Proxmox 6 and Ryzen 3700X (X570)

[xvegax]

Hi guys,

I am new to Proxmox as you can see on my other post and I have done a fresh/vanilla installation. The server has a 3700X on a Gigabyte X570 mainboard and 32 GB RAM.

I realised, that I am having a lot of connectivity problems, meaning often I cannot connect to the box without any reason and also often it disconnects me from it, while being in a SSH session. I then have to wait 5 - 20 minutes to be able to connect to it again, but it always stays pingable during this time.

When it happened I noted the times and later on checked /var/log/syslog for these times, but there is just nothing special I can find, what should lead to those disconnects or non-reachable times.

Is there a problem with the new architecture? Do I need a newer kernel?

 

[xvegax]

Anyone having experience with Ryzen 3000 or an idea how to troubleshoot this?

Alone today I lost remote access (or got "connection refused") five to ten times without even doing anything.

Firewall is turned off, fail2ban is not installed.

 

[Stoiko Ivanov]

hmm - on a hunch - could it be that something else is having the same IP in the network as your Ryzen Box?
* that would explain the being able to ping the ip consistently but getting timeouts on the services (you're pinging the other host with that ip)

* try pinging your gateway (or some other public ip (e.g.8.8.8.8)) from the Ryzen and see if it has constant access to the network.

else check `dmesg` and `journalctl -f` while the ryzen has no connectivity

 

[xvegax]

Alright, I checked now everything: syslog, dmesg, several tcpdump files, ... and I really found nothing. I am lost here.

But I did something just for testing purposes and I am having a session open for over two hours without a connection loss.

I changed

auto enp3s0
iface enp3s0 inet manual

auto vmbr0
iface vmbr0 inet static
        address x.x.x.x
        netmask 255.255.255.248
        gateway x.x.x.y
        bridge-ports enp3s0
        bridge-stp off
        bridge-fd 0

to

auto enp3s0
iface enp3s0 inet static
        address x.x.x.x
        netmask 255.255.255.248
        gateway x.x.x.y

#auto vmbr0
#iface vmbr0 inet static
#        address x.x.x.x
#        netmask 255.255.255.248
#        gateway x.x.x.y
#        bridge-ports enp3s0
#        bridge-stp off
#        bridge-fd 0

So, removing the bridge helped, it seems.

What the hell is wrong with the bridge? Did I do something wrong here? Does Proxmox have a problem with the bridging?

This is just my management interface, but my problem is now: How do I get all the VMs/Guests to communicate over my other bonded interface (not mentioned here)? Do not I need a bride for that, too?

 

[xvegax]

Btw, it is now nearly 24 hours with only one disconnect. So, my problem was the bridge interface definitively.

My problem stays then: How do I get all the VMs to talk with the outside world, if a vmbr0 is not possible?

Or also interesting: What is causing the problems on vmbr0? Is that a Proxmox or Debian problem or really a hardware issue?

Anyone knows how to troubleshoot Linux/Proxmox bridges?

 

[Benc]

The problem is not your hardware, or proxmox.... the problem is that we have spent a month trying to teach you how networking in linux works, and your configuration is still not correct.

I'll leave a reply in your old thread with some examples to see if that helps you get back on track.

https://forum.proxmox.com/threads/two-1-gb-nic-for-lb-and-doubling-speed.57179/#post-269310