Add additional IP-address

omvki · 19. Juni 2014

For my routers QOS configuration, I want to bind the transmission-plugin to an dedicated IP-address.
How can I add an additional IP-address to my network-card in the webgui?

davidh2k · 20. Juni 2014

You can't via GUI. Not even sure if this works via CLI... can't you set your QoS via port settings?

Greetings
David

scoobynz · 20. Juni 2014

I used to do it that way. Google "virtual ip address debian"

That said, a much better way is this; http://forums.openmediavault.o…17&hilit=transmission+vpn

omvki · 20. Juni 2014

Thanks for the information.

Zitat von "davidh2k"

can't you set your QoS via port settings?

Thats the problem with transmission. The peer-port just works for incoming connections. But for outgoing connections, transmission uses random ports. Binding to an extra-ip is the only solution to create a rule for it.

I just added

Code

auto eth0:0
iface eth0:0 inet static
  address 172.16.0.91
  netmask 255.255.0.0

to /etc/network/interfaces. Is that ok, or will the file overwritten by times?
I also bind transmission to that address and created a QOS-rule in my router for the ip-address.

One thing I have to fix: If I enable NAT-PMP in transmission, my router forward the port to the ip-address of eth0 instead of eth0:0. Any idea how to fix this?

SerErris · 30. Juli 2014

I am pretty sure that his will not deliver the expected result.

Your problem:
An Ip belongs to a host and not to an interface. So outgoing traffic is not bound to a specific interface but to the first interface in that subnet.

This is a common routing problem and a old dilemma. And what you see on your router, is exactly the effect.

The effect is: all outgoing packets are sent via eth0 regardless of what IP you tell transmission to use. This is true as long as both IPs are in the same subnet. The inbound traffic is accepted by the secondary IP, but you will never get a packet delivered to that address via the as the router get the request from the primary IP (outgoing is sent by that IP). I am sure I am not very good to explain the issue, but it does not work as you hope it should.

So for the ingoing traffic there is an easy workaround. Setup the NAT rule on your ruter manually and point it to the .91 IP. That will at least enable inbound traffic to be limited or shaped or QOSsed. But for the outgoing traffic it will allways come from eth0 and the router still cannot distinguish from the other interface.

The bind setting in transmission is meant to bind it to one interface in a multi network computer (multiple network segments/subnets).

So if you have a server with a public IP (frontend) and a private IP (backend) then you can bind transmission to the public IP and limit it to that.
Your scenario is different and routing is not working that way. A application simply cannot define the network used for sending stuff. It can only define the target and the interface for outgoing traffic is assigned by the kernel.

There is a hack available with source based routing ... but I am not sure if that works here.

DAVIDKUEK · 29. August 2018

How about using nginx to configure a virtual IP address ?

Jetzt mitmachen!