Setting 2 MikroTik for hot-standby

RouterOS General
1

Hi guys,

I’m planning to buy two decent MikroTik routers to put on a data centre.

From the data centre, they will be supplying us with two ethernet cable for WAN, so each router will have their own WAN interface.

My question is, (1) what will be the best practice to make both router hot-standby?

So in case of router A went down, I want router B to take all router A’s job and vice versa (preferably no master/slave configuration, but any of them will work)

(2) how to make both routers in sync, i.e. update each other? (so if I add a rule on router A, then it will be replicated to router B)

2

Read about the VRRP concept with MikroTik here.

3

Brilliant!

Appreciate for this.

My understanding is we’ll just need to implement VRRP and use exact LAN configuration on both R1 and R2? If so, how can we make both R1 and R2 config in sync?

Cheers

4

Beginning from the 6th post here, a fellow forum member has created a script for this purpose.

5

Do you have experience using that script?
Would you know if we implement that on an existing router (not a fresh or blank slate router) ?

6

Looking at this topology,

Can we use without Switch in the middle? (i.e. attaching the servers directly to the routers)

I’m planning to create a bridge between router.

I only have two servers though, and each servers has 4 NICs.

Any input would be appreciated. Ta

7

Yes, but if you attach a server directly to a router and that router fails you have no access to that server.

As your servers have multiple NICs you could connect each server to both routers, but you would have to either run a routing protocol on the servers to announce their presence to the two routers, or bridge the NICs and use RSTP to prevent loops.

If you feed everything through a switch that becomes a single point of failure unless you have two switches which support distributed bonded interfaces.

You need to consider what the various failure modes are (e.g. device failure, device misbehaving - an overloaded switch CPU can make RSTP behave really badly, a network cable being disconnected), and what the probability of each is to decide what is required. Sometimes building overly complex redundancy can make the system more prone to failure.