3 ms·
Was good to work with HLF but the ops background required to run it in production was staggering: Replicated Kafka + ZK backing the consensus algorithm. I put i
by harlanji 8y ago
Was good to work with HLF but the ops background required to run it in production was staggering: Replicated Kafka + ZK backing the consensus algorithm. I put it all in Kubernetes with StatefulSets. Even with my background running those systems it felt unreasonable. I quit that job in Nov so I don’t know if the situation has improved since then. I liked the architecture and the team was responsive and put care into docs. Would use again.
- brianbehlendorf 8y agoThank you! Folks are working hard to get deployment and management via Kube (and Helm Charts, etc) to be first-class supported and consistent. It's about where one would expect a 1.x level project to be, which means improvements like you point out are needed. Lots has happened since November. :) Hope you consider coming back and contributing.