changing the hello-time should affect failover time in your test environment. Did you test this? e.g. compare 1s/4s vs 10s/40s.
Nick On 28/05/2015 12:58, Mohammad Khalil wrote: > Ok , Sorry I misunderstood > So , I think combining both BFD and LFA will do the required failover time > But why changing the hello-time on both interfaces (the link between R1 and > R2) will not affect ? > > BR, > Mohammad > >> Date: Thu, 28 May 2015 12:53:35 +0100 >> From: [email protected] >> To: [email protected]; [email protected] >> Subject: Re: [c-nsp] OSPF per-prefix LFA >> >> This has nothing to do with OSPF virtual links. I'm talking about the >> virtual ethernet interfaces defined in your virtual lab. >> >> You can test out whether it's doing what I suggested by changing the >> hellotime value on the relevant interfaces. >> >> Nick >> >> On 28/05/2015 12:50, Mohammad Khalil wrote: >> > Hi Nick and thanks for the reply >> > All my routers are participating in area 0 , no virtual-link in place >> > I shutdown the interface connected to R2 (from CSR or R1 side) >> > >> > BR, >> > Mohammad >> > >> >> Date: Thu, 28 May 2015 12:36:50 +0100 >> >> From: [email protected] >> >> To: [email protected]; [email protected] >> >> Subject: Re: [c-nsp] OSPF per-prefix LFA >> >> >> >> On 28/05/2015 12:24, Mohammad Khalil wrote: >> >> > When I shutdown the interface with R2 (as I do not want to turn off the >> >> > remote interface on R2 as I cannot turn on BFD on GNS3) and I have >> >> > checked the route to 2.2.2.2 , it took about 5-6 seconds to install the >> >> > same route via the backup path (via R3) >> >> >> >> did you try this by shutting down both sides of the virtual link at the >> >> same time? The remote side will not detect carrier loss on a virtual link, >> >> so ospf failover will be detected by according to the ospf deadtime. >> >> >> >> If you handle this with bfd, the failover time should be much faster. Last >> >> time I measured this (me3600/100ms bfd), the failover time for l2vpn lsp >> >> rerouting was reliably less than 400ms. I.e. the failover for regular ip >> >> service should be a little less. >> >> >> >> Nick >> _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
