Large number of routes observed in holdown state.

lab> show route protocol bgp inactive-prefix

inet.0: 5317 destinations, 5318 routes (5314 active, 700 holddown, 0 hidden)

inet.3: 5305 destinations, 7115 routes (1808 active, 1000 holddown, 10 hidden)

JPR.inet.0: 596 destinations, 1200 routes (593 active, 800 holddown, 0 hidden)

+ = Active Route, - = Last Active, * = Both

20.0.0.1/30 [BGP/170] 00:00:01, MED 1, localpref 100, from 144.133.176.6

 AS path: I, validation-state: unverified

 Indirect

20.0.0.2/30 [BGP/170] 00:00:01, MED 1, localpref 100, from 144.133.176.6

 AS path: I, validation-state: unverified

 Indirect

20.0.0.3/30 [BGP/170] 00:00:01, localpref 100, from 144.133.176.6

 AS path: I, validation-state: unverified

 Indirect

holddown (routes that are in the pending state before being declared inactive)--> A holddown route was once the active route and is no longer the active route. The route is in the holddown state because a protocol still has interest in the route, meaning that the interest bit is set.

A persistent holddown state often means that the interested protocol is not releasing its interest bit properly.

However, if you have configured advertisement of multiple routes (with the add-path or advertise-inactive statement), the holddown bit is most likely set because BGP is advertising the route as an active route. In this case, you can ignore the holddown state because nothing is wrong.

If "advertise-external" statement is configured, it mandates BGP to advertise the best external route into an IBGP mesh group, a route reflector cluster, or an AS confederation even if the best route is an internal route.

It means internal route is the best and we have interest in it, however we’re advertising external route. Although we have not isolated exact routes which were marked as holddown, we are confident that this scenario is where we can ignore holddown state, because nothing is wrong.