MPLS route is still active in routing table even its corresponding OSPF route is not listed


Hi There,

I found MPLS route is still active in routing table even its corresponding OSPF route is not listed in routing table.

Ori Destination Gateway Mtr Flags VLAN Duration
mp 192.17.80.55/32 172.17.79.1 1 U--D---um-L-- CR1-DR1_902 0d:2h:57m:41s
mp 192.17.80.55/32 172.17.79.85 1 U--D---um-L-- CR1-DR1_782 0d:3h:37m:45s
#or 192.17.80.55/32 172.17.79.85 12 UG-D---um--f- CR1-DR1_782 0d:3h:37m:46s

Even increasing the OSPF cost of the link, its mpls route is still active in routing table and traffic is going via that link.

How can I increase the metric of mpls route ?

Best Regards,
Ronie

11 replies

Userlevel 6
The problem is with your route priority being that they are both the same subnet mask. You can change the route priority with the "configure iproute priority" command.

Switch# show iproute priority
Direct 10
MPLS 20
Blackhole 50
Static 1100
ICMP 1200
EBGP 1700
IBGP 1900
OSPFIntra 2200
OSPFInter 2300
Isis 2350
IsisL1 2360
IsisL2 2370
RIP 2400
OSPFAsExt 3100
OSPFExt1 3200
OSPFExt2 3300
IsisL1Ext 3400
IsisL2Ext 3500
Bootp 5000
Userlevel 7
I don't understand the question. In your output, only the OSPF route is selected and installed in the FIB.
Let me elaborate. Even only the single OSPF route is there, the path is chosen based on the mpls routes. There are two mpls routes as you can see in the output shared above, the first mpls route is taken as primary active path (its corresponding OSPF route is not even shown in routing table).

Is that due to the reason the cmd "enable ospf mpls-next-hop" is applied on the node ? Your further explanation on this would be highly appreciated.
Userlevel 7
Oh, then yes, I think so, you are telling the OS to use the MPLS LSP with that command. Check the User Guide (p.1222 in the 22.6 version) for "OSPF Calculated LSP Next Hops" for more details.
The question being is why following mpls route is shown in routing table in the first place ?

mp 192.17.80.55/32 172.17.79.1 1 U--D---um-L-- CR1-DR1_902 0d:2h:57m:41s

Since there is only one ospf route (best route) is selected, there should be only one mpls route with the same next-hop (as in ospf route).

Your further explanation on this would be highly appreciated.
Userlevel 7
I double-checked with a real expert.

There are no OSPF Calculated routes here. If there were, there would be a lower case “L” instead of the uppercase one. MPLS next-hops may even not be turned on (since neither MPLS route is selected for routing, but do have a lower cost than the OSPF route). Can you confirm that?

The MPLS routes are ECMP here. This could be because OSPF reports both as equal cost, which I guess isn’t the case, since OSPF only shows one route, or it could also be due to adjacent nodes. Is CR1-DR1_902 connect directly to 192.17.80.55 from this node?

Can you share the output of "show mpls ldp peer detail"?
Userlevel 7
Grosjean, Stephane wrote:

I double-checked with a real expert.

There are no OSPF Calculated routes here. If there were, there would be a lower case “L” instead of the uppercase one. MPLS next-hops may even not be turned on (since neither MPLS route is selected for routing, but do have a lower cost than the OSPF route). Can you confirm that?

The MPLS routes are ECMP here. This could be because OSPF reports both as equal cost, which I guess isn’t the case, since OSPF only shows one route, or it could also be due to adjacent nodes. Is CR1-DR1_902 connect directly to 192.17.80.55 from this node?

Can you share the output of "show mpls ldp peer detail"?

Hi,

I would expect the flags Lf for an active IP over MPLS route (using enable iproute mpls-next-hop) as in the following example:
#mp 10.0.0.2/32 10.1.0.2 1 U--D---um-Lf- t-s2-s3 0d:0h:1m:22s[/code] oa 10.0.0.2/32 10.1.0.2 14 UG-D---um---- t-s2-s3 0d:0h:14m:58s[/code]
It seems to me as if only the OSPF route is used for IP forwarding in the question above.

Thus the problem seems to be more that there are two LSPs available and possibly advertised via LDP, while only one LSP corresponding to the least cost path is expected.

This leaves the question of why there is the other LSP corresponding to a non-least cost path.

Thanks,
Erik
Yes, mpls-next-hop is turned on as follow.

DR1.6 # show configuration | inc next-hop
enable ospf mpls-next-hop
DR1.7 #

And also CR1-DR1_901 is directly connected to 192.17.80.55 but its ospf cost is set to 40 where the primary link CR1-DR1_782 is set to 2.

DR1.7# show ospf interfaces

VLAN IP Address AREA ID Flags Cost State Neighbors

CR1-DR1 172.17.79.86 /30 100.0.0.0 -rif--- 2/C P2P 1
CR1-DR1 172.17.79.5 /29 100.0.0.0 -rif--- 40/C P2P 1

DR1.9 # show ospf neighbor | inc 192.17.80.55
192.17.80.55 1 FULL /DROTHER 00:08:09:59/00:00:00:09 172.17.79.85 CR1-DR1
192.17.80.55 0 FULL /DROTHER 00:00:07:09/00:00:00:03 172.17.79.1 CR1-DR1

This must be a bug and Extreme really need to fix this.
Userlevel 7
Hi,

I'm not GTAC, so if you think there's a bug that needs to be fixed, you should open a case.
Can you please share the output of "show mpls ldp peer detail"?
OK. If you say so. I used to contact GTAC but now, I am really fed up with their support. They kept asking to provide output of cmd again and again. I better save time by looking workaround solution.

Due to the privacy, I can't share the output you are asking for. Thank you for your help till here.

Best regards,
Ronie
Userlevel 7
You need to look at that output to check if there's no LDP LSP between the adjacent nodes. You may have one created, explaining your behavior.

Reply