MX warm spare (high availbility) + ISP failover architecture

[grebyn86]

We recently moved all the company's perimeter devices to Meraki MX. All of our locations currently have one ISP into one of the MX's WAN ports, then one of the MX's LAN ports into a Cisco, non-Meraki, switch inside our perimeter. Example;

We are looking into adding a second "backup" ISP circuit and a second Meraki MX at each location for; warm spare, high availability, failover, or whatever it is called, and could use some help with the architecture. 

This is the architecture we initially came up with based on various articles, forums, ext... we found online. 

As a starting point, is this architecture correct? If not, what do we need to change? 

Assuming the first architecture is correct, we also came up with this architecture, which removes the MS hardware outside the perimeter. If this architecture is functionally correct, is it a secure risk to have the switch stack inside and outside the perimeter at the same time? 

Thanks everyone in advance for your input and assistance. 

[RWelch]

@Ryan_Miles might have a slide deck to share with you on best (suggested) practices.

[Ryan_Miles]

This past thread covers this topic. While using Meraki switches for WAN breakout works, be aware it's deemed "incorrect".

[grebyn86]

@Ryan_Miles 

Looking over the past thread you linked, and the layer 3 topology article, I am still unsure what the correct topology is for two MXs and two ISPs? Is best practice to go ISPs directly into the MX, and if an MX fails, physically move the ISP cables into the second MX? 

[Ryan_Miles]

Using breakout switch(es) for automatic failover would be ideal vs. relying on physically moving cables to restore service.

Now whether you do this with one switch, two switches, cloud managed switches, non cloud managed switches, dedicated switches, dual purpose "inside" switches like your #2 diagram shows, etc. you're never going to find one agreed upon design. People have their preferences and beliefs and you need to find the right solution to fit your requirements (cost, manageability, redundancy).

I will refrain from saying what I'd do and leave it to others here to offer their advice and experience on what's worked for them.

[grebyn86]

@Ryan_Miles 

From a security perspective, is having a dual purpose inside/outside switch(es) or Meraki MS switch(es) between the MXs and the ISPs, a concern? If so, I saw unmanaged switches mentioned to sit between the MXs and ISPs. Cisco CBS110-5T-D for example. Would unmanaged switches alleviate the security concern, or is the only secure way to have the ISPs go directly into the MXs? 

[cmr]

@grebyn86 this is exactly why I use them, reduced security risk.

[KarstenI]

I wrote a blog post some time ago summarizing the different options. Perhaps it helps:

https://cyber-fi.net/index.php/2024/02/19/connecting-your-meraki-mx-to-the-internet/

[DarrenOC]

Agree with a breakout switch upstream of the MXs to share the links between your MXs. 

[cmr]

I always go with your first architecture, but replace the MS130s with unmanaged Cisco layer 2 switches.  Cheaper, simpler, can't be hacked and do the job.  At their price you can always have a spare onsite.

[grebyn86]

Thanks everyone for the input. I believe this is the direction we're going. It will be at least a few days before we begin moving forward with this. So if anyone still has an opinion let me know.