run: print out how to break out of console on terminal about coreos-assembler HOT 10 CLOSED

Would there be some detection that the QEMU monitor is running before dropping the message?

with coreos-assembler run it is using qemu so we don't have to do any detection. The auto login on serial conolse with coreos-assembler run only happens because we drop in a unit to make it happen so this is very much specific to run. I think with your work we can just also tell ignition to drop in another file and it will give a nice message to the user telling him/her how to break out.

from coreos-assembler.

Though maybe it would be better to change run to work more like Vagrant and inject a ssh key and then use ssh. Would fix tty corruption issues I think.

OTOH, there's something really nice and clean about just wrapping qemu-kvm for this. The way I've been using run is just for quick tests, with anything more "permanent" using my regular libvirt scripts.

from coreos-assembler.

Just tried this out! Looks great!

Fedora 29 (CoreOS preview)
Kernel 4.19.6-300.fc29.x86_64 on an x86_64 (ttyS0)

coreos login: core (automatic login)

Fedora CoreOS (preview)
https://github.com/coreos/fedora-coreos-tracker
WARNING: All aspects subject to change, highly experimental
---

To exit, press Ctrl-A and then X.

from coreos-assembler.

@rfairley - do you know how easy this would be to do? I imagine it would involve us writing a file using ingnition similar to what we do for creating the systemd unit to autologin:

from coreos-assembler.

@dustymabe Creating a systemd dropin unit passed through Ignition seems clean. Would there be some detection that the QEMU monitor is running before dropping the message? If it is detected that the QEMU monitor is running, then the dropin unit would drop a file mentioning Ctrl-A in /run/motd.d/ at startup so sshd can show it. If it is not detected, then no file is dropped.

from coreos-assembler.

Will also give run a go, as I have not tried that out yet. I'll try building an system with the PAM changes and look look into how the detection ^ might work.

from coreos-assembler.

Though maybe it would be better to change run to work more like Vagrant and inject a ssh key and then use ssh. Would fix tty corruption issues I think.

(The question there would be...do we retain the "lifecycle binding" where closing ssh would kill the machine?)

from coreos-assembler.

Though maybe it would be better to change run to work more like Vagrant and inject a ssh key and then use ssh. Would fix tty corruption issues I think.

yeah that could be something we do in the future

(The question there would be...do we retain the "lifecycle binding" where closing ssh would kill the machine?)

I do like the lifecycle binding :)

from coreos-assembler.

Ahh, understood. I'm thinking the file could be dropped in the same [email protected] unit here - it would just be another line of script that the unit runs. Makes sense to go into /run/motd.d/ with it being generated at runtime (vs. /etc/motd.d/).

from coreos-assembler.

thanks @dustymabe for trying it!

from coreos-assembler.