> On 07 July 2017 at 01:10 Greg Oliver <[email protected]> wrote:
> 
.....
> I would really have completely different routing tables for this rather than 
> changing metrics.  Source routing /netfilter is your friend here, but source 
> routing also comes with security risks, so you need to make sure netfilter is 
> stopping anything you do not want.  Basically, then you can easily say the 
> interafce it comes in on, it goes out of (or any interface you want...).
>

Thanks for the suggestion, Greg (and for your persistence in steering me 
towards other ways ;)).
I didn't relish delving into the topic, but it does look like I've been able to 
set up iptables rules to get the system a bit more flexible about the interface 
to use - certainly seems to be behaving more seamlessly, with less intervention 
required. (Mind you, I'm not sure if this is what you actually meant by 
'different routing tables' - but perhaps it just demonstrates your point below!)

I still have slight niggle that what I've put in only works when my eth cable 
is unplugged - if it's an upstream failure then I guess the kernel simply can't 
know of that, and so it keeps trying on eth0 regardless.
I already have a script monitoring actual 'connectivity to the interweb', so I 
could use that to poke NM. I suspect, for that scenario, I probably ought to 
re-Up the eth0 connection anyway: dhclient ought to be re-run (and using 
autoconnect-retries=0) in case the outage was a reset of the broadband modem, 
which would warrant a renew/re-lease of the DHCP. (??)


> The good thing about the linux kernel is that there are usually ten(s) or 
> more ways of doing what you want/need.

Indeed...!
_______________________________________________
networkmanager-list mailing list
[email protected]
https://mail.gnome.org/mailman/listinfo/networkmanager-list

Reply via email to