Hello Folks!
The documentation is great, exept I do not understand what all stuff do, it is to deep down in layer 2 stuff, got to get a book and try translate it but had no time yet. Recommendation is to make a much more simple interface to deal with basic vlan, trunking, bonding, etherchannel etc. but I guess you like me, is very occupied with stuff and upper management asks for that and that all the time 
Anyhow we tried everyting now, still the CRS leaks traffic and hence can not be put in to production.
I do not know to much about vlan and layer 2 stuff, but I have all the years being able handling the same on Cisco and HP switches without problems.
With everything means:
- A brand new untouched CRS was taken ount from shelf.
- All packages disabled: hotspot, mpls, ppp, routing, wireless.
- The device was resetted fully with no scripts to be run at boot to configure it.
- Mac address access to the device was performed
- IP Settings was fully disabled, ip forward, send redirects, secure redirects, allow fast path.
- dhcp client was enabled on ether1
Then the device as accessed using the dhcp address on ether1.
The device as then upgraded to RoS6.9 and step 3 → 6 was repeated and we continued with below steps.
Following mikrotik examples by the book for CRS (http://wiki.mikrotik.com/wiki/Manual:CRS_examples) we put ether2 as “trunk port”, by our understanding it will accept all ethernet traffic coming in or going out without exeptions like ALL VLANS trunk something. Must say I do not really understand point 9 below, how can the switch know that ether3 is belonging to vlan 200, does it come from sa-learning, what is that by the way ?
- switchport ether2 was set as master for ether3
/interface ethernet set ether3 master-port=ether2
- Tag all ethernet packages coming in to ether3 to vlan 200
/interface ethernet switch ingress-vlan-translation add port=ether3 customer-vid=0 new-customer-vid=200 sa-learning=yes
- And the reverse, remove vlan tags for traffic going out on ether3.
/interface ethernet switch egress-vlan-translation add port=ether3 customer-vid=200 new-customer-vid=0
That wasnt to hard, now we connected the CRS ether2 to the trunk line containing very many vlans.
Last we connected one RedHat dhcp client to ether3, in a blink it got IP address from our dhcp server on vlan 200.
tcpdump was started on the RedHat dhcp client to see what is going on in vlan 200, sadly we saw a lot of arp requests and other oddities leaking from all other vlans. But it is isolated in some way, because we could not ping servers in other vlans that was not routed to vlan 200 from the RedHat client neither did we got any arp addresses from other vlans so that is positive.
We then went further trying to activate port isolation, but it goes back to promiscues all the time, not possible to change.
Am I doing somthing wrong in the setup ?
It could actually work this way, if it does not cause any conflicts and other oddities.
I would really like to throw out our slow Cisco switches and go gigabit now, also have noted the CRS does not consume by far as much energy as the cisco:s.
What do you experts say, is it safe to go into production with CRS at this stage if you can live with the little leakage ?
I have not experienced any port flapping etc or other problems in our environments for a very long time, and the few we had was sorted out by disabling snmp and the lcd display, and it was on CCR not CRS. Also our RB2011 has performed without any problems.