Network conflict on VM with multiple interfaces

So, I finally got this project (PiAlert) working how I'd like.

It basically uses arp to keep track of devices on your network, and let you know when new ones join. It gives some basic stats like uptime, etc and you can configure a few different notification options to be alerted when a rogue device connects.

Anyways, to get this work on my network involved setting up several network interfaces, as I have quite a few VLANs I'd like to keep an eye on. While everything seems to be working, I feel like I may have created an asymmetric-routing situation, as now when I SSH to the VM hosting this, it will freeze up after a few seconds.

My interfaces look like such. The problem is that I am accessing this VM (hosted on from my personal network ( My personal network has access to and obviously to it's own subnet, so I think packets are getting confused, as there are multiple routes they can take to this VM.

I believe this is confirmed, because if I disable the entry for in my /etc/network/interfaces file, the problem goes away.

How should I handle this? I've tried some simple UFW rules to try to force things to only use the interface, but to no avail.

Edit: Sorry for the weird markdown, not sure why it's highlighting keywords

Creat ,

Do I understand this correctly? Your PC is on .6.X, and your connecting to the PiAlert on .1.X, but it also has an interface on .6.X? You just can't do that with Linux. Weirdly enough I hink Windows handles this correctly and sends the responses back via your router (I think any stateful TCP connection will use the same interface both ways). This doesn't explain why anything actually freezes though. Did the VM lock up, or is it just ssh that's dropping?

But as for the solution: if both devices have interfaces on the same network, you should connect to that interface.

root OP , avatar

Just SSH dropping. Everything on the VM side is ok.

And yes, the computer I'm using is on .6.X (LAN VLAN) and the VM is on .1.X (MGMT VLAN).

The management VLAN is only accessible by a couple devices and this is one of them. To get PiAlert to be able to see devices on the LAN VLAN, it has to have an interface to be able to ARP from.

Creat ,

Yes I know why you have so the interfaces, but as far as I know: Linux simply can't do what you want. So if you want to access PiAlert from your main PC on .6.X, you need to make that accessible from .6.Y on that VM. If you want to have the management port (UI) only open on the management interface, you would need to remove it's interface on .6.X.

As I said, as far as I'm aware Linux simply can't not route packets properly in an environment like that. I won't respect that the interface packets came in on needs to also be the outgoing interface for the return trip. I also had that problem and eventually j I've just given up.

root OP , avatar

Understood. Thanks so much!

Taleya ,

Try coding a static route?

root OP , (edited ) avatar

Would that be similar to telling SSH to listen on only one interface? Because I did try that but it unfortunately did not resolve the issue

Edit: Found what you mean. I'll give this a try, thanks!

Taleya ,

Not quite. Static route is coded on both ends. Tells your machines 'if you want to talk to network B use this ip / route and no other.' And then on the other end tell the machine 'you want to talk to network A via this ip/route and no other.'

You can jigger with the subnets obvs, to cover an entire network range or just a specific ip / machine

jet ,

Packet capture and verify your hypothesis about the routing.

  • All
  • Subscribed
  • Moderated
  • Favorites
  • random
  • [email protected]
  • All magazines