| opendevreview | Iury Gregory Melo Ferreira proposed openstack/ironic-specs master: [WIP] Firmware Updates Interface https://review.opendev.org/c/openstack/ironic-specs/+/878505 | 04:16 |
|---|---|---|
| dtantsur | two older backports if anyone has time: https://review.opendev.org/c/openstack/ironic/+/876178 and https://review.opendev.org/c/openstack/ironic/+/876174 | 10:08 |
| kaloyank | Hello Ironic o/ | 11:29 |
| iurygregory | morning ironic | 11:31 |
| iurygregory | dtantsur, looking now | 11:31 |
| kaloyank | TheJulia, dtantsur, I want to implement the snapshot spec (https://specs.openstack.org/openstack/ironic-specs/specs/not-implemented/snapshot-support.html). I have the following questions: | 11:51 |
| kaloyank | 1. You mentioned that the spec might be changed as it's a little out of date. I'd like to make some suggestions, how can I do that? | 11:52 |
| kaloyank | 2. I'm currently running Yoga, will update to Antelope once it hits RDO but still it remains unknown to me how will I test my code. SHould I setup a testing env? | 11:53 |
| dtantsur | kaloyank: #2 you can use https://docs.openstack.org/ironic/latest/contributor/dev-quickstart.html#deploying-ironic-with-devstack or https://docs.openstack.org/bifrost/latest/install/index.html | 11:54 |
| kaloyank | dtantsur: Thanks, I'll check it out | 11:55 |
| kaloyank | also, as there are changes to be made to IPA, I presume that I'll have to rebuild the IPA, which is fine. How can I test that the IPA is doing the correct thing? | 11:57 |
| kaloyank | Shall I use a VM that network boots or there's some other way? | 11:57 |
| dtantsur | kaloyank: both devstack and bifrost allow creating a testing environment with VMs serving as fake bare metal machines. | 11:59 |
| kaloyank | awesome, thanks | 12:01 |
| opendevreview | Maksim Malchuk proposed openstack/bifrost master: Create the log file for the disk-image-create command https://review.opendev.org/c/openstack/bifrost/+/822895 | 12:25 |
| opendevreview | Maksim Malchuk proposed openstack/bifrost master: Create the log file for the disk-image-create command https://review.opendev.org/c/openstack/bifrost/+/822895 | 12:33 |
| opendevreview | Merged openstack/ironic stable/yoga: Do not move nodes to CLEAN FAILED with empty last_error https://review.opendev.org/c/openstack/ironic/+/876178 | 14:08 |
| opendevreview | Merged openstack/ironic bugfix/21.2: Do not move nodes to CLEAN FAILED with empty last_error https://review.opendev.org/c/openstack/ironic/+/876174 | 14:08 |
| opendevreview | Merged openstack/ironic master: Use main branch of metal3-dev-env to run metal3 integration job https://review.opendev.org/c/openstack/ironic/+/877600 | 14:08 |
| iurygregory | dtantsur, quick question regarding for rpc api impact, you don't think we would need because we will re-utilize the current update_firmware clean step we have? | 15:26 |
| iurygregory | I'm also a bit puzzled since you mentioned we need to settle on a new standard name for the clean step... | 15:27 |
| dtantsur | iurygregory: clean steps don't require new RPC's, they all go through one generic call | 15:33 |
| dtantsur | do_node_clean or how is it called? | 15:33 |
| iurygregory | yeah | 15:35 |
| iurygregory | only raid had specific ones | 15:35 |
| dtantsur | yeah, because it has conductor-side validations IIRC | 15:36 |
| iurygregory | gotcha | 15:36 |
| iurygregory | do we plan on re-use the clean step update_firmware or would be a new one? | 15:37 |
| iurygregory | and I think one possible approach to get the current versions of FW available would be maybe via verify_steps... | 15:37 |
| dtantsur | iurygregory: update_firmware is not even on the correct interface :) | 15:42 |
| iurygregory | yeah, it lives in management I think | 15:43 |
| iurygregory | so it would be a new clean step (I'm worried about the name, since we have update_firmware in management), will update things in the spec based on your answers o/ | 15:44 |
| iurygregory | tks! | 15:45 |
