=== juergh is now known as Guest58246 [08:34] Hi [08:35] is there any plan to let desktop image use autoinstall ? [08:45] as you removed kickstart support since 20.04, what is the recommended way to automate desktops desployments ? [08:46] deployments* [08:57] what is the nickname of Ryan Harper here please ? [12:15] eoli3n_: i'm not sure what you mean by kickstart support being removed. [12:15] eoli3n_: desktop installation automation has not been changed for years, and can be done in 20.04 the same way it has always been possible to do. [12:15] actually i was wrong, just reworking on this. Problem is not that kickstart is removed, problem is that i can't deploy in EFI mode with kickstart [12:16] eoli3n_: what's not working for you? [12:16] i deploy ubuntu desktop with EFI always. [12:16] with kickstart ? [12:16] yes [12:16] hm [12:16] eoli3n_: your partitioning does have ESP right? [12:16] one kind of needs that =) [12:17] how do you trigger EFI deploy mode with kickstart ? [12:17] by booting the machine in EFI mode === jamesh_ is now known as jamesh [12:17] with either EFI iPXE or EFI Http boot [12:17] yes, but when i tested, my booted EFI host deploy legacy system, but maybe i did forget to adapt my partition table [12:24] xnox what iso do you use ? [12:24] the main desktop one ? [12:25] i used the netboot one, which can't deploy in efi mode [12:25] eoli3n_: yes i always use the desktop one. [12:26] and i network boot initrd & kernel from it, using shim + network build of grub grubnetx64.efi. And pass url=http://path/to/matching-desktop.iso [12:26] eoli3n_: i only always deploy desktop using the desktop.iso with preseeding. [12:27] that's the only way to deploy Ubuntu Desktop the right way. [12:27] any link which describe the process ? [12:29] xnox is there any chance to adapt it to nfsroot the iso ? [12:30] i manage 800 nodes, deployed by group of 40 [12:30] i can't let each node get a full 2G iso [12:31] eoli3n_: "i can't let each node get a full 2G iso" => but they will need to get 2G of data one way or another to complete desktop install. [12:31] so not sure it matters _when_ they do that =) from initrd on boot, or elsehow. [12:31] But there is nfs root mount option too, if you can export nfs mount. [12:31] https://discourse.ubuntu.com/t/netbooting-the-live-server-installer/14510 => is how to setup pxeboot, it does work with UEFI too if your UEFI supports pxe (see like the first comment). Instead of the live-server.iso you just use the desktop one. [12:34] xnox nfsroot is much more efficient [12:34] because of the server itself [12:35] 40x dl of a big file is worse than letting clients get a lots of little files [12:35] boot=casper netboot=nfs nfsroot=192.168.122.73:/srv/iso-mount/ => make sure that you export the full mounted desktop iso; and that initrd&vmlinuz you boot are from it. [12:35] if you can't mount the iso make sure to copy _all_ of it, including hidden files. [12:36] there are some details on automating the preseed of Ubiquity https://wiki.ubuntu.com/UbiquityAutomation [12:36] i.e. booting with automatic-ubiquity and supplying the preseed. [12:37] i.e. cp -r /mnt/. /srv/nfs/somewhere => note the '.' in /mnt/. to ensure all the hidden files are copied and exported. [12:37] i already does this [12:38] do* [12:41] eoli3n_: cool. so yeah, my preference is to boot in uefi, either download/mount iso from the initrd (url=http://...) or do nfsroot mount, and automate that. There are no reasons not to boot in EFI mode. [12:42] eoli3n_: if you don't want to use PXE / tftp, you can point your servers to perfrom UEFI HTTP boot => point them at shim + grubnetx64.efi [12:42] but you use kickstart or ubiquity only ? [12:42] i use ipxe and tftp [12:42] eoli3n_: there is no native support for kickstart, and never was. it always translated kickstart into d-i/ubiquity compatible preseed. [12:43] eoli3n_: thus i prefer to use native formats, thus yes, i simply write/serve/boot with preseed for ubiquity. [12:43] i know, but i already have a kickstart script