RSPF routes that linger on

[klemets@sics.se]

> all the routes were still there; very old but
> still there. ROUTE confirmed this and an RSPF STATUS showed all my
> adjacencies as TENTATIVE rather than BAD as I would have expected.

Interesting observation. When the suspect timer for an adjacency
expires, it goes from the OK state to SUSPECT. RSPF tries to ping the
adjacency a number of times as specified by the maxping command. If
the pings fail, the adjacency will enter the BAD state, where it will
remain for one supect-time interval until it is deleted from the
adjacency and routing tables.

Now suppose that you have more than 5 adjacencies. If you just unplug
your TNC, they might all become suspect within a short period of time.
RSPF imposes a limit on the number of processes it will create to ping
suspect adjacencies. This limit is defined by RSPF_PROCMAX which set
to 5. If an adjacency becomes suspect while there are already 5 other
adjacencies being pinged, it will not enter SUSPECT state but rather
the TENTATIVE state. And I forgot to reenable the timer in this
particular case, so the adjacency will remain in that state until
maybe you plug in the TNC again.
I guess this must be the problem you have experienced.

Anders