OSPF fast hello query

tinu_karkitinu_karki Member Posts: 21 ■□□□□□□□□□
Had a question regarding OSPF fast hello. Lets say i configure fast hello on an interface using the command:

ip ospf dead-interval minimal hello-multiplier 5



This means my dead interval is 1 second (this value never changes)

This also means 5 hellos are sent out within 1 second which equals a hello message every 200ms



I have two question

1. Firstly dead interval doesnt follow the normal behaviour of 3 times the hello interval, with fast hellos dead interval is set to 1 second irrespective of hello interval - is this correct?



2. If dead interval is set to 1 second the what is the point of specifying the hello multiplier ie, does it really matter what the multiplier is set to? Even if it set to 100ms per hello or 200ms per hello dead interval will not happen until 1 sec.

Comments

  • SettSett Member Posts: 187
    The hello and dead interval are independent timers, there isn't some formal ratio between them. Even that with the defaults it's 4 to 1 you tune each of them independently and all the routers do is to expect just 1 hello within the dead interval. The reason that the dead interval is bigger then the hello timer is to count the possibility that some packets might get drop on the way and you don't want to bring the whole OSPF adjacency down for just one dropped hello packet.

    If you need such a fast detection of a dead neighbor why you just don't go with BFD. It's more lightweight.
    Non-native English speaker
  • tinu_karkitinu_karki Member Posts: 21 ■□□□□□□□□□
    Thanks for the reply Sett it has clarified a few things, only question - with fast hellos i assumed dead intervals are not tunable, they are always set to 1 second?
  • thadizzythadizzy Member Posts: 72 ■■□□□□□□□□
    Check this link OSPF Support for Fast Hellos - Cisco Systems
    But yes, with my understanding dead timer will always be set to 1sec.
    Should be easy to put up a lab with? just enable timestamps in the console log and proper debugging.
Sign In or Register to comment.