12:55 <fifofonix> dustymabe: chronyd worked fine on newly provisioned test-devel (thought I'd responded yesterday - sorry)

13:29 <dustymabe> fifofonix: thanks! If you want to respond in the issue you opened and say you tested the fix and that it worked, that would be good information to have

13:32 <fifofonix> dustymabe: done

13:38 <dustymabe> fifofonix++

13:55 <dustymabe> jlebon: bgilbert: I have a question about https://github.com/coreos/coreos-assembler/pull/3409 - before that merged I was able to launch older FCOS versions in the upgrade test, but after that merged I can't (they complain about a ignition config version that's not supported). Before I was able to successfully down-version the configs by specifying on old butane spec version

13:55 <dustymabe> (https://github.com/coreos/fedora-coreos-config/blob/2b0fe09b75c7200b538724be97081f035085164d/tests/kola/upgrade/extended/config.bu#L2-L4)

13:57 <dustymabe> in the staging pipeline yesterday (where I was testing out the upgrade test) this manifested by instances entering emergency.target in the initramfs

13:58 <dustymabe> when I try to repro this locally I see: No translator exists for variant fcos with version 1.5.0

13:58 <dustymabe> https://paste.centos.org/view/5699af0d

14:07 <jlebon> dustymabe: i'll let bgilbert answer that one :)

14:46 <dustymabe> jlebon: NVM - I was accidentally using an old `kola` binary that was in my local dev directory

14:47 <dustymabe> the test is working now

15:40 <millerthegorilla> I am writing an ansible script to provision a rpi4 with coreos.  I am currently using coreos-installer, but for portability I wondered if there is a script that can be used by machines that are running distros' other than Fedora?  I have seen this, but it looks quite old...

15:40 <millerthegorilla> https://raw.githubusercontent.com/coreos/init/master/bin/coreos-install

15:42 <millerthegorilla> is there a container with coreos-installer (looks..)

15:46 <millerthegorilla> no official container, but I guess it should be easy to make one, as long as its possible to mount the sd card...

17:18 <javierm> travier[m], walters: related to https://github.com/coreos/bootupd/issues/440, something that I was thinking was that bootupd could do an atomic rename exchange of dirs in the ESP

17:20 <javierm> that is, create a new directory that's a copy of the ESP, apply the diff to that and then do a renameat2(..., RENAME_EXCHANGE)

17:21 <javierm> I see that openat::Dir already has that support https://docs.rs/openat/latest/openat/struct.Dir.html#method.local_exchange

17:26 <dustymabe> millerthegorilla: https://docs.fedoraproject.org/en-US/fedora-coreos/bare-metal/#_installing_from_the_container

17:27 <millerthegorilla> dustymabe many thanks, precisely the thang... cheers!

17:37 <walters> javierm: worth commenting on the issue, I think it'd help, though the obvious cost is without reflinks it incurs temporary storage doubling

17:37 <walters> or really even sym or hard links...

17:38 <walters> why did they have to pick FAT 😢

17:43 <javierm> walters: yeah, but it's just temporary and after all the shim, fwdup, grub, etc binaries are not that big

17:44 <javierm> walters: because UEFI is really MS-DOS in disguise :)

17:56 <javierm> walters, travier[m]: I've filed as a separate issue because I think both are complementary https://github.com/coreos/bootupd/issues/454

18:05 <millerthegorilla> I am using coreos-installer that is installed on fedora silverblue by rpm-ostree, with the script at https://docs.fedoraproject.org/en-US/fedora-coreos/provisioning-raspberry-pi4/#_edk2_combined_fedora_coreos_edk2_firmware_disk  but everytime I run it, it works perfectly but along the way it produces an error message, which must emanate from the

18:05 <millerthegorilla> coreos-installer itself,    ---  mount: /tmp/FCOSEFIpart: can't find in /etc/fstab.  ---  this then leaves a non-zero exit code, which bothers my ansible...  should I open an issue at coreos-installer github?

18:28 <dustymabe> millerthegorilla: the instructions say `sudo mount $FCOSEFIPARTITION /tmp/FCOSEFIpart`

18:29 <dustymabe> are you setting $FCOSEFIPARTITION ?

18:54 <millerthegorilla> this is the correct link for the script, sorry.  https://docs.fedoraproject.org/en-US/fedora-coreos/provisioning-raspberry-pi4/#_edk2_combined_disk_mode_alternate_machine_disk_preparation

18:56 <millerthegorilla> The line with fcosefipartition should be correct, I will have to test it - it takes about 3.5 minutes to run.  back soon.

19:03 <millerthegorilla> no, fcosefipartition is correctly set to /dev/sda2

19:16 <dustymabe> millerthegorilla: the error you posted `mount: /tmp/FCOSEFIpart: can't find in /etc/fstab` only happens when you only provide one argument to mount

19:16 <dustymabe> which is why I asked if the variable was set to some value

19:16 <dustymabe> if you manually run the steps it should work just fine (I just executed the steps earlier this week as part of FCOS 38 test week)

19:17 <dustymabe> so the question is:

19:17 <dustymabe> 1. if you manually run the steps (outside of ansible) does it work fine?

19:17 <dustymabe> 2. if the answer to 1. is yes, then the bug is in your ansible script, correct?

22:21 <millerthegorilla> dustymabe I found the problem.  Coreosinstaller is not releasing the disk until after it has run, so unless I put a sleep command after the coreos-installer line, the partitions couldn't be found.

22:22 <millerthegorilla> it was certainly working previously, and I can't be certain that it is something else that I have done.