Gigabyte GA-990fxa-ud3 2 weird issues No network connection.

[eminence]

I have done a fresh install of latest Proxmox 5.1-3 on 3 machines to create a PVE/ceph home setup. Machine 1 ( HP z420 v2) works fine. Machine 2 ( custom build FX 8320e / Asus M5A78l-m/USB-3 mobo) works fine. Machine 3 ( FX 6300 Gigabyte GA-990FXA-UD3 ver.3) has 2 separate issues.

GA-990FXA-UD3 does not have onboard GPU, I must use dGPU for install. Removed dGPU after install. Asus M5178l does have an onboard AMD gpu.

Fresh install of 5.1-3, non-subscription repository updates applied. For future reference I have performed clean installs, once with onboard Realtek NIC and once PCIe Intel NIC. Same issues below.

Issue1: This issue is a non issue long term as I am planning to run headless without a gpu. flashing cursor on screen during "shutdown -h now" or "halt". Hard drive thrashes for a bit and then I hear the heads park and mobo/ fans still run....forever, until I hold power for 5 seconds. However, this only happens on shutdown. It will reboot with no issues at all. I have tried both 1 AMD and 2 NVidia gpus. I have added acpi=force and apm=power_off=1 to /etc/default/grub. This MOBO has some iommu bugs that are solved if you add amd_iommu=1 and iommu=pt ( iommu=soft also works) to grub to avoid a bunch a error messages popping up. I have not tried any pass-through configs yet, nor will I. I cannot seem to fix this issue. It shouldn't be an issue as I want to run headless. This shutdown issue seems to disappear if I run headless..... which introduces/causes issue 2

Issue 2: booting headless/ no gpu. It appears to boot but NO network connections. During "boot" : Hard drive randomly thrashes/HDD LED activity like it is booting properly and eventually settles down as if it fully booted. Here a weird part with reference to issue 1: If I push power button ( just like my other 2 working nodes), hard drive starts thrashing for a bit, and node shuts down properly after a few seconds. Issue 1 goes away, with no GPU( I think...can't log on to see), but I have no network connections up to SSH into machine or webgui config it. I can't find this issue on google or proxmox forums.

I am not the most linux savy but get around well enough. Which logs should I be looking at for troubleshooting this network issue...after I reinstall a GPU to get into the physical terminal. I will need to look at previous logs as once I boot with GPU it will boot fine.

 

[eminence]

update 1: clean install of debian 9.4, get the same shutdown hang using shutdown -h now or halt with a GPU installed. However, "shutdown -h -f now" actually powered off the PC. Also, have the same network issue once I pull the GPU out and boot Debian 9.4 without GPU.

So I ruled out a Proxmox specific problem with my hardware. So it is at least a Debian issue.

Also, I did a "more" messages(I forget the command exactly, but it was on tecmint.com) log lookup prior to installing Debian. It mentioned booting Intel enp5s0, but lscpi states intel as 6 not 5. I then edited the /etc/networking/interfaces file to read enp5s0 for fun....still no go. I think I will test some CentOS for the above behaviors, and either get a differnet PC for that node, or use my old Dell t3500 Intel e5630( also works fine with proxmox 5.1-3). I thought I would try the AMD FX 6300 as it uses the same watts and has more cpu power and has a case/mobo with more drives/ports....off to test CentOS .....or maybe I won't bother...undecided.

It is like the devices are shifted or pointed to differnet hardware once I remove the GPU...or not since switching enp5s0 didn't help.

 

[eminence]

UPDATE 2: installed CentOS 7....same exact behavior despite 3.10 kernel. However I have resolved the shut down halt + hang thanks to CentOS displaying more things on shutdown, I found this which corrected issue(at least on Centos 7):

Solution: add the kernel boot parameter "acpi=noirq" in the quotes with "quiet splash....etc" of /etc/defaults/grub

However, It did still have the "no network" after pulling out the GPU to run headless problem still