| clarkb | fungi: TheJulia can probably confirm but I think we're in a good spot for a dib release. I think we'd be doing a 3.34.0 tag since we've changed things like the default centos version to build. Unfortunately I've been bad and haven't dug my gpg key out of cold storage. Any chance you might be willing to do that again | 18:07 |
|---|---|---|
| fungi | sure, no problem. i'll take a quick look at it first | 18:08 |
| TheJulia | I'll take a quick look at open items, but I know my pain points have merged | 18:08 |
| opendevreview | Julia Kreger proposed openstack/diskimage-builder master: docs: add two contextual warnings to the replace-partition element https://review.opendev.org/c/openstack/diskimage-builder/+/928819 | 18:09 |
| TheJulia | https://review.opendev.org/c/openstack/diskimage-builder/+/927781 | 18:09 |
| TheJulia | but that might force 4.0.0 | 18:09 |
| JayF | ++ on doing it as a major version | 18:10 |
| TheJulia | https://review.opendev.org/c/openstack/diskimage-builder/+/923896 has some testing improvements | 18:11 |
| clarkb | the one edit is only for supporting python2 which we haven't in a long time and the other is for python3.1 or 3.0 which barely got any traction. If we don't want to do a 4.0.0 I think we can just tag it 3.34.0 but also versions are basically free | 18:11 |
| JayF | I'm doing a review run through, but tbh I don't remember if I have actual-full-core here or just "babysit-gentoo" core | 18:11 |
| TheJulia | we likley want to workstream https://review.opendev.org/c/openstack/diskimage-builder/+/922352 | 18:12 |
| JayF | yeah I was going to but then realized I shoudl ask ^ | 18:12 |
| clarkb | https://review.opendev.org/c/openstack/diskimage-builder/+/923896 can go in before or after a release without any real impact to users | 18:12 |
| TheJulia | agree | 18:12 |
| clarkb | re MBR you may be surprised, shcoekd even, to know that basically none of the clouds we work with really document boot requirements anywhere and I'm positive at least one of them cannot efi | 18:16 |
| clarkb | (that was sarcasm if it didn't carry through the text) Unfortunately we're likely going to need to support both for a bit. It annoys me that even though gpt boot without efi is supposed to be a thing it isn't anywhere. Thought at the same time its not like things are getting updated for efi so why would they be udpated for gpt | 18:16 |
| clarkb | the good news is we test both so should have decent coverage of them | 18:17 |
| TheJulia | That is really really really unfortunate | 18:17 |
| TheJulia | Since bios booting will fully deprecated on centos 10 (like, expect stuff to get peeled away as it lives) and entirely gone in centos 11 | 18:17 |
| fungi | 25 changes merged (so far) since 3.33.0, https://docs.openstack.org/releasenotes/diskimage-builder/ has a couple of upgrade notes, the consequential one being the centos/rhel 7 removal which might warrant a major version bump but i'd be okay with that as minor | 18:18 |
| TheJulia | That likely needs to be solid feedback that operators need to keep in mind | 18:18 |
| clarkb | but ya it would be cool if gpt was a thing everywhere then you can do bios and efi | 18:18 |
| fungi | chat-gpt is everywhere, does that count? | 18:18 |
| * TheJulia is disrupted by cat finding the cooking pork in the crock pot interesting | 18:18 | |
| clarkb | fungi: ya and I just approved a chagne to rip out python2 and 3.0 and 3.1 support in some places. But again thats minor enough given we haven't done python2 in forever and python3.0 and 3.1 were never really a thing that I'm not sure we need a 4.0 release | 18:18 |
| clarkb | but if others think 4.0 is safest those numbers are basically free | 18:19 |
| fungi | right, for a project whose lifecycle includes rolling additions and removals of support for distro versions, and python interpreter versions for that matter, i would count those things as sufficiently non-surprising as to not warrant signaling major backward-incompatibility | 18:20 |
| TheJulia | clarkb: well, fun thing is you *can* bios boot gpt partitioned disks, since the pointer is to location on disk, not a partition structure data | 18:20 |
| TheJulia | bootloader just needs to have a clue | 18:20 |
| clarkb | TheJulia: supposedly qemu/kvm arm64 implementations explicitly cannot do that even though the spec says they should (not sure if it is a must) | 18:21 |
| TheJulia | clarkb: all 64 bit arm is UEFI | 18:21 |
| clarkb | or maybe I have the logic backwards and it is mbr + efi | 18:21 |
| TheJulia | and thus is not bios booting | 18:21 |
| clarkb | ya I must have things backwards then its mbr + efi that is supposed to exist according to the spec for efi but they only grok gpt | 18:22 |
| TheJulia | yeah, that is semi-common on physical hardware side as well | 18:22 |
| clarkb | re https://review.opendev.org/c/openstack/diskimage-builder/+/922352 this change implicitly drops support for centos that is not stream | 18:22 |
| clarkb | I'm going to -1 it for error handling that is missing, but are we ok with ^ as well? | 18:22 |
| TheJulia | (the lack of groking it even though they could easily JFDI) | 18:22 |
| TheJulia | do those streams of packages even still exist? | 18:23 |