for legacy mode [12:43] or autoinstall.yml for the subiquity live-server iso. [12:43] xnox oula oula oula oula [12:43] lets restart fresh, to let you understand from where i come, to where i want to go, and where i'm now [12:43] eoli3n_: kickstart should work too. I can't remember if we build preseed-kickstart into the ubiquity images. [12:44] let me check that. [12:44] eoli3n_: you also did say taht your kickstart doesn't work, rigth? =) in EFI... hence if it doesn't work, is it really there? [12:45] i used ubuntu ""kickstart"" to deploy legacy hosts with pxelinux/tftp. Now i need to deploy in EFI mode so i worked a lot on server iso to use autoinstall, i'm really near to get it work [12:45] but i wonder if its the right way [12:46] now i don't have much time so i need to choose a quick and proper way [12:46] there is no good documentation about this [12:46] that's sad [12:47] a kind of guide [12:47] eoli3n_: when you say you used ubuntu ""kickstart"" can you show me what you mean? [12:47] yes, just a sec [12:47] just a few snippets / quotes of it. [12:48] eoli3n_: your request is not unique. I am pondering to run a public demo netboot/install server that "works" and then let people copy it. [12:48] xnox https://x0.at/UWZ.txt [12:48] i need to deal with dualboot [12:49] eoli3n_: the only bad thing about your preseed is the part of "generate partitioning script" [12:49] i can't do without it [12:49] because of dualbooting [12:49] eoli3n_: if you change the sfdisk < it will start to work in EFI mode too. [12:50] ahhh [12:50] ok [12:50] eoli3n_: you need to adjust that one, to create ESP [12:50] that was my question [12:50] eoli3n_: that's what i did tell you at the very start of our conversation =) [12:50] "xnox> eoli3n_: your partitioning does have ESP right?" [12:50] yep yep [12:51] if you have no ESP in your partitioning, when installer gets to "now i install shim+grub into ESP, it goes oh oh" [12:51] eoli3n_: and it's best to always create ESP and always use GPT => it works with legacy mode, efi mode, dual boot, with any windows&linux combo, etc. [12:51] ok [12:52] even if the deploy is done in bios or efi mode, it will installs and boot in either. [12:52] our ubuntu desktop is versatile like that, and allows to boot itself irrespective of how one installed it. [12:52] so lets try to adapt this instead of working with autoinstall [12:53] xnox http://archive.ubuntu.com/ubuntu/dists/bionic/main/installer-amd64/current/images/netboot/netboot.tar.gz [12:53] is this able to deploy in EFI mode ? [12:54] eoli3n_: it's incomplete. and does not support secureboot. [12:54] i don't use secureboot [12:54] what do you mean by "incomplete" ? [12:55] eoli3n_: you'd need to add shimx64.efi / uefi-ipxe / grubnetx64.efi to get secureboot (shim), uefi netboot capability (ipxe or httpboot grubnetx64.efi from focal+) [12:55] eoli3n_: one needs more than just that, to boot in efi mode is what i mean. [12:56] if i extract kernel+initrd and boot it with ipxe, no problem then, right ? [12:56] yeah, that works. [12:56] some expect the shim/signed uefi ipxe/grub/configs to be in that tarball too. [12:56] so in that case, what is the benefit to prefer the main desktop.iso ? [12:57] eoli3n_: that thing, will never deploy the same desktop as the desktop.iso will. [12:57] lets precise something [12:57] i just need the installer to deploy the minimal system [12:57] eoli3n_: no option to install minimum ubuntu desktop; no option to install correct matching nvidia drivers; no way to configure and install OEM kernel for the certified hardware [12:57] then i use ansible to deploy everything else [12:57] ansible does all those things for me [12:57] no option to have the correct defautl snaps installed; following the correct tracks [12:58] no oem config... [12:58] hm [12:58] eoli3n_: some/none of the above maybe of interest to you [12:58] eoli3n_: and/or you may already be fixing those things separately. [12:58] but then will not be any 22.04 netboot so [12:58] lets switch to the main desktop.iso [12:58] eoli3n_: i also don't see ubiquity supporting kickstarts. and there is new installer in progress which will have autoinstall.yml support just like server iso [12:58] thanks for your help xnox, i was a bit lost [12:59] "there is new installer in progress which will have autoinstall.yml" [12:59] i need this ! [12:59] eoli3n_: swithcing to the main desktop.iso / and or testing that, may help to futureproof your deployements. [12:59] ok [12:59] eoli3n_: see the New Installer articles / github repo where it is being developed. At the moment early design stages. [12:59] so two moves : move to desktop.iso, and translate the kickstart part to ubiquity script [13:00] any link about this ? [13:00] eoli3n_: yes, if you boot / deploy your kickstart, there should be somewhere in /var/log/installer the translation. Cause d-i does translates it into preseed. It should be a good enough starting point. [13:01] eoli3n_: plus you don't actually do anything special much. [13:01] you pick the right disk; and configure ansible [13:02] https://code.launchpad.net/ubuntu-test-cases => we have many preseeds automation for various things [13:02] does ubiquity automation only, support %post chrooted script and %pre ? [13:02] eoli3n_: https://bazaar.launchpad.net/~ubuntu-test-case-dev/ubuntu-test-cases/desktop/view/head:/preseeds/default.cfg is more or less all you need. [13:02] but you do need preseed/late_command to do more things kind of like what your %post does to do ansible [13:03] and i think you will want to do a hook before partma to figure out the right drive to autopartition with a shell script [13:03] cause i think you are trying to iterate things and find the one you like. [13:03] for that you can use partman/early_command [13:04] eoli3n_: or check what your kickstart got translated into preseed terms. [13:04] i am sorry, i have to go, but i will come back to read backscroll here. [13:06] thanks a lot for your help === mfo_ is now known as mfo [13:07] cjwatson: any further thoughts on my git-ubuntu rich history thread please? I think, with your feedback in mind, it still makes sense to implement my proposed spec (but with a singular Vcs-Git-Ref), with a view to switching to "push" at some far point in the future. Do you concur? [13:10] rbasak: Since you're treating the rich history as best-effort, I think I concur for now, yes. [13:11] OK, thanks! And for your detailed feedback. [13:11] rbasak: I do think you should add at least some kind of fallback-and-retry for temporary failures even in the short term - they won't be *that* uncommon [13:12] rbasak: Even if it's just the kind of fairly basic thing that lazr.restfulclient does (I mean in terms of overall algorithm, rather than specifics) [13:13] Understood. [13:14] I can add some short-lived retries that will hold up the importer worker (and therefore other imports) for an hour or two. [13:15] Leaving the import to move on to others and then come back to it is harder, since I don't currently have a soft-fail mechanism in the queue architecture. [13:15] That's something I should probably add, but isn't there yet. [13:33] rbasak: Seems reasonable. Most such temporary failures are likely to be service-wide anyway [14:26] xnox any chance to get my ks file to work as is ? [14:26] i found the translation in /var/lib/preseed/log, but it doesnt contains everything [14:40] "xnox: eoli3n_: i also don't see ubiquity supporting kickstarts." [14:41] hm.. ok [14:55] xnox some questions : [14:56] - how to generate my partman-pre script [14:56] - how to run multiple command in chroot after install [14:56] https://help.ubuntu.com/lts/installation-guide/armhf/apbs05.html#preseed-hooks [14:56] this is really unclear [15:02] found this about multiple command in chroot : https://github.com/dsgnr/Ubuntu-16.04-Unattended-Install/blob/master/preseed.cfg#L195 [15:12] eoli3n_: yeah ';' works for multiple commands. not nice, but it's what it is. [15:12] "how to generate my partman-pre script" => not sure what you mean. [15:13] https://x0.at/UWZ.txt [15:13] see the # PRE # part [15:13] eoli3n_: if you want to control everything by hand, something like this works => https://secopsmonkey.com/custom-partioning-using-preseed.html [15:13] which disables all the automatic partitioning, and does it by hand. [15:13] i do want automatic partitionning [15:14] i just want to force to restore part1 which is windows [15:14] my pre part is to not overwrite windows if present [15:14] if not present, it prepare the place [15:14] that's why i dump diskid [15:14] and regenerate as is [15:15] it actually clear everything except windows part [15:15] clears [15:15] can i wget that script ? [15:17] d-i partman/early_command string wget -O /tmp/pre.sh http://prodpeda-cesium.infra.domain.fr/preseed/fds/pre.sh; /bin/bash /tmp/pre.sh [15:17] eoli3n_: so you either have empty disk => wipe everything; or you have windows on part1. [15:17] lets try this [15:17] xnox i don't get your question [15:17] i do have both [15:17] eoli3n_: when you have part1 => do you have enough disk space after part1? or do you need to resize windows too? [15:17] if this is a new host, first use, then prepare part1, else if windows present, keep it functionnal [15:18] i have enough space [15:18] eoli3n_: cause i wonder if you can just select use largest free, and it will always do the right thing. [15:18] that's what i use [15:19] preseed partman-auto/init_automatically_partition select biggest_free [15:19] hm [15:19] oh, so you create empty part1, such that later you can have windows on it? [15:19] interesting. [15:19] better than that [15:19] or if it is there already, don't touch it? [15:19] if windows is present, is "recreate" over it to keep it intact [15:19] yes [15:19] interesting. [15:20] do we currently break windows? =/ [15:20] that was the only way i found [15:20] cause i would have hoped that biggest_free doesn't break/touch windows that is there anyway. [15:20] i had some issues yes [15:20] ah, netboot.tar breaks things yes. [15:20] windows needs a repair if disk_id changed [15:20] ubiquity doesn't => it has more logic. [15:20] oh ! [15:21] lets try without then [15:21] eoli3n_: ah, interesting. maybe we still break disk_id, but i thought we didn't. [15:21] but with GPT partition table, maybe its different [15:21] eoli3n_: it would be interesting to try without anything. Imho, it is a bug if it breaks windows and doesn't just use biggest_free. [15:22] problem is that i fix two cases [15:22] true.... [15:22] first is windows is present : so without anything it "would" work, as you say [15:22] but if nothing is already present, the place will not be prepared if i drop my script [15:22] hopefully doing "d-i partman/early_command string wget -O /tmp/pre.sh http://prodpeda-cesium.infra.domain.fr/preseed/fds/pre.sh; /bin/bash /tmp/pre.sh" will just work as you mentioned earlier. [15:22] ok [15:23] maybe i can use sfdisk in another way [15:23] to just create the part without restauring the full partition table [15:23] in 2 steps [15:23] restoring [15:30] running [15:30] just a thing : how to disable the filesystem check at startup ? [15:32] that one : https://x0.at/xsf.png [15:34] eoli3n_: it's turned off by default in hirsute; i think you can boot with fsck.mode=skip and it will not run. [15:34] what is hirsute ? [15:34] lets try "fsck.mode=skip" [15:34] eoli3n_: Ubuntu 21.04 => development version of the next Ubuntu release. [15:35] oh ok [15:38] fsck.mode=skip doesnt disable the filesystem check [15:38] https://x0.at/zL_.txt [15:45] if you have more exemples of seed files, i take [15:51] (filesystem check seems to still print but does not run (stuck at 0% then disapear)) [15:52] xnox, what's the state of https://bileto.ubuntu.com/#/ticket/4491, are you still working on it? [16:09] oSoMoN: that approach failed, but yes still working on it. [16:09] ack, thanks [16:10] eoli3n_: stuck at 0% then disappear => is not running =) the visual UX improved later. === ijohnson is now known as ijohnson|lunch === ijohnson|lunch is now known as ijohnson