[CUWiN-Dev] Troublesome Bouncing Default Route

[David Young]

On Wed, Apr 05, 2006 at 03:10:44PM -0500, Rob Simmons wrote:
> It happened again. I only got /var/log/daemon from one of the two places (I
> second guessed myself and thought I was looking for /var/db/vizlinks
> instead, but their default route was looking at each other - I was accessing
> through another node in between that was unable to connect to the internet
> as a result of the bouncing default route. All nodes were able to access the
> cuwin network, none of them were able to access anything off-network. The
> two nodes in question were 10.0.233.21 and 10.0.233.54.

In 233.54-daemon, I see this change of route:

Apr  5 00:06:35 cuw hslsd: SPF HSLS_RIB_CMD_CHANGE dst 10.0.233.21/32 nexthop 169.254.241.172 metrics [(etx, 778), (hopcount, 2), (backplane-hopcount, 0), (network-radius, 2)]

If 233.54's nexthop for 0/0 is 233.21, then there is an inconsistency
in the routing, because we know there is at least one route to 0/0 that
is shorter than 233.54 -> 233.21 -> ... -> 0/0, and that is the route
233.54 -> 241.172 -> 233.21 -> ... -> 0/0.

I would like to see the HSLS_RIB_CMD_ messages where the 0.0.0.0/0 route
is added/changed.  Also, I would like to see the routes according to
Zebra---maybe there is an inconsistency between Zebra and the kernel?
I would also like to see the SPF computation (add an -l spf option
to hslsd).

Dave

-- 
David Young             OJC Technologies
dyoung at ojctech.com      Urbana, IL * (217) 278-3933