RSPF hangups

[Anders.Klemets@CS.CMU.EDU]

> I thought the reason for using rspf was that the size of routing
> updates would be reduced and occur only when necessary. Our
> experiance around here is that rspf fires off a full routing update
> every 900 secs (the setting of rspf timer), regardless of whether the
> state of the routes has changed or not.  We tried using rspf mode vc
> locally in the hope that this might reduce the verbage; it didn't.

900 seconds is not really an optimal setting of the RSPF timer in your
case. I would recommend you to try 3600 seconds or more. This will cause
one update to be sent every hour regardless off whether the state has
changed or not. However, if you would get or lose routes, this will be
reported in special good/bad news routing updates. These updates contain
no information other than about the routes that have actually changed.

If you use "rspf mode vc", the updates will be sent using connected mode
AX.25, and this will of course increase the amount of traffic. I would
still recommend this however, because otherwise a routing update may get
lost, which is undesirable if the updates are sent only once an hour.

When it comes to your other problems, I can only say that I have not
observed them here. I most certainly have noticed how routes disappear
from the routing table. But this only happens when an adjacency goes
from the SUSPECT state (and maybe also TENTATIVE state) to BAD, and
managed to remain in BAD state for some 30 seconds.

Anders