Hello,

I have problem with IPv6 and VLANs (in a port direct to other group of pcs works always ok).
In this case, is a VLANs in a bonding and vlan ARP reply-only, i read in the forum some problems in older versions with ARP in reply-only mode with IPv6.
I set all VLAN ARP to enabled and the result its the same.

Switch on client router and all works ok.
About ten minutes... the ipv6 freezes.
Reset client router or mikrotik router... IPv6 works ok other ten minutes

I update the rc7 to rc9 and do the same.

any ideas?

I've not seen this issue, but IPv6 does not use ARP, it uses Neighbour Discovery, so you might look at your options in that menu.

As a separate point, it might be worth getting more detail on the type of failure - do devices lose their IPv6 addresses/routes, are they able to ping each other, does the router have IPv6 routes?

Yep cupis, v6 no have ARP
the thread is http://forum.mikrotik.com/viewtopic.php?f=2&t=54200
3rd post talks about the "same"

About the discovery, the multicast packets pass ok, all devices are ok, ip, route, dns...

ping directly from client to mikrotik not possible but the interface vlan of mt have only set the fe80:: (local).

from client to client router, works.
from client router to mikrotik router works.

(client)<---->(client router)<----->(2nanobridges in bridge to vlan)<--
-->(RB1100AH bonding in backup and deVLANed routed to 6to4 tunnel and routed in ipv4 to HE)

when ipv6 subnet on client pc not work, the 6to4 tunnel take in and out packets...
aaaand... the other subnet in ether2 <-to-> switch <-to-> pcs work OK... the ether2 have a global ip to autoconf and no pool and dhcp.

on testing.... i add a global ip to router vlan interface and obtain the same result.

The error is in a cisco linksys E1200 (have last firmware Ver.2.0.04 (Build 1))
all ipv6 in auto in exactly 5 minutes downs the ipv6, only globals ip

any relevance the 300 seconds timeout with ND in mikrotik?


Update

The solution for E1200 is put less of /64 prefix in pool, example: /62, and prefix lenght /64.

Why not accept prefix /64 and prefix lenght /64 to leave option of subneting??

/64 is minimum length that can be assigned to the customer. However i have not seen any issues in RouterOS if less than that is allowed. DHCP-PD even allows prefix length of 64 to be distributed to clients with pool length of 64 (that is DHCP-PD can give out only 1 prefix to PD client)

/64 is minimum length that can be assigned to the customer. However i have not seen any issues in RouterOS if less than that is allowed. DHCP-PD even allows prefix length of 64 to be distributed to clients with pool length of 64 (that is DHCP-PD can give out only 1 prefix to PD client)

yes janisk, again, router os works great, best job.


the problem is the firmware of cisco linksys.
i send mail to support...
nex step, flash router with dd-wrt or similar...

EDIT:
THE CPE ROUTER IS OK.
for CPE routers (User final or subscriber router) i see this in the itf
---------------------------------------------------
http://www.ietf.org/proceedings/78/slides/dhc-4.pdf

DHCPv6 Server Functions
Supports stateful/stateless DHCPv6 on its LAN
interfaces 
Serves addresses from its delegated prefix

Requirements
The IPv6 CE router MUST assign a separate /64
from its delegated prefix(es) for each of its LAN
interfaces.
---------------------------------------------------

can not be assigned a /64 prefix to a CPE router, the CPE router need a one /64 by interface.
for common router 4 lans + 1 wlan minimum /61, equals or more than 5 subnets.

is correct?

I am confused. In router os, the parameter in pool to configure 16 client with 16 subnets is 2001::/56 and prefix lenght /60 ????????

...SOLVED...