Hi All,
As an experiment I moved the IPs from the VLAN sub-interfaces on each
node to the parent bond interface and put that in the VRF... and it
works.
Due to the ESXi test environment and the vswitches not supporting
traffic with 802.1q tags I had used 802.1ad (0x88a8 ethertype) for the
sub-interfaces. I don't think it should make a difference, but yeah.
Moving the test off ESXi and onto real boxes with 802.1q tags to see if
that makes any difference.
This is definitely not a bird thing, let me know if you want me to post
updates anyway just in case for the archives :)
Regards,
William
On 27/08/2021 09:21, William wrote:
Hi Alexander,
Thanks for the response, I suspecting it's not just bird from the sshd
behaviours too, but thought someone here may have run into something
similar and be able to suggest something.
iptables/nftables is not in use (all chains ACCEPT in all tables), nor
is ebtables. For completeness I have unloaded the kernel modules but
no change.
IPv4 and v6 forwarding is enabled.
Just thought I'd test IPv6, getting the same behaviour too.
I'll keep investigating and see what I can come up with. I don't
think I've missed anything considering I can ping inside the VRFs on
both sides but that's kernel-space, not handing off traffic into
user-space.
Regards,
William
On 26/08/2021 23:27, Alexander Zubkov wrote:
Hi,
This does not look like bird-related. As you have rp_filter disabled
already (net.ipv4.conf.all.rp_filter too?) then you can also check
things like iptables, maybe forwarding?
On Thu, Aug 26, 2021 at 4:57 AM William <[email protected]>
wrote:
<snip>
