but still polls the BMC | 15:14 |
| TheJulia | yeah | 15:14 |
| dtantsur | anyway, I'm wondering how I can quickly and painful rip this agent power off out... | 15:14 |
| dtantsur | * painlessly | 15:14 |
| TheJulia | do we need an nc-si power overlay... or optional underlay driver?! | 15:14 |
| dtantsur | We're going to work on it. But I need something for yesterday :( | 15:14 |
| TheJulia | the driver is that nic link dropping right ? | 15:14 |
| TheJulia | of course :( | 15:15 |
| dtantsur | The "best" idea I have now is to conditionally monkey-patch the power_off call in our hardware manager | 15:15 |
| dtantsur | yeah | 15:15 |
| TheJulia | if that is the simplest, perhaps easiest and then we can figure out a proper way upstream | 15:15 |
| dtantsur | I don't quite like the 1 minute delay while polling the BMC.. but I guess that's life (and their BMC is emulated anyway) | 15:16 |
| TheJulia | blah | 15:16 |
| JayF | TheJulia: exactly, in IPMI it sends a signal through to the OS, you need (in linux) the IPMI modules loaded for it to work. We pushed hard against any use of this when I was at Rackspace because this was very breaky in our model where we disabled that interface explicitly | 15:17 |
| JayF | I think in a redfish world we can revisit that thought upstream | 15:18 |
| TheJulia | in redfish bmcs, aiui, it just triggers the soft power signaling to the OS via acpi | 15:19 |
| TheJulia | I think newer gear, even with ipmi also signals the via the acpi interface, but again, OS needs to be there | 15:19 |
| JayF | yeah, I'd believe that is how it works on reasonable hardware | 15:23 |
| JayF | I know older ipmi gear, even without our mods, had the driver-based behavior, but it always seemed silly to me | 15:23 |
| kubajj | Hello everyone o/ | 15:23 |
| TheJulia | Yeah, that is why we called the sync to it, becuase of misbehaving hardware raid | 15:23 |
| JayF | o/ | 15:24 |
| TheJulia | ... like... who doesn't commit to flushing buffers unless a sync comes down to the device. | 15:24 |
| TheJulia | I guess dtantsur might find hardware continuing to not do that, since we had those raid controllers which were super lazy about flushing buffers out | 15:29 |
| dtantsur | sigh | 15:30 |
| TheJulia | if you keep sync that is far less of a rsik | 15:30 |
| TheJulia | risk | 15:30 |
| JayF | sync;sync;sync;sync;echo s>/proc/sysrq-trigger;sync | 15:30 |
| JayF | gotta be sure | 15:30 |
| TheJulia | since that does a direct write | 15:30 |
| JayF | the original agent reboot was echo s > /proc/sysrq-trigger; echo b > /proc/sysrq-trigger | 15:31 |
| TheJulia | and I think we've only spotted one device which didn't flush out afterwards | 15:31 |
| JayF | I think that got fixed before it got upstreamed tho | 15:31 |
| TheJulia | nah, it was upstream at a point | 15:31 |
| JayF | I wrote that code and can't remember why I did it that way | 15:31 |
| JayF | maybe $container-problems? or $coreos-problems? | 15:31 |
| TheJulia | I remember coreos changed supported behavior at one point | 15:32 |
| TheJulia | And I remember that got refactored then | 15:32 |
| JayF | I patched systemd-nspawn once to try and keep the existing pattern; when they broke us again it became clear we were the only folks interested in containers that could do hardware stuff lol | 15:32 |
| TheJulia | yeah, I think that was the last straw for you with coreos | 15:34 |
| JayF | then we did chroot while figuring out a better solution | 15:34 |
| JayF | we are still looking for that better solution /s | 15:34 |
| TheJulia | python2uefi | 15:34 |
| JayF | (DIB is much better than it was, I just liked the idea of the ramdisk being separate from IPA; but they are inseparable now) | 15:34 |
| TheJulia | I *really* think the best path is a signed EFI binary, but that could also just be linux which doesn't load *any* drivers and has the modules to map system kernel stuffs to EFI calls. | 15:35 |
| JayF | You think the issues we get with nutty hardware are bad now? lol | 15:36 |
| TheJulia | but I'm also crazy like Starbuck (Kara Thrace) | 15:37 |
| TheJulia | eh, then they are not efi compliant :) | 15:37 |
| * TheJulia should resume the Battlestar Galactica re-watch | 15:38 | |
| JayF | I'm sure those raid controllers are not something-compliant too | 15:39 |
| JayF | just pick some letters outta the alphabet soup :D | 15:39 |
| TheJulia | https://external-preview.redd.it/0gVGkRu3drrCNCjFntCTrAwBVY9_3h2WSMQm0DIWOAc.jpg?auto=webp&s=15574d55285f5ab2bf093dd6357f512f7c28c1c3 | 15:40 |
| dtantsur | :D | 15:45 |
| opendevreview | Riccardo Pittau proposed openstack/bifrost master: Test 2 major ansible versions https://review.opendev.org/c/openstack/bifrost/+/925563 | 15:53 |
| rpittau | good night! o/ | 15:58 |
| TheJulia | dtantsur: at some point, we'll have AI generated movies, and I feel like "give me a rendition of all of the starwars movies where all the sith are cats in sith robes" | 16:07 |
| dtantsur | finally, a solid use case for AI! | 16:07 |
| JayF | who needs AI, get that to sundance | 16:13 |
| TheJulia | Only if I get to acquire new camera gear and the kitten/cat cast finds good homes where they get plenty of tasty wet food for the whole of their lives | 16:49 |
| JayF | cats and cast, it's so obvious, they were meant for the silver screen | 16:54 |
| TheJulia | Summer can be a producer, we'll just need someone to finance our efforts | 17:03 |
| cid | o// | 20:06 |
| JayF | https://bugs.launchpad.net/ironic/+bug/2045191 appears to have expired out due to incomplete status never getting updated on it | 21:51 |
| JayF | I triaged it | 21:51 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
