Datasets:

hjerpe

/

github-kubeflow-issues

like 0

https://api.github.com/repos/kubeflow/pipelines/issues/9299

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9299/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9299/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9299/events

https://github.com/kubeflow/pipelines/issues/9299

I_kwDOB-71UM5k-vVF

Vulnerabilities in images identified

open

NONE

### Feature Area /area backend ### What feature would you like to see? <!-- Provide a description of this feature and the user experience. --> Vulnerabilities in images: Currently there are vulnerabilities: go pkg github.com/emicklei/go-restful; CVE-2022-1996; CRITICAL; v2.9.5+incompatible; resolved in 2.16.0; authorization bypass k8s.io/kubernetes; CVE-2018-1002101; CRITICAL; v1.11.1; resolved in 1.11.2; command injection python pkg Pillow; CVE-2022-24303; CRITICAL; 8.4.0; resolved in 9.0.1; path traversal joblib; CVE-2022-21797; CRITICAL; 0.14.1; resolved in 1.2.0; remote code execution tensorflow; CVE-2023-25668; CRITICAL; 2.5.1; resolved in 2.11.1; remote code execution tensorflow-cpu; CVE-2023-25668; CRITICAL; 2.5.1; resolved in 2.11.1; remote code execution node pkg minimist; CVE-2021-44906; CRITICAL; 1.2.5; resolved in 1.2.6; prototype pollution underscore; CVE-2021-23358; CRITICAL; 1.7.1, resolved in 1.9.1; arbitrary code execution ### What is the use case or pain point? <!-- It helps us understand the benefit of this feature for your use case. --> ### Is there a workaround currently? <!-- Without this feature, how do you accomplish your task today? --> --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9299/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9298

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9298/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9298/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9298/events

https://github.com/kubeflow/pipelines/issues/9298

I_kwDOB-71UM5k9O3K

[bug] Kubeflow Pipelines issue K8s 1.22+

open

NONE

### Environment Hi, a previously reported issue https://github.com/kubeflow/pipelines/issues/8076 doesnt seem to be fixed with any of the latest KFP versions. * How do you deploy Kubeflow Pipelines (KFP)? Standalone deployment on EKS https://www.kubeflow.org/docs/components/pipelines/v1/installation/standalone-deployment/ * KFP version: 1.8.13 ### Steps to reproduce ``` > kubectl diff --kustomize "github.com/kubeflow/pipelines/manifests/kustomize/cluster-scoped-resources?ref=1.8.13" error: unable to recognize "github.com/kubeflow/pipelines/manifests/kustomize/cluster-scoped-resources?ref=1.8.13": no matches for kind "CustomResourceDefinition" in version "apiextensions.k8s.io/v1beta1" ``` ### Materials and reference The beta CustomResourceDefinition API (apiextensions.k8s.io/v1beta1) has been removed. See [Kubernetes 1.22 API changes](https://kubernetes.io/blog/2021/07/14/upcoming-changes-in-kubernetes-1-22/). ### Labels <!-- /area backend--> --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9298/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9292

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9292/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9292/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9292/events

https://github.com/kubeflow/pipelines/issues/9292

I_kwDOB-71UM5k6bJQ

[bug] Unexpected Argo behavior with background processes and emissary executor

open

NONE

### Environment We have a Kubeflow deployment with KFP and are running `gcr.io/ml-pipeline/argoexec:v3.3.8-license-compliance` with the emissary executor. ### Steps to reproduce Create and run a pipeline with this op: ``` def test_op(): return dsl.ContainerOp( name="test_container", image='public.ecr.aws/amazonlinux/amazonlinux:latest', command=["/bin/sh", "-c"], arguments=["sleep 600 & ls -a"] ) ``` ### Expected result I expect the container to behave as if I ran `docker run public.ecr.aws/amazonlinux/amazonlinux:latest /bin/sh -c 'sleep 600 & ls -a'` in which case the container command backgrounds the sleep process, runs `ls -a`, and the `/bin/sh` process and container exit immediately after. Note: I observed no difference when adding `--init`. With Argo, the container will run: ``` argoexec emissary -- /bin/sh -c "sleep 600 & ls -a" ``` However, while the `/bin/sh` process exits, the `argoexec` process waits on the sleep process. If I run this same image, command, and argument set in AWS Batch, I get the expected behavior. In my view, the `argoexec emissary` interception is breaking the abstraction I had for running my image with a specified command and arguments. Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9292/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9266

