Re: [mpls-linux-general] a bug ?

[James R. L. <jl...@mi...>]

On Wed, Jun 27, 2001 at 11:36:29AM -0000, Venisa Cabrilla wrote:
> Jim,
> 
> I use the third router(lo 10.0.0.7) directly connected to router2 (lo 
> 10.0.0.6) to support my argumentation here related to my last posting 
> ("another problem in ldp_linux").
> 
> It seems that ldp_linux works fine if I run and configure it from interface 
> (lo) whose address is smaller. In my case of using three routers, ldp_linux 
> works well, if and only if I start configuring from router1(lo 10.0.0.5) to 
> router2(lo 10.0.0.6) and then from router2(lo 10.0.0.6) to router3(lo 
> 10.0.0.7).
> 
> Am I right ? Do we always have to do like that ?

There must be a bug with ldp_linux.  LDP determins who will be "active"
based on the LSRID (router ids).  The one with the higher address is active.
When you start them in the "wrong" order the LSR that is active hasn't
had time to send a hello yet.  Thus the 'passive' LSR doen't know of it's
existence and when the 'active' LSR tries to setup the TCP session it fails.

What should happen next is that the 'active' LSR goes into backoff state for
about 10 seconds, then tries again.  If your not seeing this then something
is broken.

(BTW I had this tested and working before my recent TCP over haul it must
be a bug in there)

I'll take a look at it tonight and see if I can fix it.

Thanks for the bug report!

Jim
-- 
James R. Leu