Hi - relatively new Scale user, long(er) time Core user.
Recently migrated to Dragonfish 24.04.2. Running a couple of networks deployed as
eno1 (s/micro X11 built-in 1Gb Nic): mgt lan, IPv4 and IPv6 dual stack
bond0 (on top of 2x Intel XXV710 ports) in LACP to switch
vlan10, vlan30, vlan90 on top of bond0: IPv4 only (although the router offers SLAAC, as IPv6 is used by some other clients on the same network)
I ran a similar config in Core, and although IPv6 default gateway never quite worked through SLAAC (there are several thread on that here), I got IPv6 working for the mgt lan by setting IPv6 default gw manually.
In Scale, everything initially seemed to work as expected. I’ve got “Autoconfigure IPv6” enabled on eno1 and disabled everywhere else.
As of the other day, for unclear reasons, this all started to break. Specifically TrueNAS now picks up and sets IPv6 addresses for ALL the interfaces DESPITE only eno1 having “Autoconfigure IPv6” checked in the GUI. It gets worse when TrueNAS then continues to set multiple and conflicting default routes and gets its whole IPv6 stack in a mess. A reboot resolves it, for a short while.
You can do this directly from the UI, start with the smile icon in the top title bar area. That will allow you to file the bug and easily attach a debug and screenshots if helpful.
I hope this can be resolved… There’s been a slow burn since IETF defined IPv6 as an Internet standard but accelerating very quickly it seems. Google measures approx 50% of all public Internet traffic over IPv6 and it’s growing quickly in data centres as well. Fewer and fewer software appliances don’t support it fully and there has been rock-solid support in RHEL and Debian for some time now, ESXi and Proxmox too. Point being that even if the above and other relatively baseline issues can be resolved, IPv6 support is still clearly a second class citizen in TrueNAS next to IPv4. Why does it only allow a single interface to be auto-configured for example?
@patrickkeane since you expressed interest above… I attached all relevant information including debug logs. The ticket however was closed today without action and with comment:
Thanks for the ticket but we don’t automatically add routes in this scenario. If you’re using SLAAC, and your RA is providing routes, then TrueNAS will install them as they are advertised. This isn’t something we’re seeing anywhere else nor can we really reproduce it.
Bit of a head scratcher…
Does TrueNAS automatically add routes or not? 1st and 2nd sentence seem contradictory. What is the actual desired/expected behaviour here?
If TrueNAS automatically does install routes as per the 2nd sentence - which is also what I’ve seen - why does it do so irrespective of whether “Autoconfigure IPv6” is checked? And specially, when it’s NOT. Clearly that is broken behaviour and in addition odd in combination with the GUI restricting it to only be checked on one interface globally.
Last but not least, in all this, example in my debug logs where TrueNAS botches up the default routes when more than one interface advertises RA (not atypical) to the point where IPv6 becomes broken and unusable until reboot.
It seems to me that they actual behaviour is that if “Autoconfigure IPv6” is checked on any interface, it applies to all, and what’s worse in a way which creates a broken state with default gateways mixed up between the different interfaces.
If autoconfigure is disabled everywhere, it doesn’t apply to any. So the workaround seems to be to set everything manually instead. Which is a pain if you’ve got a dynamic prefix.
