Fork me on GitHub

@morri we use dc/os for orchestration - to update our onyx jobs, as long as the job metadata remains the same, we just update the dc/os service description and it deploys all the peers which pick up where the previous version left off (since the job metadata is all stored in zk). i imagine k8s would work very similarly


if the job metadata has changed then we have to submit the job again after updating the dc/os service, but this only happens rarely for us - e.g. when we deploy a version with sufficiently changed schema that we want to use different kafka topics