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9266/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9266/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9266/events

https://github.com/kubeflow/pipelines/issues/9266

I_kwDOB-71UM5krPKH

[frontend] tfx components like StatisticsGen and ExampleValidator does not display other splits than 'train' and 'eval' in the KFP vizualization window

open

NONE

### Environment * How did you deploy Kubeflow Pipelines (KFP)? Kubeflow Pipelines as part of a full Kubeflow deployment in Google Cloud Platform * KFP version: v1.5 ### Steps to reproduce When creating a third split named "val" (for example), visualization components windows (at least StatisticsGen and ExampleValidator) only display the 'train' and 'eval' split even if the 3 splits sub-folders are present in google storage buckets (`Split-train`, `Split-eval` and `Split-val`) . BTW, with `tfx.InteractiveContext()` in jupyter notebook, i can see the 3 visualizations splits ### Expected result Seeing other splits than `train` and `eval` in windows components vizualisations I Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9266/timeline

reopened

https://api.github.com/repos/kubeflow/pipelines/issues/9251

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9251/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9251/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9251/events

https://github.com/kubeflow/pipelines/issues/9251

I_kwDOB-71UM5kfow5

load_component_from_file() input to be parameterised

open

NONE

Issue: looking to parse a pipeline argument through to the input for load_component_from_file() ``` @dsl.pipeline(name="my-pipeline") def pipeline( model_version: str ) -> None: run_model_for_partition_task = comp.load_component_from_file( filename=f"models/run_model_{model_version}_component_spec.yaml" ``` However getting this error - `FileNotFoundError: [Errno 2] No such file or directory: 'models/run_model_{{pipelineparam:op=;name=model_version}}_component_spec.yaml'` Is there a way to do the above, as to avoid having to hardcode the location of the component spec file?

https://api.github.com/repos/kubeflow/pipelines/issues/9251/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9240

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9240/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9240/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9240/events

https://github.com/kubeflow/pipelines/issues/9240

I_kwDOB-71UM5kcIwM

[sdk] Pyright classifies use of types "not allowed"

closed

NONE

### Environment ``` [walter@cuddles Development]$ pip list | grep kfp kfp 1.8.20 kfp-pipeline-spec 0.1.16 kfp-server-api 1.8.5 ``` ### Descriptoin Pyright seems to think kubeflow pipeline's use of type constructors in the function signature is incorrect: ``` [walter@cuddles Development]$ cat kubeflow.py from kfp.components import InputPath def my_op(input_file_path: InputPath(str)): print(input_file_path) ``` ``` [walter@cuddles Development]$ pyright kubeflow.py No configuration file found. No pyproject.toml file found. stubPath /home/walter/Development/typings is not a valid directory. Assuming Python platform Linux Searching for source files Found 1 source file pyright 1.1.291 /home/walter/Development/kubeflow.py /home/walter/Development/kubeflow.py:3:28 - error: Illegal type annotation: call expression not allowed (reportGeneralTypeIssues) 1 error, 0 warnings, 0 informations Completed in 0.534sec ``` Is there a way around this? Is this the intended way to use the pipelines `InputPath` API? Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9240/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9239

