Redundancy with 2 bridge and LACP switchs

RouterOS Wireless Networking
1

Hi all,

I need to make a double wireless link with 4xRB333.

My costumer have Allied Telesyn switch LACP capable and we would use this to offer redundancy.

If I configure the 2 links with WDS and interfaces in a bridge (link in transparent mode) this should work? LACP packets will pass over the link without problems?

I have to make some changes to the links configuration?


Thanks,
Martín.

Forward LACP in bridge?
2

Hi Martin,

I am pretty sure this will not work as the Linux bridging protocol interferes with the packet stream. A good example is that we tried to set up redundant links between a Cisco 3750 and a Cisco 3560 using a Proxim/Terabeam GigaLink and a Mikrotik Link using 2xRB333’s and MSTP. This did not work as the Mikrotik links were not passing the spanning tree packets.

Hope this helps.

3

Hi and thanks Nz!


If LACP will not work, then, how can I make this 2 links with redundancy? I make an identical scenario with 2 Alvarion Breezenet B100 links.

There is any way to make it possible with mikrotik?


Thanks!
Martín.

4

Hi all!


It can work with EoIP tunnels?


Thanks,
Martín.

5

Hi Martin,

Not too sure on that one, maybe Uldis can help ?

MT always seem to go quiet when issues regarding MSTP, LACP and the like come up :wink:

It would be nice if they could integrate the ZebOS stack into RouterOS, it would fix a number of issues that currently exist and I wouldnt mind paying a bit more for MSTP and LACP support.

6

Hi! 3 days trying different solutions, but anyone seams to work..

Uldis, there is any solution to make my costumer LAN working with redundancy? I’m in a trouble now..


Thanks,
Martín

7

Hi Martin,

What you can do is run RSTP accross the bridge, this is ugly but it will work.

8

Hi nz! but there is a problem with this..

If I have 2 mikrotik in each side, connected to my switch then the RJ45 switch connector allways will work. If I lose my wireless link these connectors will be working.. I think I can lose data with this.


Martín.

9

Hi Martin,

No it will use one or the other Mikrotik link. I have this operating at a few sites with Cisco 3750 → Cisco 3560 links.


Regards,



Andrew

10

Hi all,

my trouble with LACP and all relationated with my double link continues very bad.

I tryed using RSTP in both sides (RB333) and in switchs and the link is redundant but only one link works at same time and the other link continues unused.
I tryed using EoIP tunnels in the 2 links and all works same than in last case. LACP packets can’t pass over the link and switch ports aren’t syncronized.

Can anyone help me with this? I think Uldis isn’t reading this post : /

I need to make this link working at about 100 Mbps UDP (50 of one RB333 link and 50 with the other link)

My costumer will crush my head with the 4 RB333 : O!


Thanks,
Martín.

11

have you already configured per the bonding requirements on RouterOS? It also looks to use LACP to handle things.

http://www.mikrotik.com/testdocs/ros/2.9/interface/bonding_content.php

Sam

12

Hi changeip,


thanks for reply.

I think I can’t make bonding with nothing, see my link in the attached JPG.


thanks,
Martín.
LACP-MAPRO.jpg

13

One more thing:

LINK must be transparent, nothing routed.

LAN1 is 192.168.1.x and LAN 2 is 192.168.1.x too.


Thanks,
Martín.

14

Hrmmm,

Why dont you just use a single RB333 at each end with two cards installed, then run these in turbo mode with bonding ?

15

im guessing for bandwidth reasons …

so the main problem is that LACP packets are not passing between the switches thru the routeros boxes? i read below (above) that its the linux bridging protocol thats stopping this ? Is this a known limitation or a bug ? I was thinking down the road of a setup like this as well and would like to know if it can’t be done ahead of time so I can investigate alternatives.

16

Nz_monkey,

If I use only one RB333 on each side and this RB333 fails, my link fails too. I need to use 2 RB333 on each side for redundancy, also for better bandwidth.

I think that my schema/diagram isn’t a strange redundancy solution. I need to have all electronics duplicated.


changeip I can confirm this. If I put 2 ethernet wires between my 2 switches configured with LACP, ports are syncronized. If I make this simulated ethernet wire with WDS link or with Nstreme link or AP-station or with EoIP tunnels ports aren’t syncronized.

I think that if Mikrotik makes LACP working on RouterOS we can make better solutions. We will have support for this or anything to make redundancy 100%?



Thanks,
MArtín.

17

Let me see if I got the point…
side1: RB333 → side2:RB333
side3: RB333 → side4:RB333
Symmetric bonding:
RB600_1+side1+side3
RB600_2+side2+side4
side1 plugged via rj45 on ether1 of rb600_1 and side3 on ether2
side2 plugged in ether1 of rb600_2 and side4 on ether2.

rb333s wds bridge all
rb600_1 add bonding interface(use 802.3af if smartswitch behind) → bond ether1+ether2/see manual on the wiki/
rb600_2 add bonding interface → bond ether1+ether2
create bridge and add ports ether3+bonding interface on both sides.
u r done. :slight_smile:
bonding.png

18

Hi promind,

I have to avoid using more equipement, due to redundancy of all components. If you see, there is no redundancy in RB600s. What is the problem if RB600 die?.. This is solved using LACP but I can’t use here.


Thanks for your reply, it is very clear.


Martín.

Edit: If my costumer thinks that this is ok we will make this solution. This is the second best solution I think, the best solution is to use LACP.
Thanks.

19

Hi Martin,
There should be any solution…probably if using lacp over vlan…but right now I don’t have the equipment to confirm if it’s possible.
I’ve found this:
http://www.juniper.net/techpubs/software/junos/junos76/swconfig76-network-interfaces/html/interfaces-ethernet-config19.html

20

We recently had the same problem while trying to acheive the same result. Complete failover, higher capacity, efficient frequency useage. We were implementing this using SMC8024L2 switches, 2 x x86 dc powered rackmount MT units on each end using Ubiquiti XR5 cards into Andrew dual polarisation dishes with nstream over 50-100km distances.

What we found with this was that the link failed over perfectly when the wireless dropped. The problem was that the switches didn’t register when the link came back up again and a packet storm ensued. The only way to make the switch re-negotiate the LACP was to change the port status on the routerboard via a script.

For instance setup a script on a schedule to check the registration table and if the contents changes then change the mode of ether1 from autonegotiate to 10Mbit and back again. You cannot simply disable and re-enable the the ether1 port as it doesn’t turn off the port only blocks the traffic so the switch doesn’t detect the status change.

The other way that a LACP type implementation can be acheived is to use Cisco switches at each end and use Cisco’s ‘etherchannel’ technology. Because this is proprietry Cisco technology the etherchannel negotiation packets pass through the MT bridge ok. This implementation is not as good as LACP because it’s slower (up to 30s to reconfigure) but at least works.

This information may be a little out of date as the last time i tried this was when V3 was still in beta, i haven’t tried this with V3 stable or using vlans, MPLS etc…