| clarkb | opendev doesn't care because we only build stream images at this point | 18:23 |
| clarkb | TheJulia: I think they do in the archives somewhere let me dig | 18:23 |
| TheJulia | I know centos7 and I think centos8 packages are gonezo | 18:23 |
| TheJulia | ++ | 18:23 |
| fungi | pretty sure the 8 to 8-stream transition happened quite some time back | 18:23 |
| TheJulia | I had someone trying to do centos7 package installs recently and their links were all dead (for good reason) (except epel, which had archive urls) | 18:24 |
| clarkb | https://vault.centos.org/8.5.2111/ | 18:24 |
| TheJulia | ack, makes sense to -1 | 18:24 |
| clarkb | the packages do seem to still exist but you must repoint things to the vault (they are removed from the nomal location) | 18:24 |
| TheJulia | we also don't inherently support that, but folks might try it | 18:24 |
| clarkb | I'm going to post the comment without a +/- and see what others think | 18:26 |
| fungi | i'm on standby ready to push a 3.34.0 once folks are happy that whatever needs to be merged has been | 18:26 |
| fungi | also, ianw tagged 3.33.0 so might want to have some input on what goes into 3.34.0 | 18:27 |
| TheJulia | The train must ship though, fwiw | 18:28 |
| clarkb | fungi: thanks. 923896 and 927781 have been approved but also don't have major impacts on the release. We can definitely see if ianw has input later today | 18:28 |
| clarkb | or we can just go ahead with 3.34.0 as is then decide if we want a 3.35.0 or 4.0.0 in the near future | 18:28 |
| fungi | in the immortal words of esr, release early release often | 18:29 |
| clarkb | after fiddling with bash regexes and [[]] test behavior I think this code actualyl does work since we weren't doing anything special for regular centos | 18:31 |
| clarkb | all of this will simply be skipped over. I'll approve the change | 18:31 |
| clarkb | it turns out "" is not -ge "9" | 18:32 |
| clarkb | I should've expected that because bash | 18:32 |
| clarkb | fungi: 922352 has also been approved if we want to wait for the three changes before tagging | 18:32 |
| fungi | can do | 18:35 |
| clarkb | side note: at the summit the software factory folks were deploying using centos 7 | 18:36 |
| TheJulia | clarkb: heh, it was one of the folks loosely adjacent to that which grumbled in my direction last week about epel7 | 18:58 |
| TheJulia | "why are you doing that!? nooooo" | 18:58 |
| clarkb | those dib changes are failing to pass ci due to dubious git ownership. I'm surprised we haven't fixed all of those at this point | 20:24 |
| fungi | huh, that's been... years? | 20:24 |
| clarkb | I think the most recent update was this year but like 6 months ago | 20:25 |
| clarkb | https://zuul.opendev.org/t/openstack/build/41611afd76ca48b09a54bae007cf4932/log/nodepool/builds/test-image-bc7092e3e9ba41b4827e5352810b1ddb.log#1905-1910 | 20:26 |
| clarkb | that is indeed doing a local clone of the zuul home dir repos into the dib build cache. I'm confused as to why this was never an issue before, but I suspect we can do a simple chown or git config edit in the jobs to address it | 20:27 |
| fungi | file exclusions hid the failure up to now i guess | 20:27 |
| fungi | as in that job never ran | 20:27 |
| clarkb | maybe? I think those jobs run often | 20:28 |
| clarkb | ya most recently about a week ago | 20:28 |
| clarkb | that runs outside of the chroot | 20:33 |
| clarkb | which means it is running in the context of nodepool-builder's container image I think | 20:33 |
| clarkb | do we bind mount in the zuul repo dir to make that work? | 20:34 |
| clarkb | looking at nodepool nothing has changed since august so I don't think we're seeing side effects of a nodepool update | 20:34 |
| clarkb | I'm going to need to do a school run soon so not sure I can really page in everything right now. It might be easiest to debug this on a held node simply because there are so many moving parts, though in theory we've got enough configs and logs to inspect that via the job results | 20:35 |
| clarkb | confirmed we bind mount /home/zuul into the container | 20:36 |
| clarkb | remote: https://review.opendev.org/c/zuul/nodepool/+/929573 Set git repo ownership for nodepool dib integration testing | 20:42 |
| clarkb | this is the "I don't have time to think this through so lets try the most obvious thing and learn what we learn" approach | 20:42 |
| clarkb | oh that change doesn't actually run the source jobs so I need a depends on | 20:47 |
| opendevreview | Clark Boylan proposed openstack/diskimage-builder master: Update functests https://review.opendev.org/c/openstack/diskimage-builder/+/923896 | 20:47 |
| clarkb | I updated that one since it was already testing related | 20:47 |
| ianw | hey no strong opinions on tagging a release | 23:24 |
| ianw | i don't really know if we need to go to 4.0.0 for https://review.opendev.org/c/openstack/diskimage-builder/+/927781 | 23:26 |
| opendevreview | Ian Wienand proposed openstack/diskimage-builder master: Drop logic for old python versions https://review.opendev.org/c/openstack/diskimage-builder/+/927781 | 23:27 |
| ianw | i'm not really sure why that wanted a rebase | 23:28 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!