Is the customer connecting via PPPoE? That would generate a /32 route that is 
being advertised.

No.  There is a hotspot involved, but we’ve already tried bypassing that and no 
difference.

Any routing marks involved?  Any mangle or firewall rules which would match 
this customer and not others?

No

When you're saying "traceroute stops at", is that the site giving the last 
reply?

Yes

One other thing....I had a couple of anomalies go away when I switched OSPF 
interfaces from broadcast to point-to-point mode.  I thought the only 
difference was the OSFP messages are sent unicast, but a few goofy things 
stopped happening when I did that.

That’s something I hadn’t thought to try.
Would I need to do it on all interfaces involved? or just from Tower B to Tower 
A? (that’s where the issue seems to be, even though I can still reach the 
customer from Tower B, just not the NOC)

Thanks,
Justin

From: Af [mailto:[email protected]] On Behalf Of [email protected]
Sent: Friday, March 31, 2017 1:34 PM
To: [email protected]
Subject: Re: [AFMUG] Strange OSPF problem

Is the customer connecting via PPPoE? That would generate a /32 route that is 
being advertised.

On Fri, Mar 31, 2017 at 1:27 PM, Adam Moffett 
<[email protected]<mailto:[email protected]>> wrote:
Any routing marks involved?  Any mangle or firewall rules which would match 
this customer and not others?

When you're saying "traceroute stops at", is that the site giving the last 
reply?

What if you specify different source IP's on your traceroute from the NOC?  
Maybe tower A is missing one route, but not others.

One other thing....I had a couple of anomalies go away when I switched OSPF 
interfaces from broadcast to point-to-point mode.  I thought the only 
difference was the OSFP messages are sent unicast, but a few goofy things 
stopped happening when I did that.


------ Original Message ------
From: "Justin Marshall" <[email protected]<mailto:[email protected]>>
To: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>>
Sent: 3/31/2017 12:55:43 PM
Subject: [AFMUG] Strange OSPF problem

Hi,

I have an odd situation with OSPF.

There are 2 towers (Tower A and Tower B) and a NOC involved that feeds both 
towers.
Tower A is directly connected to the NOC via Mikrotik Backhaul.  Tower B is 
directly connected to the NOC via Fiber.
Tower B is connected to Tower A via Fiber.

When tower A is running straight off the NOC (via Mikrotik Backhaul) everything 
works as it should.

We are trying to make a switchover to have Tower A run off Tower B so it will 
be a Fiber connection all the way to Tower A (through tower B) and eliminate 
the Mikrotik Backhaul.

When I change cost on the OSPF interfaces to make this happen, all customers 
work as they should and following the normal path to/from the Internet as 
expected.  However, for one of the customers on Tower A, traceroutes to that 
one customer stop at Tower A (through tower B) from the NOC.  Traceroutes 
towards the NOC, from the customer stops at Tower B.

All customers are on the same subnet.

I can directly ping the customer in question from both Tower B and Tower A (and 
the traffic is taking the correct path), but not from the NOC.  From the NOC 
all traffic stops at Tower B

The only real difference for the customer that is not working is they are using 
a Sonic Wall, and the other customers have Mikrotik routers behind the CPE's.  
We have tried elimating the sonc wall, and replacing it with a laptop for 
testing, and the traffic flow failed in the same way.

Tried all kinds of things and we are quite stumped.

Anyone have any suggestions?

Thanks,
Justin
[email protected]<mailto:[email protected]>



Reply via email to