Hello P.B,

Thanks for your response.

You are right, I desperately want to avoid adding further complications so will stick to using a single area for now.

I was trying to design a scalable solution but dont want to have to break the existing ospf config on all the other sites that work by making changes there.

Cheers,

Martha.

Hi Giuseppe,

Thanks so much for taking the time to respond to me.

As advised, I will keep the edge devices in area 0 for now and consider multi-area ospf at a later stage.

Cheers,

Martha

Sorry Giuseppe,

one quick question regarding what you wrote about implementing a multi-area domain. Currently all the distribution switches, the core devices and the core devices at all the other 5 sites are in Area 0.

If I wanted to make the edge switches and all the other site core devices (WAN links) into a different area and make the core an ABR so both areas can be seen, could I do that?

Please give some pointers as to how I can configure this.

I think I might just stick to the single area but I need to convince my boss that this is the best solution for now or at least show how we can alternatively configure multi-area domains but explain taht we should probably do the less complicated implementation for now.

I appreciate your help.

Regarding the use of 4 point-to-point routed links to the sister sites (if thats what you meant), that is not possible as the inter-site connections go into a separate switch and there is a trunk port connecting this to the edge switch.

Much obliged,

Martha.

Hi Giuseppe,

You are absolutely fantastic. I really appreciate you taking the time to answer all these questions.

It is really difficult to get downtime so I will try to make as few changes as possible.

The idea of putting all the site edge switches (WAN links) in area 0 appeals to me but i am not in a position to change any but the main site at this point.

I will add the additional area to the core devices so they can be easily changed over when I get to that stage.

I will also read the document on the link you sent.

At the moment, there is only a sole switch for the intersite connections but I cant see anyway around that.

Again, many thanks.

Martha.

Hi Giuseppe,

This should be the last email on this subject, honest :)

Just to be sure I understand your advice, please review how I have currently scripted my changes.

on the Core router

(recall this is currently in Area 0 and has connection to the 5 other sites whose networks are also currently defined in Area 0. The 6 distribution blocks are defined as stub areas in separate areas 100-105)

router ospf 1

log-adjacency-changes

area 100 stub

area 101 stub

............. (all other DB's)

network local loopbk int 0.0.0.0 area 0

network other local int 0.0.0.0 area 0

network ext site 1 0.0.0.0 area 0

network ext site 2 0.0.0.0 area 0

network (other ext sites) 0.0.0.0 area 0

network DB1 0.0.0.3 area 100

network DB2 0.0.0.3 area 101

.............. (all other DB's)

Now I need to add another area to this core device and make it the ABR with links in both area's??

As the core routers are still part of area 0, presumably the distribution blocks are still fine and I have not formed any contiguous networks??

Please let me know if I have understood correctly, I would as advised like the core routers at every site to eventually be in a different area and to have the edge devices at all sites in Area 0 but as we have designed the DB's in separate areas, I hope this does not complicate things.

Many thanks,

Martha.

Hello Martha,

>> Now I need to add another area to this core device and make it the ABR with links in both area's??

Actually, a router becomes an ABR when it has multiple network .. area commands specifying two different areas.

your core is already:

ABR between area 0 and area 100

ABR between area 0 and area 101

and so on.

You don't need to do else in OSPF.

note:

the most important action even before using OSPF multi-area is to remove the single point of failure represented by that switch terminating all links to other sites.

This is really important and needs to be done to make a good job.

So I would introduce a section on migrating these links to new edge nodes

It may take two/four hours of night time to do that but without this the design is not fault-tolerant.

Hope to help

Giuseppe

Thanks again Giuseppr, I'll work on getting them to fork out for an additional switch.

Would you advice changing the distribution blocks so they are all part of the same area e.g change all to 100(as I mentioned, there are 6 DB's, each connected to about 4 or 5 switches per cabinet) and the have the core as an ABR for this this single amalgamated area plus the backbone - area 0?

Each cabinet does not need to know about the other and that is whay I have made them stub areas. I need to ensure any change I make does not mess things up.

Cheers again.

Hello Martha,

>> Would you advice changing the distribution blocks so they are all part of the same area e.g change all to 100

yes, this is reasonable otherwise the work for the ABR is 6 times worse:

it is common to use a single stub area for a site.

To be noted that even if you filter inter-area routes using totally stub areas the default route allows communication between the different cabinets.

you may want to use a single totally stub area to keep area 100 database size to a minimum.

in this case you need

area 100 stub no-summary on ABR nodes ony

the distribution switches need

area 100 stub

only

in the future you can deploy other stub areas in the other 5 sites.

Hope to help

Giuseppe

I apologize for intruding on this conversation, but I am wondering one thing. If the OP had connections to the Internet from each independent location, making the other areas stubs would not be a wise choice am I right?? The reason I ask this is the default route that gets installed on the stub routers would forward all traffic back over and through the backbone when it may actually need to go out via the Internet connection. This is more of a hypothetical design question more so than a question related directly to this issue. Please tell me if I'm understanding this correctly. Thanks.