Mmmm, pistol whip.
https://www.youtube.com/watch?v=fH185Hn38m8

OK, I’m guessing that wasn’t what you had in mind?


From: George Skorup 
Sent: Sunday, February 28, 2016 8:11 PM
To: [email protected] 
Subject: Re: [AFMUG] Mikrotik OSPF route selection oddness

I think you said it best...

"sometimes RouterOS rides the short bus"

You'd really need to look at the LSA table, OSPF route table and the system 
routing table to figure out what was really going on.

The thing that used to drive me insane was a BGP session going down would 
withdraw that LSA as expected, but the routers using that wouldn't inject ANY 
other default route in their routing table, even though they existed in the LSA 
and OSPF route lists. I'd have to bounce an interface so that SPF would 
recalculate and reconverge the entire network. I thought I had something 
configured wrong for the longest time. No, just RouterOS shenanigans. It needed 
a pistol whip. (if anyone gets that reference, +1 to you)


On 2/28/2016 7:48 PM, Adam Moffett wrote:

  The two routers saw each other as neighbors.  
  I did not look closely at the LSA....but why would OSPF routes not get 
injected into the routing table? 


  On 2/28/2016 2:48 PM, George Skorup wrote:

    That would be my question too. Were the adjacencies up full at the time and 
not stuck in 2-way or something like that? If they were full but no traffic, my 
money would be on OSPF routes simply not getting injected into the routing 
table.

    Pretty much all of my OSPF issues went away after moving from 5.25/26 to 
6.32.2. However, I did have a similar situation a week or so ago. No traffic 
across a link for several hours. And the link was definitely up. It happened in 
the middle of the night, so I didn't get a chance to see what was going on. It 
just fixed itself. But the routers showed no adjacency loss. So I have no idea.


    On 2/28/2016 10:50 AM, Darin Steffl wrote:

      Have you checked to see if there is still an OSPF connection running over 
this licensed link? As in do you see the neighbor router listed in OSPF for the 
other side of the link?

      On Sun, Feb 28, 2016 at 10:18 AM, Adam Moffett <[email protected]> 
wrote:

        I have a 500mbps licensed microwave path that was up and looking 
perfectly fine, but had no traffic on it.
        Instead traffic was moving on a 5.8 Radwin to a different site. 
Everything was up, but I don't understand this route selection.  The path via 
the Radwin is longer to lots of places, so OSPF should have used the licensed 
path for at least SOME traffic simply based on it being shorter.  I also set 
the costs on the interfaces on both ends of that Radwin to 150, so it normally 
doesn't get used at all. Somehow this morning at 3:15am the routers decided to 
use the backup path.  Everything returned to normal after I rebooted one end.

        It may be coincidental (or not), but I know a truck took out a 
telephone pole last night which caused an upstream fiber issue, and we had a 
UPS battery explode.  Is this all coincidence?  Sometimes UPS's explode, 
sometimes That One Guy Steve burns down a house, sometimes RouterOS rides the 
short bus, some things are just outside our control.  Or maybe a truck hit a 
pole, which caused a power surge that blew up a UPS and made a Mikrotik fart 
inside of it's head.  I could believe either one.

        In any case.  Is there a different RouterOS version that will trust my 
opinion when I tell it a path has high cost?  Currently running 6.32.3 I think. 
 It's pretty non-sensical that this happened at all.






      -- 

      Darin Steffl 
      Minnesota WiFi
      www.mnwifi.com
      507-634-WiFi
       Like us on Facebook





Reply via email to