mjhjmtu

I'm having a problem with my CentOS 7 host, installed last weekend. I use it with KVM to host a few VM's, which are bridged to the same network as the host via software defined bridge br0. The NIC on my CentOS box is also bonded to br0.

I had this same setup prior to last weekend, until I tried Proxmox as my host OS, didn't like it and reinstalled CentOS.

When I got in from work yesterday I noticed that I couldn't SSH into any of my VM's or the host OS. When I logged into pfSense, the DHCP leases showed that all of my VM's were 'offline'. The CentOS host looked like it was still running but I could not ping the host or any VM's from inside my LAN. I decided to hard reset my CentOS host by forcing a power cycle. I checked the logs once it rebooted and there didn't seem to be anything interesting like a kernel panic or anything. One of my VM's does an hourly rsync via cron and I could see that it hadn't reached an external site for a few hours. Therefore I assume that networking between my CentOS host and pfSense box had gone down some hours before I returned from work.

I put it down to a 'brown-out' or some kind of electrical disturbance causing my host to stop networking and forgot about it.

This morning I woke early and I went through my usual routine. I SSH'ed into one or two of my VM's and everything seemed normal. Then I watched TV for about an hour and next time I looked the same thing has happened again! The CentOS host has randomly stopped communicating with the pfSense box. This has happened literally an hour while I've been up and watching TV for a while: nothing in the home has flickered and all of my other electrical home stuff has done anything weird. I live in a built-up area of the UK and we rarely have electrical outages anyway.

I tried unplugging the LAN cable from my pfSense box and plugging it into my UniFi switch and I see lights flashing like it is trying to negotiate DHCP but it won't, because I've got my switch and pfSense ports on two discreet subnets. It proved that Linux is still running on the box, though. The box is headless with no GPU, so I can't plug it into a monitor without forcing a power off and plugging in a spare graphics card to get a physical screen up.

What should I do to troubleshoot? I have a spare GPU. I'm thinking of forcing a shutdown, inserting the GPU and just starting everything up again and next time it happens I'll have a visual display to see what's happening.

I'm going to plug the problem machine back into pfSense and reboot pfSense first, to see if it renews a DHCP lease.

It's all a bit weird. I'm sure that when I set up my host again last weekend that I configured it the same as I'd had it before and everything seems to be working as normal.

To give a bit of background information, my network setup is like this: CentOS host is plugged into dedicated pfSense box via gigabit LAN cable, pfSense box is plugged into UniFi 8-port switch which powers my UniFi AP and I have a Raspberry Pi plugged into the switch, running UniFi controller software.

All wireless devices are functioning normally. I'm typing this from my laptop which is connected to the UniFi AP.