Network Dies from Plex + Metal LB

[psiinara]

Router: Ubiquiti Dream Machine Pro (192.168.10.0/24, DHCP only hands out addresses above .50)

PC1: Gaming PC Custom Built (Windows 10) (192.168.10.69)

PC2: Asrock 4X4 Box 4800U (Debian 11) (192.168.10.16)

 

Not sure where else to post as this problem has been plaguing me for the past few months.

 

I'm running k3s (kubernetes) on my Debian 11 Box (pretty vanilla, nothing special) with a few services such as Plex, Radarr, Sonarr, Prometheus, ect..  In order for me to access these services on my network, I use Metal LB in layer 2 mode so it will assign IP's to the various services running in k3s (for example Plex is running on 192.168.10.23, Radarr is running on 192.168.10.24, and Sonarr is running on 192.168.10.25). Every few days the Debian box just stops responding.  It becomes unpingable on the network and I havent been able to hook up a display to it to investigate what is going on.  I reboot it and when it starts back up, its pingable again.  The moment the Debian box dies the Windows box network also dies and doesnt come back until i disable/re-enable IPv4 or bounce the Debian box.  I'm thinking of trying a nodeport instead of MetalLB but thats kind of unsatisfying.  I'm also thinking of cutting another VLAN and putting this server on that VLAN to prevent the Windows machine from dying.

 

Things to note or things I've tried:

 

- STP is enabled on the network

- The ARP table becomes incomplete for both the Debian box (192.168.10.16) and all IP's running on it.

- Used to have an edgerouterX with the same exact problem.

- Tried Ubuntu 20.04 and had the same problem

- All other wireless clients on that subnet (via wired Access Points) are completely fine and operate normally.

 

I know this is a super specific problem, but I've been ripping my hair out on this one and can't figure out what the problem is and any ideas/help would be awesomely appreciated.

Edited January 24, 2022 by psiinara
router name change

[HeroRareheart]

32 minutes ago, psiinara said:

Every few days the Debian box just stops responding...The moment the Debian box dies the Windows box network also dies

If the box dies every few days relatively consistently, EG almost every 40-50 hours, I suppose you could setup the machine to restart itself every night after you'd be in bed. It doesn't find and solve the issue at it's core but it could prevent it. Unfortunately I have no idea what the issue could be, seems super bizarre.

[psiinara]

On 1/23/2022 at 9:49 PM, HeroRareheart said:

If the box dies every few days relatively consistently, EG almost every 40-50 hours, I suppose you could setup the machine to restart itself every night after you'd be in bed. It doesn't find and solve the issue at it's core but it could prevent it. Unfortunately I have no idea what the issue could be, seems super bizarre.

Thanks for the response!  Yeah theres plenty of workarounds I can think of, but I feel like I would have lost...

[HeroRareheart]

On 1/25/2022 at 1:13 AM, psiinara said:

Thanks for the response!  Yeah theres plenty of workarounds I can think of, but I feel like I would have lost...

Unfortunately you may have to settle for a work around if nobody can help you get to the root issue.