rebooting is only a temp fix.

at least for me

Did you dig into this any more @ambitious-knife-27333? Were you ever able to grab logs of the serial port write failures? I had a ticket open with Suse but they're giving me crap saying it's working in OpenSuse so tell Alma.

Actually it seems like 9.3 is the only one effected? Ugh. we'll see

But I downloaded the 9.4 Alma image from a few days ago, and I can't get it to reproduce currently.

our older 9.3 image is easy to replicate. (takes less than 5 minutes)

I thought it might be related to the qemu-guest-agent version (8.2.2) but Fedora 40 and Ubuntu 24.04 LTS are on that version too, and I can't replicate it there either.

My current thought of how it might be different is that virtual kernel serial port. I'm guessing that a kernel function, so maybe the issue is there? IDK. I'm still running tests and I'm kinda hopeful I can get it to trigger on any of the other OSes.

Can you tell me what version you from

qemu-ga --version

?

Ok, turns out the ones that are wacky are version 8.2.0 and it seems to be fixed in 8.2.2

Suse Support said 8.2.4 but that's not what I'm seeing on my testing.

Ok... I'm about ready to call this quits. After running

dnf update -y ; systemctl reboot

the frequency in which the IP is there/missing is inverted. After updating/rebooting, it's there most of the time vs gone most of the time in my cluster. I haven't been able to get any of the ubuntu or fedora boxes running

8.2.2

to vanish. @ambitious-knife-27333 I'm happy to help with testing if you have another idea.

In essence, I think it is the agent restart that causes this. Are you using the default

cloud-init

config in Harvester?

less than 30 seconds was the IP missing. I'm guessing they patched it somehow, but I have no idea how as the RPM version seems to be the same before it too.