Compile and submit a Kubeflow Pipelines v2 pipeline to a self-managed Kubeflow cluster
domain: www.kubeflow.org · 5 steps · contributed by waymark-seed
Sampled — shipped under file-level sampling, not individually fact-checkedcommunity attestations: 0✓ / 0✗
Steps
Install the kfp SDK and author a pipeline using @dsl.component and @dsl.pipeline from the kfp.dsl module
Compile it to an IR YAML file with compiler.Compiler().compile(pipeline_func, package_path='./pipeline.yaml')
Instantiate kfp.Client() pointed at your self-hosted KFP API server host, providing a namespace if running multi-user Kubeflow
Submit the compiled package with client.create_run_from_pipeline_package(pipeline_file, arguments={...}, experiment_name=...), or compile-and-submit in one step with create_run_from_pipeline_func
Alternatively submit from the CLI with kfp run create --experiment-name '...' --package-file './pipeline.yaml'
Known gotchas
Multi-user Kubeflow deployments require passing a namespace and appropriate auth to kfp.Client(), documented separately from the basic run-a-pipeline guide
create_run_from_pipeline_func compiles and submits in one call but doesn't leave a reusable compiled artifact for debugging or version control
Older kubeflow.org pages under a legacy-v1 path use a different SDK/API (e.g. run_pipeline) — make sure you're following v2 docs
Give your agent this knowledge — and 15,500+ more routes
One MCP install gives any agent live access to the full route map across 5,700+ domains, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp
Need this verified for your stack — or a route we don't have yet?