06-30-2010 02:09 PM
I'm trying to understand the configuration options for routing between two separate EX stacks connected through a MAN Ethernet connection...
Currently... I have two connections, one 200Mbit and a second thats 100Mbit connecting to remote buildings together. I have static routes setup currently on each switch like so:
----Core
static {
route 10.6.0.0/16 {
qualified-next-hop 10.6.1.6 {
metric 10;
}
qualified-next-hop 10.6.1.2 {
metric 100;
}
}
}
--- Remote Office
static {
route 0.0.0.0/0 {
qualified-next-hop 10.6.1.5 {
metric 10;
}
qualified-next-hop 10.6.1.1 {
metric 100;
}
}
}
Which routing protocol should I use to make the link failover between the two lines based on complete link state... I was looking at OSPF and is-is but... from the examples online I'm not seeing how to configure them correctly.
Solved! Go to Solution.
06-30-2010 11:25 PM
Hi,
OSPF should be fine (IS IS needs an extra license). The only thing that we have to take into account is that OSPF needs different interfaces for different links. That means that the primary link should be connected to a different interface as the backup link.
I would suggest, you configure these uplink interfaces as layer 3 interfaces. Let's say you use ge-0/0/0 as the uplink for the primary and ge-0/0/1 as uplink for the backup.
then you sould configure
interfaces {
ge-0/0/0 {
unit 0 {
family inet {
address 10.6.1.17/30;
}
}
}
ge-0/0/1 {
unit 0 {
family inet {
address 10.6.1.5/30;
}
}
}
}
Then you can activate OSPF on these interfaces. We have just on thing to consider: Both interfaces have equal cost from OSPF point of view. That these line are rate limited by the provider is not obvious to the switch/router. We have to manually adjust the metrics:
protocols {
ospf {
area 0.0.0.0 {
interface ge-0/0/0.0 {
metric 10;
}
interface ge-0/0/1.0 {
metric 20;
}
}
}
}
You could also consider using authentication with OSPF. Or you could activate BFD for fast link down detection.
On your other switch, you have to mirror your configuration.
I hope this helps.
Regards,
Dominik
07-01-2010 07:25 AM
Thanks for the detailed reply. How are the proper routes transfered?
the one side only needs to know how to go to 0.0.0.0, but the core side only needs to know that anything 10.6.0.0/16 is accross those links? How does OSPF know about the local networks on the remote side? or is that just something built into it... and its going to work as you put it?
07-01-2010 07:41 AM - edited 07-01-2010 07:42 AM
Hi,
OSPF is a link state protocol. That means, all routers within the same area have to maintain the same graph where the shortest path algorithm runs on.
If you want to filter routers, you can introduce areas different to the root area (id 0), let's say area 1, 2 and so on. So you could put your remote office router/switch in area 1 and your main office router in area 0. The be accorate, you put the interfaces facing the remote side into area 1 in the main office and you have to assign at least one interface into area 0, possibly a loopback interface.
If a router belongs to more then one area, it is called an area border router (ABR). On an ABR you can do filtering. If you want that the router in the rmeote office receives only a default route, area 1 would be called totally stub area. On the other hand, in area 0 you will still find the routes of area 1.
To explain everything in detail would exceed the possibilities of this thread. But you should be able continue working with this.
Here is a short example:
protocols {
ospf {
area 0.0.0.0 {
interface lo0.0 {
passive;
}
}
area 0.0.0.1 {
stub default-metric 10 no-summaries;
interface ge-0/0/0.0 {
metric 10;
}
interface ge-0/0/1.0 {
metric 20;
}
}
}
}
On your router in the remote area, you have to just mention the stub statement, without the default metric and no-summaries verbs.
I hope this helps a bit.
Regards,
Dominik
07-10-2010 11:31 AM
I did this... however I never saw any routes get pushed to either side... how are the routes published?
07-10-2010 11:36 AM
Does the OSPF adjacency get established? You can verify this by issuing a show ospf interface command in operational mode.
