[c-nsp] route processor redundancy basics (RPR)

[murchison link]

Lukas,
thank you very much.

On Tue, Aug 2, 2016 at 5:44 PM, Lukas Tribus <luky-37 at hotmail.com> wrote:

> > RPR: standby RP is partially initialised. That is, only the
> startup-config
> > of the active router and standby router are sync'ed.
>
> Correct.
>
>
> > I lost about 1 or 2 minutes of pings during this stage, then RP1
> stabilised
> > and the pings started going through.
> >
> > Is this expected behaviour?
>
> Yes. You just said it yourself: RPR only sync's the startup configuration,
> meaning that a "switchover" (which actually isn't a switchover) will reload
> both RP's.
>
>
> > Also, why would anyone use RPR over SSO?
>
> RPR: RP redundancy (not a switchover)
> SSO: stateful switchover
>
>
> You can use RPR (which again, is not a switchover) if you have a problem
> with SSO. For example if you have reason to believe that an inconsistency
> (due to a bug) may spread to the secondary RP due to stateful
> synchronization, and you don't need an actual switchover (but only hw
> redundancy).
>
> We have some ancient GSR12k as P (therefor no customers are connected to
> those boxes) with dual RP configured in RPR, because we don't care about
> stateful switchover, as there are backup paths around the box. We do want
> the box to come back after a few minutes if there is a hardware failure
> though.
>
>
> On a box with single homed customers connected to it I would always run
> SSO.
>
>
>
> Regards,
>
> Lukas
>



-- 
regards,
Prakash Ganesh