| prometheanfire | dtantsur: got stuff working, fat fingered a neutron config, does the 'no root device specified' when booting a node mean that I need to pass in root device hints? | 18:12 |
| prometheanfire | nevermind, it looks like ipa is trying to use the internal url to talk back to ironic instead of the public endpoint | 19:28 |
| JayF | That is intentional, iirc | 19:31 |
| JayF | you can set an external url per node or in config | 19:31 |
| JayF | if you have network segmentation such that the URL an IPA needs to hit is different than what's internal | 19:31 |
| prometheanfire | happen to know which flag to set? trying to google isn't helping | 19:31 |
| prometheanfire | ya, internal url is not https, so sits on it's own non-routed vlan | 19:32 |
| prometheanfire | (OSA deploy) | 19:32 |
| JayF | https://docs.openstack.org/ironic/latest/admin/dhcp-less.html#deploying-outside-of-the-provisioning-network | 19:32 |
| JayF | that is covering a specific case, but I think the abilities there might be helpful to you | 19:33 |
| prometheanfire | ya, looks useful | 19:33 |
| JayF | hope it's useful :D | 19:33 |
| prometheanfire | ya, I think it's specifically the callback url | 19:35 |
| jrosser | prometheanfire: you know i updated the OSA documentation for ironic very recently with a worked example? | 19:39 |
| prometheanfire | I have been using the osa ironic docs :D | 19:41 |
| jrosser | prometheanfire: including all the settings you need to make the callback work with an isolated network for ironic provisioning | 19:41 |
| prometheanfire | ok, didn't see that before | 19:41 |
| prometheanfire | just looked at latest os_ironic docs and still didn't see it | 19:42 |
| jrosser | this https://docs.openstack.org/openstack-ansible-os_ironic/latest/configure-lxc-example.html | 19:43 |
| prometheanfire | didn't mention the external callback url? | 19:43 |
| prometheanfire | set the external callback url and it's still trying with the internal ip :| | 19:44 |
| jrosser | look in the example config for `endpoint_override` | 19:45 |
| prometheanfire | ok, see that now | 19:45 |
| jrosser | in my example the haproxy endpoints are not reachable *at all* from the provisioning network so the override reflects that | 19:46 |
| jrosser | you can set it to whatever you need | 19:46 |
| prometheanfire | yep, that downloaded an image, thanks | 20:01 |
| JayF | Thanks jrosser \o/ congrats prometheanfire | 20:09 |
| prometheanfire | thanks | 20:54 |
| prometheanfire | now I just have to wipe the node since I think the existing software raid is making the deploy to disk sad, that or add some hints | 20:55 |
| opendevreview | Verification of a change to openstack/ironic master failed: Enables boot modes switching with Anaconda deploy for ilo driver https://review.opendev.org/c/openstack/ironic/+/860821 | 21:04 |
| prometheanfire | ya, not sure why, but cleaning, quick, full or express are all not actually cleaning it seems, md127 still found and still messing things up | 22:36 |
| prometheanfire | I tried adding {"interface": "raid","step": "delete_configuration"} in front of the erase_devices step but that wasn't allowed during a clean (agent aborted) | 22:37 |
| jrosser | prometheanfire: is this a previous raid config you need to get rid of? | 22:41 |
| prometheanfire | ya, the host had an OS installed | 22:43 |
| jrosser | ok this was a massive PITA | 22:43 |
| prometheanfire | I've tried manually cleaning | 22:43 |
| prometheanfire | heh | 22:43 |
| jrosser | there might be a better way(!), but.... | 22:43 |
| prometheanfire | easier to erase the hard disks out of band? :P | 22:44 |
| jrosser | disable the raid config `openstack baremetal node set --reset-raid-interface <NODE_NAME>` | 22:44 |
| jrosser | specify a single disk `openstack baremetal node set <NODE_NAME> --property root_device='{"name": "/dev/sda"}'` | 22:44 |
| jrosser | deploy an instance to the node, this will remove the raid signature but fail because of the cached partition table still showing the raid signature | 22:45 |
