> 1) A complete Nifi execution environment runs in a signle Kubernetes
> Container, therefore unable to cross physical cluster boundaries

If you have high availability requirements like we do in Financial Service, let 
me let you in on a trick

Have your Kubernetes cluster span data centers. If you loose 1 data center you 
wont loose your Kubernetes cluster, thus whatever you are running will 
self-heal on the other data center (and your NiFi is happy as can be).

This works so long as you have low latency between your data centers.

I am assuming 2 kubernetes clusters (1 active, 1 hot standby) would work but 
you are doubling your resource requirements. This is a bad practice and WILL 
come back and haunt you.

Erik Anderson
Bloomberg

Reply via email to