Loopback addresses? 1

[drewdown]

I have more of a concept question then an actual problem.

When configure lookpback addresses for network devices is there a typical rule you follow? Or does anything go?

IE Lets say I have a network that consists of 2 main offices and 4 satellite offices. Total of roughly 8 routers, any suggestion on how to number the loopback interfaces? IE all on the same subnet? Seperate by location?

Thanks in advance.

 

[rodjor]

What is your reason for using the loopback addresses? For OSPF purposes?

 

[drewdown]

Yes sir.

 

[burtsbees]

You'll need to decide on a DR and BDR, and that depends on your topology. You could force them by priority, but I prefer the loopback myself, as it serves as an "always up" management interface. The highest IP will be the DR of the area...does that sound right?...crap...can't remember now...

Burt

 

[drewdown]

OSPF is already in place and running but I am using physical interfaces as the router-id's. I have the DR/BDR set with priority's as well, but I don't want the router-id's to be acutal physical interfaces anymore.

Only in the event of an identical OSPF priority will the highest RID (can be loopback) be used to select the DR.

Now I just need to come up with an addressing scheme and am hoping someone can help me out. Best practice would be /24 or /32 addresses per device?

 

[lerdalt]

We use /32 addresses on our loopbacks. We've chunked off an entire class C just for loop backs..not that we are using it all, but saved room for growth.

One thing we've done too is have more or less a site number (in many cases matches the subnet). For example. The site that has address to clients of 10.0.32.0/24, the router might have a host name of xx32rtr and loopback of 10.0.0.32/32.

 

[drewdown]

Thanks man.

So if you had 2 routers at the same location servicing the same network would you use a seperate subnet for the other router or just another /32 in the same loopback subnet?

 

[Belushi]

Separate subnet for the loopback interfaces is necessary. You don't want it to be a part of another subnet containing other addresses. I agree with the stance of using a /32 mask for loopback interfaces although there is certainly no reason for it to be that way functionally. My network has thousands of devices and we carved out quite a few address spaces for loopbacks, all of which are displayed to the network as /32 host routes.

Got some strange responses here I wish to address:

1. Not sure how DR/BDR election got involved in the conversation. You made no mention of OSPF network types which wold facilitate the need for those. Personally, I hate using network types that require DR/BDRs unless absolutely required but to each his own.

2. Best practice according to Cisco is to always specify router-id for the various routing protocols BGP and EIGRP as well, not just OSPF. Not that it won't work without them, but it is best practice and a good administrative policy.

 

[burtsbees]

With all the loopbacks as /32 addresses, they are all in their own "subnets"...

Burt

 

[ADB100]

It all depends on the size of the OSPF network and whether you have multiple areas or not.

If you have split your network up into areas and assigned IP Address space to specific areas AND summarised address space at ABRs then the IP addresses of the Loopbacks need to be area specific. If you have a single Area (0) then the IP addresses of the Loopbacks are not that important (they need to be routable but not within a specific summary).

It's a design thing..... Loopbacks typically have /32 masks and are routable. If you have multiple areas then the major network or summary they belong to is relevent, if you have a single area then this isn't that important. If you are designing for the future then it might be?

HTH

Andy

 

[drewdown]

Thanks guys.

1 single area at this point, but may change in the future.

I configued loopbacks for my 4 main devices.

172.233.201.1/32
172.233.202.1/32
172.233.203.1/32
172.233.204.1/23

All routebale via OSPF.