[02:19] <xnox> hm
[06:41] <eoli3n> Hi
[06:41] <eoli3n> i don't know you nickname (raharper ?) but i added required files : https://bugs.launchpad.net/subiquity/+bug/1906829
[06:42] <eoli3n> is there any "clean" workaround for this ?
[13:47] <eoli3n> seriously
[13:48] <eoli3n> i'm getting crazy with this new autoinstaller
[13:48] <eoli3n> each time i test something it is not working...
[13:48] <eoli3n> i spent 2 days on timezone setting
[13:48] <eoli3n> which is ignored
[13:48] <eoli3n> and still not working
[13:49] <eoli3n> https://x0.at/_cG.txt
[13:49] <eoli3n> this is config i'm using and its ignored !
[13:50] <eoli3n> locales not working, timezone not working
[14:00] <eoli3n> i'm spent so many time on this
[14:00] <eoli3n> i'm exausted
[14:00] <eoli3n> i want to finish, each step is a mess
[14:00] <eoli3n> i opened 10 issues
[14:00] <eoli3n> please help
[14:09] <eoli3n> https://bugs.launchpad.net/subiquity/+bug/1907107
[19:10] <eoli3n> mwhudson does post install script which runs at first startup uses the first sudo user ?
[19:19] <mwhudson> eoli3n: which script is that? cloud-init does it's stuff as root
[19:31] <eoli3n> here's my problem
[19:32] <eoli3n> http://sprunge.us/QooXyU
[19:33] <eoli3n> when i comment chroot line with ansible-pull run, then locales and timezone works
[19:33] <eoli3n> but when i run my ansible-playbook it fails. So i need to identify which tasks is breaking post script which setup timezone and locale generation at startup
[19:33] <eoli3n> the playbook has 400 tasks :/
[19:34] <eoli3n> one is that i remove the first sudo user created by the installer
[19:34] <eoli3n> i'm testing piece by piece
[19:34] <eoli3n> but its really long
[19:34] <eoli3n> each iteration needs a new install to run
[19:35] <eoli3n> (i use qemu/kvm with ovmf then i can't use snapshots)
[19:37] <eoli3n> as this is running in chroot late_command could it break something ? (tasks which remove first user)
[20:38] <mwhudson> eoli3n: the late-commands run as root but before first startup
[20:38] <eoli3n> yes i'm not talking about that
[20:39] <eoli3n> lets restart fresh
[20:39] <eoli3n> in late-commands, i run ansible
[20:39] <mwhudson> ok
[20:39] <eoli3n> inside that ansible playbook, there is one tasks which break something
[20:40] <eoli3n> because when i first boot, there are some scripts which finish to setup things right ?
[20:40] <mwhudson> right
[20:40] <eoli3n> those scripts doesnt work because timezone and locales are not working
[20:40] <eoli3n> if i comment out ansible-playbook run in late command, it works
[20:40] <eoli3n> then i need to identify which tasks break those first startup scripts
[20:40] <mwhudson> oh so something your ansible tasks are doing are breaking cloud-init
[20:41] <eoli3n> yes i think so
[20:41] <mwhudson> hmm
[20:41] <eoli3n> one of these tasks is first sudo user remover
[20:41] <mwhudson> well the sudo user is created by cloud-init, it won't be there when you run your tasks
[20:41] <eoli3n> as it runs as late command, it removes that user before first reboot
[20:42] <eoli3n> it may break some latelate scripts ?
[20:42] <eoli3n> oh ok
[20:42] <eoli3n> then i need to search more
[20:42] <mwhudson> actually
[20:42] <eoli3n> where are stored those first startup scripts ?
[20:42] <mwhudson> if you remove the identity section from your autoinstall config you won't get a user created at all
[20:42] <eoli3n> i can do that ??
[20:43] <mwhudson> yes
[20:43] <eoli3n> nice
[20:43] <mwhudson> it's even documented!
