[KFP] Obtain the Ray cluster run ID from the user for KFP v2. #956
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Revital Sur <[email protected]>
|
Please update the PR explanation: |
roytman
reviewed
Jan 22, 2025
kfp/doc/simple_transform_pipeline.md
Outdated
| # In KFPv2 dsl.RUN_ID_PLACEHOLDER is deprecated and cannot be used since SDK 2.5.0. On another hand we cannot create | ||
| # a unique string in a component (at runtime) and pass it to the `clean_up_task` of `ExitHandler`, due to | ||
| # https://github.com/kubeflow/pipelines/issues/10187. Therefore, meantime the user is requested to insert | ||
| # a unique string created at compilation time. |
roytman
reviewed
Jan 22, 2025
roytman
reviewed
Jan 23, 2025
kfp/doc/simple_transform_pipeline.md
Outdated
| func=compute_exec_params_func, base_image=base_kfp_image | ||
| ) | ||
| else: | ||
| compute_exec_params_op = comp.create_component_from_func(func=compute_exec_params_func, base_image=base_kfp_image) | ||
| # create Ray cluster | ||
| create_ray_op = comp.load_component_from_file("../../../kfp_ray_components/createRayComponent.yaml") |
There was a problem hiding this comment.
should it be component_spec_path + "createRayClusterComponent.yaml"
and below
kfp/doc/simple_transform_pipeline.md
Outdated
| @@ -57,11 +57,16 @@ Ray cluster. For each step we have to define a component that will execute them: | |||
| ```python | |||
| # components | |||
| base_kfp_image = "quay.io/dataprep1/data-prep-kit/kfp-data-processing:0.0.2" | |||
| @@ -107,6 +112,7 @@ The input parameters section defines all the parameters required for the pipelin | |||
| The parameters used here are as follows: | |||
|
|
|||
| * ray_name: name of the Ray cluster | |||
| * ray_run_id_KFPv2: Ray cluster unique ID used only in KFP v2 | |||
There was a problem hiding this comment.
it is not in the pipeline parameters; it should be around line 91
| @@ -111,9 +99,11 @@ def {{ pipeline_name }}( | |||
| ray_name: str = "{{ pipeline_name }}-kfp-ray", # name of Ray cluster | |||
| # Add image_pull_secret and image_pull_policy to ray workers if needed | |||
| {%- if image_pull_secret != "" %} | |||
| ray_run_id_KFPv2: str = "", | |||
There was a problem hiding this comment.
why do we need to define it twice? Here and at L 106?
| @@ -139,6 +128,7 @@ def code2parquet( | |||
| """ | |||
| Pipeline to execute NOOP transform | |||
| :param ray_name: name of the Ray cluster | |||
| :param ray_run_id_KFPv2: string holding the id used for the Ray cluster used only in KFP v2 | |||
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Revital Sur <[email protected]>
roytman
reviewed
Jan 27, 2025
| @@ -113,6 +101,7 @@ def compute_exec_params_func( | |||
| ) | |||
| def code2parquet( | |||
| ray_name: str = "code2parquet-kfp-ray", # name of Ray cluster | |||
| ray_run_id_KFPv2: str = "", # Ray cluster unique ID used only in KFP v2 | |||
There was a problem hiding this comment.
Why do we share it with both versions if we don't need it for KFPv1?
| @@ -21,6 +21,7 @@ Another useful feature of the KFP v2 is the `Json` editor for the `dict` type in | |||
| - It creates just one run that includes all the nested transfroms and their sub-tasks. | |||
| - No need for additional component as `executeSubWorkflowComponent.yaml`. All the implementation in the same pipeline file. | |||
| - In superpipelines of KFP v1 there exists an option to override the common parameters with specific values for each one of the transforms. This option is missing in the KFP v2 superpipelines. | |||
| - In kfp V2 pipelines the user is requested to insert a unique string for the ray cluster created at run creation time (called `ray_run_id_KFPv2`). This is because in KFPv2 dsl.RUN_ID_PLACEHOLDER is deprecated and cannot be used since SDK 2.5.0. | |||
There was a problem hiding this comment.
"and we cannot generate a unique string at run-time, see kubeflow/pipelines#10187"
roytman
reviewed
Jan 27, 2025
| @@ -62,6 +62,7 @@ def super_pipeline( | |||
| p2_skip: bool = False, | |||
| p2_noop_sleep_sec: int = 10, | |||
| p2_ray_name: str = "noop-kfp-ray", | |||
| p2_ray_run_id_KFPv2: str = "", | |||
There was a problem hiding this comment.
can it be p1_ parameter, so users don't have to add it for each step?
Signed-off-by: Revital Sur <[email protected]>
Signed-off-by: Maroun Touma <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Why are these changes needed?
In KFPv2 dsl.RUN_ID_PLACEHOLDER is deprecated and cannot be used since SDK 2.5.0. On another hand we cannot create
a unique string in a component (at runtime) and pass it to the
clean_up_taskofExitHandler, due tokubeflow/pipelines#10187.
Therefore, meantime the user is requested to insert a unique Ray cluster name at Run creation time.
Issue
#977