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9239/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9239/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9239/events

https://github.com/kubeflow/pipelines/issues/9239

I_kwDOB-71UM5kax7m

[bug] Possible Dependency over specification/typo in GCPC-2.0.0b1

closed

NONE

Apologies if this is the wrong place to post this, my issue seems to be in the GCPC==2.0.0b1 python package hosted on Pypi (https://pypi.org/project/google-cloud-pipeline-components/#history) and its dependencies. Since GCPC seems to live in this repo I am posting here, but I am unsure if it is a Backend, Frontend or SDK bug. ### Short Summary: It seems the version of aiplatform used in GCPC==2.0.0b1 was pinned to less than 2 but also EXACTLY 1.14.0. The dependencies file for GCPC before this version and throughout this repo generally specify "aiplatform <2, => some_version" but what i think is the GCPC 2.0.0b1 release branch (https://github.com/kubeflow/pipelines/blob/google-cloud-pipeline-components-2.0.0b1/components/google-cloud/dependencies.py) has it specified as changing from: "google-cloud-aiplatform>=1.14.0,<2" to "google-cloud-aiplatform==1.14.0,<2", In the following commit: https://github.com/kubeflow/pipelines/commit/04d7fa3c5ec333e31052b1c31fabf85bc7af8dbc Was this intentional or is it a typo? If a typo how quickly can this be fixed as it is blocking me using GCPC components in Pipeline Templates on GCP Vertex AI (since Cloud Scheduled Pipeline Jobs template_path argument expects the https uri of the Pipeline Template in a KFP Artifact Repo and https pulling of KFP artifacts was only added in aiplatform after 1.14) ### Full Context: I am trying to write a KFP Pipeline containing Google Cloud Pipeline Components, store said pipeline in a KFP Artifact repository in Google Cloud Platform and then schedule this pipeline template using Cloud Scheduler to call a Cloud Function which will pull the template from a HTTPs URI and run the template as a pipeline job. I am essentially following the steps outlined here: https://cloud.google.com/vertex-ai/docs/pipelines/create-pipeline-template?hl=en#vertex-ai-sdk-for-python However trying to install both GCPC==2.0.0b1 and aiplatform==1.24.1 in a clean environment causes PIP to complain that aiplatform 1.24.1 doesn't match the dependencies specified for the GCPC package. ``` The conflict is caused by: The user requested google-cloud-aiplatform==1.24.1 google-cloud-pipeline-components 2.0.0b1 depends on google-cloud-aiplatform<2 and ==1.14.0 ``` Note that PIP is somehow requesting <2 and _exactly_ 1.14.0 which to me seems odd. If I force the install locally and try running my Cloud Function to pull the artifact using https uri as a template path it builds the job and submits it to GCP fine, so I suspect this is a accidental over pinning of the dependency. The non-zero exit status of this PIP build means that I cannot deploy the Cloud Function needed to schedule a KFP which makes use of GCPC components, as GCP refuses to deploy a requirements file that causes PIP to technically fail (even though the build seems fine when testing it). This problem can be reproduced by installing a conda env (attached below), or by installing GCPC==2.0.0b1 and aiplatform==1.24.1 into a clean environment using pip directly. ### Expected result I would expect pip to allow me to install aiplatform==1.24.1 alongside GCPC 2.0.0b1 as both are required to use GCPC in Pipeline templates and I suspect the pinning of aiplatform to less than 2 but _exactly_ 1.14.0 is a typo. ### Materials and reference: Conda File to reproduce issue: ``` channels: - conda-forge dependencies: - python==3.10 - pip - pip: - google-cloud-aiplatform==1.24.1 - google-cloud-pipeline-components==2.0.0b1 ``` ### Labels <!-- Please include labels below by uncommenting them to help us better triage issues --> /area components --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9239/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9224

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9224/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9224/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9224/events

https://github.com/kubeflow/pipelines/issues/9224

I_kwDOB-71UM5kQSqL

[feature] Time fields in kfp RecurringRun should inform users of what timezone the field is in.

open

NONE

### Feature Area /area frontend (Documentation) ### What feature would you like to see? The documentation and possibly display for the timezones of the recurring run time components as they appear in the kubeflow pipelines dashboard should be better. Specifically, what timezone each of the following fields is assumed to be in: - Trigger cron expression (seems to be UTC) - start_time (appears to display a local time) - end_time (appears to display a local time) ### What is the use case or pain point? It is not clear to a user of the kfp frontend how various datetimes should be interpreted or supplied. ### Is there a workaround currently? No. --- Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9224/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9220

