[j-nsp] router reflector clients and non-clients

Mark Tinka mark.tinka at seacom.mu
Mon Jun 4 08:42:46 EDT 2018

On 4/Jun/18 12:43, Victor Sudakov wrote:

> Dear Christopher,
> What do you mean by "done correctly" in this case?

Having memory and CPU to throw at the problem, I presume :-).

We have plenty of RAM - and CPU - but we run a single Cluster-ID per
PoP, having mitigated the risks of doing so. Been working fine since
2014, when we went virtual for RR's.

I suppose the point is if you know what you're doing, either option will
work fine. Single Cluster-ID's have a higher risk, but if you account
for that, you will be as trouble-free as I have for the last 4 years.


More information about the juniper-nsp mailing list