Ray Jobs and Source Submission

About 167 wordsLess than 1 minute

Use direct Python execution when your driver already runs inside the cluster environment. Use Ray Jobs when you want a repeatable remote submission boundary.

Stable cluster image

If RayOrch and workload dependencies are already installed:

run = bench.submit("http://ray-head:8265")
report = run.wait(timeout_s=3600)

No wheel is required by the Benchmark API.

Development from a source checkout

from rayorch.benchmark import LocalSource

run = bench.submit(
    "http://ray-head:8265",
    source=LocalSource(
        project_root="/path/to/RayOrch",
        modules=("/path/to/another-local-module",),
    ),
)
report = run.wait(timeout_s=3600)

LocalSource maps the project to Ray working_dir, optional modules to py_modules, and by default keeps the workload's declared dependency installation. Set install_dependencies=False when dependencies are already managed by prepared stage environments.

Shared artifact path

BenchmarkRun.wait() reads summary.json after the Job succeeds. Therefore the artifact path must be visible to both the Job and the submitting process. wait(timeout_s=...) limits how long the client waits; it does not stop the remote Job automatically. Call run.stop() explicitly if desired.