[20:43] <eoli3n> hm, i didn't see that it was optionnal
[20:43] <eoli3n> sorry for that
[20:43] <eoli3n> i opened an issue about this
[20:43] <eoli3n> i will close it
[20:43] <mwhudson> https://ubuntu.com/server/docs/install/autoinstall-reference#user-data
[20:43] <mwhudson> i am pretty behind on bug mail :(
[20:44] <eoli3n> i didn't get what you mean
[20:44] <eoli3n> oh !
[20:45] <eoli3n> if i set identity section, will it ignore user-data ?
[20:45] <eoli3n> because i set timezone in user-data
[20:45] <mwhudson> no they get merged
[20:45] <eoli3n> hm ok
[20:45] <mwhudson> (in a fairly unintelligent way, but that shouldn't matter here)
[20:46] <eoli3n> i fire all my questions because we are reverse about sleeping hours
[20:47] <eoli3n> what about multilines ? should "early-commands: |" works ?
[20:47] <mwhudson> no early-commands has to be a list
[20:47] <mwhudson> the entries in the list can be strings or lists though
[20:47] <eoli3n> that's sad
[20:48] <eoli3n> i open an issue about this too
[20:48] <mwhudson> ryan explained the trick to make these more comfortable right?
[20:48] <eoli3n> haharper gave me a curtin way
[20:48] <eoli3n> ah it was ryan
[20:48] <eoli3n> yes he does
[20:48] <eoli3n> he did
[21:37] <eoli3n> ok then i found that there is no problem
[21:37] <eoli3n> its just that scripts are triggered late
[21:37] <eoli3n> 4 min after startup
[21:42] <eoli3n> mwhudson why that behaviour ? why installer doesnt configure everything during install process ?
[22:29] <mwhudson> no especially good reason, it's just easier
[22:29] <mwhudson> 4 mins seems very slow
[22:35] <eoli3n> slow ?? :D
[22:35] <eoli3n> its an eternity
[22:37] <eoli3n> the problem seems more to be about cloud-config service (cloud-init ?) dependencies
[22:37] <eoli3n> it should run before anything else
[22:37] <eoli3n> just after networking.target
[22:37] <eoli3n> i fixed my ansible-pull service, adding cloud-init.target as Wants and After
[22:38] <eoli3n> that service has also Before=lightdm.service
[22:38] <eoli3n> so it should force the order, cloud-init -> ansible-pull -> lightdm
[22:39] <eoli3n> for now i get ansible-pull -> lightdm -> ... -> cloud-init
[22:39] <eoli3n> hope this will not slow down boot process
[22:41] <blackboxsw> driveby sorry, I highlighted on cloud-init. also double check whatever your systemd dependency chain looks like now to make sure as well that something in the image didn't introduce an systemd ordering cycle issue journalctl -b | grep ordering cycle'   as well as checking the system-analyze critical-chain on your units you are manipulating.
[22:42] <blackboxsw> helpful tools to debug `cloud-init analyze blame` and `systemd-analyze blame` to checkout the worst offenders in boot
[22:48] <eoli3n> thanks blackboxsw, it's 23h48 here, lets check it tommorow
[22:48] <eoli3n> i note your commands
[22:49] <blackboxsw> yowsa +1
[22:49] <eoli3n> yowsa ?
[22:49] <blackboxsw> yowsa ==  a very strange way to say "ohh my"
[22:50] <eoli3n> :)
[22:51] <eoli3n> blackboxsw but by default, cloud-init service should run before i get tty1 login ?
[23:03] <blackboxsw> honestly not certain. but systemd-analyze critical-chain says system-logind waits on cloud-init completion https://paste.ubuntu.com/p/YQC9BSzQzr/
[23:05] <eoli3n> strange, because, even with a "classic" autoinstall, after first reboot, i get tty1 login, then cloud-init runs and override prompt
[23:05] <eoli3n> that's dirty
[23:08] <eoli3n> lets sleep a bit, i on it since 7am without any stop, its 0am