| opendevreview | Will Szumski proposed openstack/diskimage-builder master: Remove vm element from rocky-container image https://review.opendev.org/c/openstack/diskimage-builder/+/924421 | 16:03 |
|---|---|---|
| opendevreview | Will Szumski proposed openstack/diskimage-builder master: Upgrade curl-minimal for RHEL based images built from containers https://review.opendev.org/c/openstack/diskimage-builder/+/924424 | 16:28 |
| clarkb | TheJulia: can you think of any reason https://review.opendev.org/c/openstack/diskimage-builder/+/923815 might be problematic? I think on the opendev and zuul side of things its fine, but I didn't want to single core approve it (that is why the -2 is there I accidentally approved it and then removed my +A causing the failure to merge and a -2) | 17:14 |
| JayF | looking | 17:15 |
| JayF | +2, not who you asked for but hopefully that's what you wanted :) | 17:16 |
| clarkb | ya mostly just a sanity check that installing podman won't create unwanted side effects | 17:17 |
| clarkb | I mean we're already installing a bunch of package manager systems and qemu image tooling, but container runtimes can sometimes be a bit more invasive | 17:18 |
| TheJulia | I think it is likely fine | 17:18 |
| clarkb | though I think podman is a bit less so than say docker | 17:18 |
| TheJulia | the exclusions on that list are slightly alarming but it makes sense | 17:18 |
| clarkb | TheJulia: JayF yall may know the answer to my question in https://review.opendev.org/c/openstack/diskimage-builder/+/924421 too | 17:33 |
| JayF | I really don't know, I don't use DIB much except for VM-specific images or baremetal ramdisks. Most places I've been that use openstack got their images separately | 17:34 |
| TheJulia | so... hmmm | 17:36 |
| TheJulia | well, VM is a horribly named element | 17:36 |
| clarkb | yes no argument there | 17:36 |
| TheJulia | and it just extracts the kernel/ramdisk for partition image usage | 17:36 |
| TheJulia | baremetal *just* does some cleanup and invokes bootloader | 17:36 |
| clarkb | baremetal does ^ yup. I really don't think teh two elements are at odds with each other | 17:36 |
| TheJulia | so removal of vm seems logical | 17:37 |
| TheJulia | IPA has it's own extraction stuffs if memory serves | 17:37 |
| clarkb | right removing vm is fine as its over prescriptive. My concern is that if we land that commit as is it tells everyone they can't use the two elements to gether but as far as I can tell you can | 17:37 |
| clarkb | and there really isn't any harm to doing so | 17:37 |
| TheJulia | yeah, they really are not, but we should really stamp out the VM element anyhow | 17:37 |
| clarkb | uh no | 17:37 |
| clarkb | the vm element is important for other use cases | 17:37 |
| clarkb | we might want to rename it | 17:38 |
| clarkb | or alias it or something that doesn't brea kthe world | 17:38 |
| clarkb | the disk partitioning is important for being bootable in clouds iirc | 17:38 |
| clarkb | *in some clouds. In others you may upload fs and kernel separately | 17:38 |
| TheJulia | ummmmmm | 17:38 |
| TheJulia | that is not the purpose behind vm | 17:39 |
| TheJulia | i mean, removal overall, will break folks | 17:39 |
| TheJulia | oh | 17:39 |
| clarkb | the readme says Sets up a partitioned disk (rather than building just one filesystem with no partition table). | 17:39 |
| TheJulia | so vm now just removes udev links | 17:39 |
| TheJulia | and calls block-device and bootloader | 17:40 |
| clarkb | oh indeed | 17:40 |
| TheJulia | what a mess | 17:40 |
| clarkb | oh right block-device is what creates the partition table | 17:40 |
| TheJulia | yeah | 17:40 |
| clarkb | you get a heuristic setup by default but you can pipe in your own map | 17:40 |
| TheJulia | ... so... | 17:40 |
| TheJulia | ugh | 17:40 |
| clarkb | but there shouldn't be anything that prevents you from using both the baremetal and vm elements together | 17:41 |
| TheJulia | overall, it is the one filesystem/partition image case which is kind of disappearing | 17:41 |
| clarkb | if you do you'll get a complete disk image with partition table as well as the kernel artifacts | 17:41 |
| clarkb | no it isn't.... | 17:41 |
| clarkb | the vast majority of lcouds that we work with continue to boot thay awy | 17:41 |
| clarkb | uefi support has been very slow uptake | 17:41 |
| TheJulia | for VMs it has | 17:42 |
| TheJulia | if you have a bootable image overall, with proper EFI setup, the kernel/initrd artifact is useless | 17:42 |
| clarkb | yes, but that is the baremtal element | 17:42 |
| clarkb | anyway thats all orthogonal. I just don't want to encode that baremetal and vm elements are not compatibile with one another. They are as far as I can tell. | 17:42 |
| TheJulia | yeah, the baremetal element is what I was thinking | 17:43 |
| TheJulia | that one can be tossed, as far as I'm concerned, in the recycling bin | 17:43 |
| TheJulia | bit recycling bin at that | 17:43 |
| clarkb | heh | 17:43 |
| TheJulia | maybe rename "partition_image" | 17:43 |
| TheJulia | yeah, they are afaik compatible | 17:43 |
| clarkb | ok so maybe the fixup in that change should be to rewrite teh commit message to something like "Remove the vm element from rocky as some users may want a raw filesystem image instead of a complete image" | 17:44 |
| TheJulia | just baremetal shouldn't really be used | 17:44 |
| TheJulia | At least, in modern scenarios | 17:44 |
| clarkb | then separately we can think about aliasing/renaming vm to something like "partition_image" | 17:44 |
| TheJulia | clarkb: I think that is likely the right path, getting at the root context of the change | 17:44 |
| clarkb | thanks I'll respond to myself and ask for that more specific update and call out the idea for improvement to the naming scheme | 17:46 |
| TheJulia | ++ | 17:46 |
| TheJulia | Sorry for the confusion | 17:46 |
| clarkb | I don't think you're to blame for that. The confusion started a decade ago with some unfortunate naming ;) | 17:49 |
| TheJulia | Oh, the person who named it orinally has already admitted blame :) | 17:49 |
| clarkb | NeilHanlon: do you have a quick moment to look at https://review.opendev.org/c/openstack/diskimage-builder/+/924424 ? I'm not sure I understand all the nuance there and I seem to recall you dealt with something similar at one time. Otherwise teh change looks fine just want to amke sure there isn't some sublte behavior I'm missing | 18:08 |
| opendevreview | Merged openstack/diskimage-builder master: Fix regexp for Fedora element SUBRELEASE parsing https://review.opendev.org/c/openstack/diskimage-builder/+/922103 | 19:12 |
| opendevreview | Merged openstack/diskimage-builder master: Add podman to bindep https://review.opendev.org/c/openstack/diskimage-builder/+/923815 | 19:31 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!