Prefered method to manage several routers?

Hello,

I manage several routers, and access to them mostly via ssh.
I use for this the fe80:: , and sometimes the fd* one. But mostly the fe80.

so i added some type of customized hostnames pointing to link local ipv6 from owrt-abd6 (for example).

But i saw it "conflicts" to a specific ipv6 VPN, which also generate a fe80 one, which i use too.

How do you manage ssh access on LAN by not using 192.168.x.x ?

thank you vm

Are you saying that the VPN provider uses the exact same subnet, and it somehow interrupted your ability to local administer your routers?

Hostname of the devices. :man_shrugging:

Every interface has an fe80 aka link local address with IPv6 but those are not routed.

You should use the unique local addresses (fd*) for that...

Well, tbh i would say that i "discovered" the advantage of fe80 when it was automatically generated on any connected NIC, plus was unique and sort of "static", means would be the same fe80 on a laptop's wifi on any sort of AP.

sort of "Unique" IP address that wont be taken by any other.

was a bit "tired" of removing the 192.168.1.1 ssh fingerprint of .ssh/known_hosts each time i setup or reset a owrt router :slight_smile:

Other advantage of it, was that fd addresses looks like to be providen only by router, radvd or dhcp6 server, where fe80 can be handled eg to do wired-transfer on switch or wifi connection with only computers, without router/dhcp, that is a bit equivalent of 169.254, and thus can handle IP for file sharing needs.

i dont really know what differenciates fe80 and fd, as the second looks like only provided by router/radvd/dhcp.

i would add : the fe80 address of the router, is the last one, as "default gw" of the ip -6 r command output :
default via fe80::aaaa:bbbb:cccc...... dev wlp2s0 proto ra metric 600 pref medium

But im still a bit learning :slight_smile:

recorded in /etc/hosts on the owrt router, eg :
fe80:aa:bb:cc:dd owrt6
fe80:aa:b1:c1:d1 pc1-6

im still thinking about how to make it a bit better..

well, it's a zerotier with automatically generated fe80 interface (closed/private vpn), and im used to connect via those fe80 because i found it quite easy in a first time. I didnt wanted to use ipv4 for the moment, as i use it almost only under linux, for little needs (ssh).

Yeah cool but the disadvantage is that those LLA are only valid on the link layer, layer 2.
You can not route these addresses.

Edit. Your VPN Setup works because all these links seam to be peer to peer, direct connections.

  • You know OpenWrt has configuration files for stuff like this?
  • Is there a reason you're refraining from using IPv4?

Then configure your ssh client to just ignore the fingerprint or build your firmware with the host key already included.

Most of these approaches (scripts, Ansible, etc.) make sense and work well to start with.

Where I’ve seen things get tricky is over time — especially when changes happen incrementally across devices. Config drift starts creeping in, and it becomes harder to reason about what’s actually running vs what was intended.

That part seems harder to solve than just pushing config.

Curious how people are dealing with that long-term — do you periodically reconcile state, or mostly rely on discipline + tooling?

EDIT:

Disregard, I see this user has been hijacking threads.

Here

And here

Indeed, @s13884 please don't do this. Apart from the hijacking, which is very impolite towards the respective OPs, this is now the your third (sub-)thread I'm seeing about basically the same topic, that will really get on everyone's nerves.

if you know another reason to add a hostname to only ipv6 address, tell me :slight_smile:

because i'd like on some "hostnames" to ping it only in v6, for experiment :slight_smile:

I'm guessing there's a language barrier.

ping6 <hostname>

it's not only ping :wink:

the idea is router get a long name, i want to add a short name via dns. And one only affected to ipv6 address (owrt6 for example)
i found out only /etc/hosts can handle this, imho.

You could solve the issue of your device "getting a long name".

Unfortunately, it's difficult to discuss solutions for your use case, when you keep adding more conditions to that use case with each post. It's not clear if you're experimenting or having an issue managing multiple OpenWrt devices on LAN (yet, somehow need to use IPv6 instead of IPv4).

That's not an OpenWrt config file, but yes it works OK too (as you already mentioned).

but you were saying there is a better way to do? (i mean including v6 only i guess)

It's not "that much" complicated :wink:

initially i wanted to use exclusively ipv6 to hit directly each device on a unique private address, eg fe80 or fd something. I found it's not that bad idea.

The second thing is eg to assign short name in addition of long ones, to hit a specific router (ssh, web..)
eg there is archer-c2-v1 as hostname. But archer6 as "shortname" to ipv6 fe80 or fd, when ssh it from a different computer, eg a freshly installed one.

but if you get oothers manners to manage a dozen (or several dozen) of owrt routers, from various computers (basic linux or windows), i listen it :slight_smile:

thanks

Pardon me but it seams you have not understood the difference between link local addresses and uniq local addresses....

Within /etc/config/dhcp add alias entries aka cname records for that....

Why? /etc/hosts has had aliases since Sun Micro and about 1980. Why add another location?

Please review OpenWrt documentation.