> On 5 Aug 2016, at 12:37, Juliusz Chroboczek <[email protected]> 
> wrote:
> 
> It's difficult to debug remotely, since you're not giving the relevant
> information.  (Which interfaces are the routers connected over?  How are
> they configured?)
> 
> Did you install the "ip" OpenWRT package?  (Check with "opkg status ip".)
> We've seen silent failures when busybox ip was being used with hnetd.
> 
> If that's not the issue, I suggest checking whether the routing table is
> sane and whether babeld and the kernel agree on the routing table
> (ip route show, killall -USR1 babeld and check /var/log/babeld.log).
> I also suggest checking the neighbour cache (ip -6 neigh show).
> 
> -- Juliusz
Understood. I tried to simplify the test case. Missed out the nic names, I now 
realise.

ip was a separate package on both.

I’ve decided that I need to be more confident that the underlying openwrt’s are 
behaving themselves. I’ve regressed (2) to a new, clean, minimally configured 
Openwrt DD. Although this seemed to route correctly to start with, rebooting 
(2) failed to get a response to IA_PD: I think that I am seeing the bug that 
Tore identified in the CC release of odhcpd.

So I need to fix odhcpd to get a repeatable test with manually configured 
routes before I go any further on the hnetd front.

To Rich’s point wrt tooling. I’m having to construct similar tools, although I 
tend to run them using ssh from a separate machine so that I’ve got centralised 
control of the environment. That also has the benefit that I’ve got a more 
complete toolkit than busybox et al (diff, git, rake, etc), making it easier to 
create repeatable tests and compare results.
_______________________________________________
homenet mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/homenet

Reply via email to