02:43 <baaash[m]> hey team, I'm currently learning the ways of fedora, and cutting my teeth on coreos here... i'm having a bit of trouble getting my disks partitions setup right using butane config. I must me misinterpreting the docs somehow. Can someone provide a basic butane disks config for creating /boot /root and /data partitions on /dev/sda for me please so i can compare?

02:57 <baaash[m]> ah ok i've just found a little further down in the examples that a '0' value in the size_mib key means to use all available space... getting warmer...

03:01 <baaash[m]> ok i guess this isn't a very active channel for this kind of quick support stuff.. reading the spec it is.

03:03 <baaash[m]> aha.... looks like number doesn't mean what i assumed it to.

07:24 <dngray[m]> does ignite support the ssh_host_xxxx keys for each VM?

07:24 <dngray[m]> i kind of don't want them to change when i reboot the VM

07:24 <dngray[m]> would be nice to be able to provision /etc/ssh/ssh_host_ed25519_key

07:26 <dngray[m]> seems to be the only thing missing from https://docs.fedoraproject.org/en-US/fedora-coreos/tutorial-containers/

07:28 <dngray[m]> okay so seems storage is the only way https://github.com/coreos/butane/issues/210#issuecomment-793412228

09:20 <travier[m]> baaash[m]: What are you trying to do? The default setup will get you a working installation. You only need to provide a config for storage if you want to change that.

09:23 <travier[m]> dngray: The host keys do not change when you reboot a system, unless your systems are ephemeral

09:27 <dngray[m]> yeah i used the storage api, to load the same ones, because my fcos vm loads from a pxe server on boot (has no disk)

12:41 <dngray[m]> https://serverfault.com/questions/763805/how-to-place-docker-images-ontop-of-an-nfs-share-in-coreos ahh

12:41 <dngray[m]> so using NFS to mount my shares from my host in my guest VM, which runs fcos

12:45 <dngray[m]> wait should be possible, fcos is a KVM, so should be able to mount the NFS share in that VM and then start docker

12:47 <dustymabe> dngray[m]: just embed your host ssh keys in your Ignition config and they should get written out on boot

12:47 <dngray[m]> dustymabe: that's what i did with success

12:47 <dngray[m]> now i'm contemplating what problems i will have passing /var/lib/docker in from my host

12:47 <dngray[m]> the fcos vm is ephemeral, so no real storage

12:48 <dngray[m]> loads from iPXE on each boot

12:49 <dustymabe> i'd think it would just work, but I haven't tried it

12:49 <dngray[m]> i think i might have issues because of SELinux though.

12:51 <dustymabe> maybe.. you might need to just label the file appropriately before you start the VM

12:51 <dustymabe> files*

12:51 <dustymabe> you can do a proof of concept with selinux in permissive mod efirst

12:57 <dngray[m]> suppose that would work, but does SELinux even work over NFS?

12:58 <dngray[m]> `mount -t nfs -o vers=4 192.168.2.253:/var/lib/docker /var/lib/docker`

12:58 <dngray[m]> hmm, when i mount t hat

12:58 <dngray[m]> i can't even ls in /var/lib/docker

12:59 <dngray[m]> its owned by root:root on the host

12:59 <dngray[m]> actually, that's a bad idea.

12:59 <dngray[m]> i should create a "new" docker share

13:00 <dngray[m]> because the host which ran that, actually had native ZFS support

13:00 <dngray[m]> now ill be running docker as file support, the intention is to migrate over to podman

13:02 <dustymabe> dngray[m]: let us know how it goes!

13:26 <dngray[m]> is it my imagination or docker-compose is falling out of fashion

13:26 <dngray[m]> i guess because people have other ways of building

13:27 <dngray[m]> i'm aware of podman-compose, but i had some issues with it

13:28 <dngray[m]> `docker run -it --rm busybox` guess we'll see if it works

13:32 <dngray[m]> Filesystem Size Used Avail Use% Mounted on

13:32 <dngray[m]> 192.168.2.253:/var/lib/docker 38T 0 38T 0% /var/lib/docker

13:32 <dngray[m]> hmm no such luck

13:33 <dngray[m]> i disabled selinux first with `setenforce 0` so that's not the issue

13:34 <dngray[m]> oh, maybe cos it was owned as root:root the dir on the host

13:57 <dngray[m]> hmm, i think running my docker containers from NFS is going to be a massive pain in the ass

13:57 <dngray[m]> maybe i should just add a storage device lol

14:07 <dngray[m]> dustymabe: heh i see you in https://github.com/coreos/fedora-coreos-tracker/issues/1131

14:07 <dngray[m]> seems i can only find issues with people trying to mount NFS shares inside docker/podman

14:07 <dngray[m]> not run the actual containers from a share

16:05 <dustymabe> saqali: can we close your issues here? https://github.com/coreos/fedora-coreos-streams/issues

16:05 <saqali> yes I'll do that

19:15 <justJanne> hey, so, stupid question, but if I configure the file /etc/containerd/config.toml via butane/ignition (even if I keep the content and mode the same), fcos refuses to boot. Is that file managed by rpm-ostree or something?

19:15 <justJanne> And if so, how am I supposed to configure the CNI directory?

20:08 <walters> justJanne: It sounds like you're getting a fatal error in Ignition; you should try to figure out what that is...hmm, how do we not have a doc page for this? Basically you should see an error on the console

20:10 <justJanne> walters: I don't have access to the serial console if ignition fails. I've only got ssh access to the server (serial console access requires additional hardware, at a per-hour price point, so I'd like to avoid that)

20:11 <walters> ok yeah, https://github.com/coreos/ignition/issues/585 would help your case too

20:12 <walters> justJanne: at an offhand guess, do you need `overwrite: true` in your config?

20:12 <walters> one thing that can be helpful is to test your configs in VMs before applying them to bare metal

20:13 <walters> hmmm actually after https://github.com/coreos/ignition/pull/1285 we could make it convenient to test the filesystem section in a container too (or of course after https://github.com/coreos/enhancements/blob/main/os/coreos-layering.md actually ship them in containers)

20:32 <justJanne> walters: I'll try that, thanks! Sorry, I'm very lost in modern-day ignition configs

20:32 <justJanne> Still too used to the pre-ignition coreos

20:36 <justJanne> Back when fleetctl was new and cool

21:03 <justJanne> walters: you’re super helpful, so I’ll ask one more question: what’s the easiest way to enable and start an installed, but disabled, systemd unit?

21:04 <justJanne> just add it to units, with enabled: true but without contents?

21:19 <walters> justJanne: Honestly I'm not sure, but I think that'd work...(ok at a quick glance at the source code, particularly `writeSystemdUnit` looks like it would indeed work

21:29 <bgilbert> justJanne: yes, that's right

22:46 <baaash[m]> hey team so i'm adding the line: "coreos.inst.ignition_url=http:192.168.1.1:8000/config.ign" to the kernel command line options when installing a coreos vm and this address resolves to apache running in a container on my localhost. I'm using the podman0 bridge for the vm network, the same that the apache container is using. Installer won't acknowledge config exists, despite it resolving fine. it appears no networking devices are

22:46 <baaash[m]> brought up after the installer has completed. I'm obviously doing something incorrect here.. any ideas?