slow redistribution between OSPF processes

[martap]

Hi All,

Looking for some advice, will appreciate if you can help.

I run two OSPF processes and redistribute prefixes between them. I also run BFD to allow for faster failure detection which works great. What I notice is that in case of a link down, bfd immediately signals OSPF who then withdraws the prefixes. All good so far. However, the other ospf process only seem to "notice" the change about 15 sec later... Redistribution seems slow. Is there anyway to speed this up?

[Georg Pauwen]

Hello,

updates should be immediate, in theory. Hard to say why it takes 15 seconds...can you post the full configs of your OSPF router(s) ?

[martap]

Hi Georg,

Absolutely. see attached.

slowness is seen between ospf1 and ospf100, where we disabled ospf1 interfaces, and monitored ospf100 database. prefixes were removed instantly from ospf1 database but only removed 15 sec later from ospf100 database.

running IOS XE 16.3.6 on a WS-C3650.

[martap]

forgot the attachment....

[Georg Pauwen]

Hello,

I did some testing but in all cases, as soon as OSPF detects the link failure, BFD kicks in and the routing table is updated immediately.

I wonder if it has something to do with OSPF process 100 running on an SVI and not a physical interface. It shouldn't make a difference, but what happpens when you enable bfd for all interfaces under the OSPF process ?

router ospf 1

bfd all-interfaces

router ospf 100

bfd all-interfaces

[martap]

So what I'm doing is the following:

shut ospf1 interfaces

bfd signals ospf1 and routes are instantly removed ("sh ip ospf 1 database" shows that)

"sh ip ospf 100 database" shows the redistributed prefixes for another 15 sec.

I have bfd running on all relevant interfaces individually, including the SVI, but bfd on the SVI should not be an issue here I'd think, because the down link event happens on the ospf1 interfaces not the vl100 interface which is part of ospf100.

[Georg Pauwen]

Can you post the output of 'debug ip ospf sfp' when the link goes down ?

[martap]

Hi Georg, find attached.

thanks for the help...

[Georg Pauwen]

Hello,

where is 10.41.16.13 in your topology ? Actually, can you post a brief schematic drawing of your entire topology ?

[martap]

attached a diagram...

[paul driver]

Hello

Just to confirm, you are looking into the OSPF Database not the rib table correct? -  As you can still have prefixes in the database that are not in the route table?

[martap]

Correct, Paul. Looking into the db of ospf100 for prefixes redistributed from ospf1.

@georg, I’ll do my best to upload a quick diagram tomorrow.

[paul driver]

Hello

There will be some convergence time for each LSA to get updated - Can you post the output from both rtrs

sh ip ospf interface brief
sh ip ospf  interface xxx
sh ip interface xxx
sh ip ospf

[martap]

Paul,

It's the same router, just redistribution between 2 different OSPF processes seem to take up to 15 sec.

attached the output requested.

[martap]

I still haven't figured this one out... Any idea's from you guys?