MPLS-VPLS-OSPF, fallover

RouterOS Forwarding Protocols
1

i was following this example. Every thing i good, except fall over. It takes long time to switch to the other link. For example R4 and R5 have clients A1 an A2 like a picture. R1 is PPPoE Server. When the link between R3 and R4 go down,it takes long time to switch to the other link. (min. one min) . Does anyone like to give me advice .


https://wiki.mikrotik.com/wiki/Manual:MPLSVPLS

Thanks a lot :wink:

2

Issue is caused by bridge, that join vpls and physical interface to client. There is rstp protocol used. You can try to set spit horizon (same for all vpls) and disable rstp at bridge.

3

on this exsample, you mean on R1,R4 and R5?
even if i have just three MT, R1 (main) R2(MPLS LSE) and R3 like a client. if i do disable enable on the main link it takes long time to connect pppoe. OSPF works perfect. pppoe makes problem.

4

in scenario from mikrotik … yes on R1,R2,R3 … but only if you have bgp signalised vpls (reconnect occur change vpls tunnel in bridge) … sorry, my mistake

5

im using like OSPF no BGP . I didn’t cake, what do i need to set up? Disable rstp on bridge or..

6

you need use ospf and can use bgp for signalising vpls insted static configuration

7

OSPF and BGP? Can you point me about that on MikroTik wiki, just about BGP

Thanks

8

of couse, in simply version

scenario …

  • you have large network with IGP (like ospf) for connection your internal network

  • you have border routers (to you ISP / to your customers), peering with other ISP/customers is realized with BGP

  • BGP need full mesh or route reflector between all border router - you need use L2 tunnels (to many static configuration) or you can use MPLS (work like cham)

  • BGP can signalise/create dynamic VPLS tunnel between border routers (there is that issue with split horizon and rstp timeout)

  • BGP can make VPLS very simply between multiple customers end points

edit: in all scenario is critical to use loopback with proper IP configuration

https://wiki.mikrotik.com/wiki/Manual:MPLSVPLS
https://wiki.mikrotik.com/wiki/Manual:BGP_based_VPLS

9

i change something in my configuration means MPLS,know is working good, i will test out. Many people who know more about MT thera recommended OSPF instead BGP. However i will have look, just let me test my configuration. know is good. I will let you know

You are good man :slight_smile: :wink:

Thanks a lot

10

BGP and OSPF are not comparable protocols… each of one should by used for different scenario. In real world, you need both.

11

i make some test … icmp echo - standard interval from /tools ping

Cli-----R2—(vpls)----R1
aaaaaaa _R3__/

(ignore “aaa”)

ping from Cli to R1 with static vpls

  • Unplug cable between R2-R1 = 23 icmp echo lost
  • Replug cable between R2-R1 = 20 icmp echo lost
    this time is caused by rstp, without rstp is 3/0 icmp lost
    more complex topology with multiple tunnels and wrong setting split horizon can cause malfunction created by rstp, its very easy to create loop with mpls/vpls

ping from Cli to R1 with BGP signalised vpls with bridge horizon

  • Unplug cable between R2-R1 = 3 icmp echo lost
  • Replug cable between R2-R1 = 0 icmp echo lost
12

AAA you mean on ppp? how come?


About your testing i will try. I got ping 60ms-70ms between clients Let me fist check out my story :wink:

Thanks

13

Leter “a” is used as substitute “space” … multiple “space” are shown as single. Just for schema.

14

:slight_smile: Sorry AAA means A1A2A3 :slight_smile: i have no idea what you mean :slight_smile: Could you like..

15

From Tomas Kirnak about MPLS l VPLS:
oo.png

16

forgot “a”

of cause, icmp is summary for icmp echo (mpls/vpls) is used as ip accelerator, it is his purpose
my test with icmp show reconnect time from main to backup path

17

Have you got idea about split horizon? Eaves if i set up 1 on splint horizon the client can ping to etch other. did you do before?

Thanks

18

Sorry, but I don’t understand your question. What communication you try to block? Please make some scheme/picture.

Split horizont settings “1” is only identification of group of ports, that should NOT pass traffic between them. It should be used when you have complex tunnel VPLS topology like full-mesh to ensure, that traffic will go the shortest way to destination.

In full-mesh topology (without bridge horizon) with bridged ports can occur situation, when traffic don’t go shortest way and some bridges are disabled by rstp and traffic cannot pass.

Or you try to block client communication between PPPoE sessions?

19

i mean the client from R4, R5 and R1. They can ping to ech other. For e.g. One client got ip ad 192.168.1.2 and thiother 1.3, they can ping to echother. Yer they got ip from pppoe server. Any addvice?

Thanks

20

you can block communication between pppoe sessions simply with firewall

/ip firewall filter add action=drop chain=forward in-interface=all-ppp out-interface=all-ppp