Solved: OSPF looped back interface issue, LSA update??

Sanghee Han · ‎06-22-2009

Hi, all

if router's serial line(ex,45Mbps) have a problem,

we conduct a loopback test for the interface.

before this test,

I delete the ospf config about that interface subnet to prevent data transfer through it.

ex)

router ospf 10

network 10.10.10.1 0.0.0.0 area 0

int se1/0

ip addr 10.10.10.1 255.255.255.0

above configuration,

when I conduct a loopback-test for int se1/0

i delete the network for int se1/0 in ospf

config t

router ospf 10

no network 10.10.10.1 0.0.0.0 area 0

end

but i wonder whether above deleting is need?

when serial interface looped back for testing,

ospf doesn't any LSA update?

for example, LSA update for the interface network.

if it does,

i don't need deleting operation.

anyone knows about this ospf issue?

Giuseppe Larosa · ‎06-23-2009

Hello Sanghee,

OSPF protocol can deal with the loopback test:

the router will simply complain about receiving its own OSPF hello packets back and this can produce some syslog messages but no real issues.

OSPF hello packet contain the OSPF router-id of the sender so the device can safely detect the loop at the OSPF level.

I wouldn't remove the network ... area command actually only traffic for the ip subnet on the serial link could be sent out it during the loopback test if the interface state ip up/up (looped)

No OSPF adjacency can be built on the looped link so no other IP subnets can use it as outgoing interface.

The previous OSPF adjacency fails in 40 seconds (dead interval) and the SPF is recomputed for all destinations taking in account this event.

It can be a good practice to remove the network ... area for the reasons above:

elimination of warning messages in log

reduction of time to detect that no neighbor is available.

But it is not stricly necessary.

Hope to help

Giuseppe

View solution in original post

Giuseppe Larosa · ‎06-23-2009

Hello Sanghee,

OSPF protocol can deal with the loopback test:

the router will simply complain about receiving its own OSPF hello packets back and this can produce some syslog messages but no real issues.

OSPF hello packet contain the OSPF router-id of the sender so the device can safely detect the loop at the OSPF level.

I wouldn't remove the network ... area command actually only traffic for the ip subnet on the serial link could be sent out it during the loopback test if the interface state ip up/up (looped)

No OSPF adjacency can be built on the looped link so no other IP subnets can use it as outgoing interface.

The previous OSPF adjacency fails in 40 seconds (dead interval) and the SPF is recomputed for all destinations taking in account this event.

It can be a good practice to remove the network ... area for the reasons above:

elimination of warning messages in log

reduction of time to detect that no neighbor is available.

But it is not stricly necessary.

Hope to help

Giuseppe

Sanghee Han · ‎06-23-2009

Thanks for your kindly answer, Giuseppe.

as i understanding base on your comment.

router will detect neighbor down with dead-interval during loopback test so it takes maximum 40 seconds.

so the deleting operation is good to avoid a packet loss during dead-interval.

am i right?

i wonder that ospf generates the LSA relative to loopbacked interface network.

for example,

when int se1/0 is looped,

ospf doen't any LSA?

(LSA that notify se1/0 network=10.1.1.1/24 is down)

I can't find the above operation in RFC2328.

Thanks.

Giuseppe Larosa · ‎06-23-2009

Hello Sanghee,

good question:

does the router LSA reports a serial interface up/up looped as up or as down ?

Link state can be up or down there is no additional state.

I'm afraid that OSPF can be fooled, but again because no neighbor is out the link only traffic destined to the IP subnet of the link could be sent out the looped interface.

So as you see this is not a real issue because the looped link cannot become a black hole for user traffic.

When the neighbor is removed all the information is recomputed with SPF.

Hope to help

Giuseppe

Sanghee Han · ‎06-23-2009

Thanks for kindly answer.

Thanks Giuseppe.

Sanghee Han · ‎06-23-2009

there isn't any LSA about int se1/0 down or up.^-^