Regards,
Dominik
07-10-2010 03:39 PM
This is what I have currently on both switches:
> show ospf interface
Interface State Area DR ID BDR ID Nbrs
ge-1/0/1.0 DR 0.0.0.1 10.2.1.254 0.0.0.0 0
ge-3/0/8.0 DR 0.0.0.1 10.2.1.254 0.0.0.0 0
> show ospf interface
Interface State Area DR ID BDR ID Nbrs
ge-0/0/0.0 DR 0.0.0.1 10.6.1.2 0.0.0.0 0
ge-1/0/2.0 DR 0.0.0.1 10.6.1.2 0.0.0.0 0
07-11-2010 01:38 AM
Hi,
first, there must always be a root area in OSPF (=area with id 0). Different areas with id > 0 can be created,e.g. to reduce the size of the link state database, to filter out routes etc. As it seems you don't have other OSPF enabled interfaces, please change your area id from 1 to 0 on both routers.
You can see from the column neighbor, that no interface has established an adjacency to its neighbor. If this remains the same once you change the area id to 0 we can do the following checks:
I assume, that the virtual chassis stacks are connected this way:
VC1 VC2
ge-1/0/1 <-> ge-0/0/0
ge-3/0/8 <-< ge-1/0/2
First verify the link between the two interfaces by using ping and specify the IP address of the remote if as target and the local IP address of the link interface as source. And vice verse, for both links.
Then verify the MTU on both interfaces. They must match so that a adjacancy can be formed. The MTU size is displayed when issuing a show interface ge-x/y7z command in the third line.
If even these three steps don't help, we have to make use of the traceoptions. Issue these commands on both VC stacks and then post the content of the trace file:
set protocols ospf traceoptions file ospf.debug
set protocols ospf traceoptions flag hello detail
set protocols ospf traceoptions flag error detail
Regards,
Dominik
07-11-2010 08:10 AM
I think we are almost there...
mitcha@mkt-core# run show ospf route
Topology default Route Table:
Prefix Path Route NH Metric NextHop Nexthop
Type Type Type Interface Address/LSP
10.6.1.2 Intra Router IP 10 ge-1/0/1.0 10.6.1.6
10.6.1.0/30 Intra Network IP 20 ge-3/0/8.0
10.6.1.4/30 Intra Network IP 10 ge-1/0/1.0
I'm seeing some routes now... they aren't showing up in the actual routing table... nor is the default route showing up on the remote switch...
mitcha@pnt-sw01# run show ospf neighbor
Address Interface State ID Pri Dead
10.6.1.1 ge-0/0/0.0 Full 10.2.1.254 128 34
10.6.1.5 ge-1/0/2.0 Full 10.2.1.254 128 32
{master:0}[edit]
mitcha@pnt-sw01# run show ospf interface
Interface State Area DR ID BDR ID Nbrs
ge-0/0/0.0 DR 0.0.0.0 10.6.1.2 10.2.1.254 1
ge-1/0/2.0 DR 0.0.0.0 10.6.1.2 10.2.1.254 1
{master:0}[edit]
mitcha@pnt-sw01# run show ospf route
Topology default Route Table:
Prefix Path Route NH Metric NextHop Nexthop
Type Type Type Interface Address/LSP
10.2.1.254 Intra Router IP 10 ge-1/0/2.0 10.6.1.5
10.6.1.0/30 Intra Network IP 20 ge-0/0/0.0
10.6.1.4/30 Intra Network IP 10 ge-1/0/2.0
The correct connection topology is:
VC1(mkt-core) VC2(pnt-sw01)
ge-1/0/1 <-> ge-1/0/2
ge-3/0/8 <-< ge-0/0/0
mkt-core having many 10.x/16 networks on it including being the gateway out... and pnt-sw01 having the 10.6.0.0/16 network, with the mkt-core switch being its gateway.
I'm adding the trace options, I will post those results in a bit.
07-11-2010 08:13 AM
Hi,
if the routes your are missing are e.g. static routes, you won't see them on the other stack unless you export them into OSPF that transport it to its neighbors!
I can give you an example if you tell me what routes you like to leak into OSPF. All static ones?
Regards,
Dominik