The orchestration system you're talking about is Nomad. We run a huge Nomad cluster (we're in the process of moving away from it, but it's one of the harder parts of our infrastructure to wrangle).