VRRP physical questions

[Ctek]

Hi,

Maybe a dumb question but let's say I have a network with a single L3 core switch on top that acts also as a internal router. All the used ports on this switch are other switches (some with LAG's).

Then I want to add another same core switch for VRRP redundancy, do double wire all the other switches to the backup one also? I think I would need to use STP too for this to work.

I am a little clueless since it's the first time I'm toying with VRRP.

Thanks

 

[Speedsterworld]

Disregard this comment and read comment further below!

Yessir you definitely need STP, either RSTP or MSTP depending on your requirements.

I've yet to set VRRP up with 2 core switches myself as well, close but not yet. All my studies and network knowledge would suggest requiring STP since STP runs at layer 2 while VRRP is a layer 3 protocol. STP will ensure the redundant ports in your backup core stay down until they are actually needed(your other switch goes down).​

 

[payphone7456]

You could do this with port-channel, ether-channel, LAG, ETC. if:
a. you have at least two switches stacked
b. your switch lets you make a port-channel with a port on each switch
c. the second switch doesn't die if the first one does

On a Cisco 3750 one switch can die while the other still works. I've seen a stacked 3750 on fire while the other one worked before. The fan made it look like a flame thrower until the fan melted. I'm sure there are some faults that would cause both to die though. Fire apparently not being one of them though...

Not sure if this would work on their small business switches like the SG300. Also not sure if this works on other vendors. I think it's a lot easier to setup than something like VRRP.

You are more likely to lose connectivity due to power or cable problems most of the time though. Enterprise level switches are pretty reliable these days. I have several 3750's that have been on without fail for years at this point.

If you have a real mission critical setup and you still feel the need to go this route, make sure you duplicate everything you can. Power coming to the top of my mind when I think redundancy at the core level. Using a core switch that has this sort of thing in mind is a good idea also. A Cisco 6500 is what I usually use. Your budget may vary.

I think STP is fairly assumed these days. I don't think it's a good solution for redundant links. I think of STP as more of a safe guard against accidental loops.

-Sean

 

[Speedsterworld]

You're right Sean and I am wrong. STP is to be relied on for accidental loops. I wasn't sure about Cisco's port channel option across multiple switches.

You made me do some research for myself since i'm primarily on HP switches with all of my clients.

HP's term for this is:

Distributed trunking — enables loop-free and redundant network topology without using Spanning Tree Protocol; allows a server or switch to connect to two switches using one logical trunk for redundancy and load sharing

In my case this is covered in the 5400zl series which aren't done through stacking. Just two separate switches I believe.

That sums up everything I need to know about it!

 

[jjeff1]

You have 2 different problems here.

First, you need to make sure that each edge/distribution switch has access to both switches which you intend to use for VRRP. You can do that with STP, you might also investigate HPs switch meshing. You also need to make sure that the core switches have access to each other.

Second, you need to configure and test VRRP on 2 devices.

Normally VRRP is used on larger, more distributed networks. Not where a single switch forms both the core and distribution layer. You have to worry about things like: if core A has a module fail, will the edge switches have layer 2 access to core B, but no layer 3 access to your VRRP master on core A?

Remember that if core A loses connection to core B, both switches will assume they are the master, and you'll have split brain and all sorts of problems.

Likewise, where are your servers and Internet connected? If the building with core A loses power, does having routing operational even do anything for you?

Finally, understand that setting these things up adds significant complexity to your network. How much additional downtime will you incur due to problems or extra maintenance/configuration time needed every time you touch the network? What is the organizations tolerance for downtime?

You might be better off writing a script you can just cut and paste to your 2nd core switch to let it do routing, in the rather unlikely event that your primary fails.