Reading into OSPF again after not touching it for a while.
From what I'm reading - Stub area, allows LSA 1 and 2 within the stub area and the ABR sends LSA 3 into the stub area. This LSA 3 is a default route - 0.0.0.0 /0 pointing to the ABR
A Totally stubby area is the same, LSA 1 and 2, but instead of LSA 3 - ABR injects a default route of 0.0.0.0 /0 pointing to itself?
Is this not exactly the same outcome for both areas? I'm obviously missing something here
Type 3 - Summary LSA - an Area Border Router (ABR) takes information it has learned on one of its attached areas and summarizes it before sending it out on other areas it is connected to. This summarization helps provide scalability by removing detailed topology information for other areas, because their routing information is summarized into just an address prefix and metric. The summarization process can also be configured to remove a lot of detailed address prefixes and replace them with a single summary prefix, helping scalability. The link-state ID is the destination network number for type 3 LSAs.
Logically it is correct, the stub has no where else to go apart from one exit point in the area. Why else would it be stub? It does not need to know a summary of other routes, instead the ABR generates a summary LSA with the link state ID of 0.0.0.0 - associated with a default exit point.
OK Kinda got it now after reading here with the examples
Why would you use a STUB over a Totally Stubby Area though? The TSA looks to do everything a stub does but more, and use less resources
It provides other areas information/routes, actually its not a summary route. All routes are there by default. but you can configure it as a summary.
secondly, in sense of summary mean these LSA's have not detailed information, only summarized.
Basically if you are going to configure an ospf area as a stub and prohibit E1/E2 routes which will shrink the opsf database, Then why not go further and make it a totally stubby area which will also prohibit OIA routes and shrink the D/B even more so:
sh ip ospf database database-summary
Once thing you need to be aware of is if the ospf adjececies hellos are oritignating from a interarea/external prefix when you make the area a stub/totally stub the adjacency's will fail.
area x stub - abr injects default route (stub area no lsa 4-5 = E1/E2 routes
area x stub no-summary - abr (no summary) injects default route ( totally stuby area - no lsa 3-4-5) = OIA/E1/E2 routes
(the only type LSA 3 allowed is the injected one by the abr but the abr will stop all other lsa type 3 from other areas)
Still a bit unclear, sorry :-(
So for a Stub - could there be multiple different LSA type 3 injected into the STUB? various summaries? and with the totally stubby there is only ONE route - 0.0.0.0 /0?
I believe that your statements are correct. In a stub area there are likely to be multiple LSA type 3 entries providing routing information about networks and subnets in other areas of the OSPF routing domain (and no routing entries for routes external to the OSPF routing domain other than the default route). In a totally stubby area there are no LSA type 3 summaries from other areas. The only routing entry for destinations outside the area is the default route.
That's the understanding I came to once reading more so thanks for confirming. I guess my next logical question would be, why would you ever use a STUB over a Totally Stubby Area?
Let me begin my response with a rhetorical question "Do you believe that when you have more information available that you make better decisions?"
I think that most of us would agree that more information leads to better decisions. So how does that apply to OSPF and stub vs totally stub areas?
I suggest that stub areas present more information (about networks and subnets outside the area and within the OSPF routing domain) while the totally stub area presents no information about any resource outside the area.
So a stub area requires more resources and may lead to more optimum routing decisions while a totally stub area requires fewer resources and may lead to some sub optimal routing decisions. So in designing a network you choose one or the other depending on what you believe satisfies the requirements better.
One particular scenario that I've come across.
Site A (third party) is connected to the enterprise network via WAN router. This WAN router is in its stub area, it has the following static routes and OSPF route to its internal organisational networks and default route for outbound.
*O IA 0.0.0.0/0
A PC in this third party site needs to get to 10.1.1.1 (server A which exists in the enterprise network)
If we had totally stub, we have one default route out. In which case we could potentially have a routing loop. This WAN router has a more preferred route back into its internal network. Why? there is a longer match prefix (10.0.0.0/8) - As Rick pointed out, It's making the wrong decision.
If this was perhaps a stub area we could "leak" routes (using very loosely - allowing type 1,2 and 3) to other inter area routes, therefore we have this in the routing table...
*O IA 0.0.0.0/0
In this case it will send out the packet correctly taking the OSPF path out rather than back in to its local network, as a result of having the extra information required to make these types of scenario's work.
Yes if you assume that a stub area has only one way out then the extra information does not provide much value. And a number of people do seem to have the idea that a stub area should have only a single exit. That may be the case sometimes but it is certainly not a requirement. When I configure networks with OSPF I always try to make sure that every area, including stub areas, has at least two exit points. And with at least two exit points the extra information may provide value.
I have configured multiple OSPF stub areas with at least two exit points and I assure you that this works fine.