| jrosser | deploy an instance again, this will now succeed with the single disk | 22:45 |
| prometheanfire | cool, first phase deploying now | 22:46 |
| jrosser | login to the node and `wipefs -a -f <partition>` for all the old partitions | 22:46 |
| JayF | prometheanfire: your feedback is noted; I was worried this would be a side-effect of some of our raid support changes recently to less aggressively delete software raids | 22:46 |
| jrosser | undo the root device `openstack baremetal node unset <NODE_NAME> --property root_device` | 22:46 |
| JayF | prometheanfire: I'll make a note to look into that on Monday, unless jrosser specifically is enabling the code to not-erase-MD-raid in nodes | 22:46 |
| jrosser | put back the raid config you actually want `openstack baremetal node set --raid-interface <whatever>` | 22:47 |
| prometheanfire | should I remove the root_device property after the first phase? iirc the hint shouldn't be needed and is generally discouraged | 22:47 |
| JayF | 3/win 12 | 22:47 |
| JayF | whoops | 22:47 |
| prometheanfire | eh, I can figure out how to add raid later, for now I just want to be able to ssh in :D | 22:47 |
| jrosser | JayF: we also had a terrible time with software raid on nodes that had been previously used for something else that invloved an md setup | 22:47 |
| JayF | Yeah that is not supposed to happen | 22:47 |
| JayF | if you don't have raid configured, and you don't have the option enabled to skip erasing the raid, it should be erased aiui | 22:48 |
| jrosser | those steps i gave were what we came up with the rescue the situation to something usable | 22:48 |
| JayF | and if we somehow changed that behavior I would state it was likely ... misguided | 22:48 |
| prometheanfire | a clean option for all ssd type devices would wipe raid signatures I'd think | 22:48 |
| jrosser | there might totally be something we missed to do that more easilt of course | 22:48 |
| JayF | I've made a note to review this on Monday. My EOD is in about 12 minutes and you don't want a week-old brain looking at this | 22:48 |
| prometheanfire | or nvme, or whatever | 22:48 |
| prometheanfire | yarp | 22:48 |
| prometheanfire | thanks for all the help | 22:48 |
| JayF | prometheanfire: can you do me a favor, and just give me the output of a node show --detail on the imacted node | 22:48 |
| JayF | prometheanfire: assuming this is Zed Ironic/IPA? Or Antelope? | 22:49 |
| prometheanfire | sure, ya, zed | 22:49 |
| JayF | ack | 22:49 |
| prometheanfire | detail not recignized as an argument? | 22:49 |
| jrosser | relatedly we abandoned trying to use software raid because of https://storyboard.openstack.org/#!/story/2010479 | 22:50 |
| prometheanfire | openstack baremetal node show --detail, not using the ironicclient (if that's still even a thing) | 22:50 |
| JayF | prometheanfire: just give me the node show, I don't know what /node/{}/detail translates to in the CLI right now | 22:51 |
| JayF | jrosser: ack | 22:52 |
| JayF | will look at that too | 22:52 |
| prometheanfire | yarp | 22:52 |
| prometheanfire | have to clean it up a little, also note that I did the reset-raid-iterface and set root_device | 22:53 |
| prometheanfire | JayF: https://paste.openstack.org/show/819334 | 22:56 |
| prometheanfire | also, ltns, how's it going? | 22:56 |
| prometheanfire | hmm it's still seeing md127 on the second boot | 22:58 |
| jrosser | prometheanfire: it’s too late here for me to go look but we had some cleaning step config too… | 23:01 |
| prometheanfire | I'll poke at it over the weekend | 23:02 |
| prometheanfire | jrosser: thanks | 23:02 |
| * prometheanfire is seeing what happens when I said root_device to /dev/md127 out of pure curiousity | 23:02 | |
| jrosser | I’ll see if I can dig something out tomorrow | 23:02 |
| opendevreview | Iury Gregory Melo Ferreira proposed openstack/ironic-specs master: Firmware Interface https://review.opendev.org/c/openstack/ironic-specs/+/878505 | 23:35 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!