Daft can run on Kubernetes using the Daft quickstart Helm chart. This allows you to test drive both simple single-node jobs and distributed Ray-based workloads on your Kubernetes cluster without any additional dependencies.
# Install the job with your script
helminstallmy-joboci://ghcr.io/eventual-inc/daft/quickstart\--set-filejob.script=my_script.py
# View logs
kubectllogs-fjob/my-job-quickstart-job
# Cleanup
helmuninstallmy-job
# Install with Ray cluster (3 workers)
helminstalldistributed-joboci://ghcr.io/eventual-inc/daft/quickstart\--setdistributed=true\--setworker.replicas=3\--set-filejob.script=distributed_script.py
# Access Ray Dashboard
kubectlport-forwardservice/distributed-job-quickstart-head8265:8265
# Open http://localhost:8265 in your browser# View job logs
kubectllogs-fjob/distributed-job-quickstart-job
# Cleanup
helmuninstalldistributed-job