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9220/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9220/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9220/events

https://github.com/kubeflow/pipelines/issues/9220

I_kwDOB-71UM5kPj6q

[04/24/23][infra] e2e test build failure

closed

COLLABORATOR

E2e test failed to build images with error: > Reading package lists... > W: The repository 'http://security.debian.org/debian-security stretch/updates Release' does not have a Release file. > W: The repository 'http://deb.debian.org/debian stretch Release' does not have a Release file. > W: The repository 'http://deb.debian.org/debian stretch-updates Release' does not have a Release file. > E: Failed to fetch http://security.debian.org/debian-security/dists/stretch/updates/main/binary-amd64/Packages 404 Not Found > E: Failed to fetch http://deb.debian.org/debian/dists/stretch/main/binary-amd64/Packages 404 Not Found > E: Failed to fetch http://deb.debian.org/debian/dists/stretch-updates/main/binary-amd64/Packages 404 Not Found > E: Some index files failed to download. They have been ignored, or old ones used instead. > The command '/bin/sh -c apt-get update && apt-get install -y cmake clang musl-dev openssl' returned a non-zero code: 100 > ERROR > ERROR: build step 0 "gcr.io/cloud-builders/docker" failed: step exited with non-zero status: 100

https://api.github.com/repos/kubeflow/pipelines/issues/9220/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9201

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9201/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9201/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9201/events

https://github.com/kubeflow/pipelines/issues/9201

I_kwDOB-71UM5kEMM7

[Help Wanted] Use large files from notebook server in KubeFlow Pipeline

closed

NONE

I have a kubeflow notebook server that houses a large file (100 GB+) and was trying to read it into my pipeline but I am getting a file not found error. Not sure what's the best step since most tutorials I see download their data which isn't feasible in my case. 

https://api.github.com/repos/kubeflow/pipelines/issues/9201/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9200

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9200/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9200/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9200/events

https://github.com/kubeflow/pipelines/issues/9200

I_kwDOB-71UM5kD2lp

[sdk] TabularDatasetCreateOp raises TypeError: create() got an unexpected keyword argument error

open

NONE

### Environment * KFP version: 2.0.0b14 * KFP SDK version: google-cloud-pipeline-components==2.0.0b1 * All dependencies version: kfp 2.0.0b14 kfp-pipeline-spec 0.2.2 kfp-server-api 2.0.0b1 ### Steps to reproduce Running this code (project id, location, dataset name, table name fields are redacted): ``` from google_cloud_pipeline_components.v1.dataset import TabularDatasetCreateOp TabularDatasetCreateOp( display_name="dataset_name", bq_source="bq://**********.dataset_name.table_name", project="********", location="********", ) ``` ### Expected result I'm expecting table to be created but I'm getting errors with the following logs (see screenshot) <img width="802" alt="tablename" src="https://user-images.githubusercontent.com/12820124/233678623-c1dd062d-cb86-4a2a-b3fc-86ff4fe6f40b.png"> <img width="627" alt="tablenamerogue" src="https://user-images.githubusercontent.com/12820124/233678599-9fc7e717-a25e-4c01-bc19-2cf80e53cc27.png"> <img width="1442" alt="ss1" src="https://user-images.githubusercontent.com/12820124/233678607-dffcfc18-08c5-4b6c-927b-a93c8af72c1f.png"> I noticed that: 1) --method.bqsource is not parsing the values properly: instead of the value being in the next line of the log, it's being parsed as `--method.bqsource********.dataset_name.table_name` 2) The log message is saying `TabularDataset` is being called with parameters: ``` {'project': '********', 'location': '********', 'display_name': '********', 'TABLE_NAME': '--method.labels', 'labels': {'vertex-ai-pipelines-run-billing-id': '********'}} ``` There's a rogue field there with the value of my table name 3) The job then quits with a `TypeError` with message `create() got an unexpected keyword argument TABLE_NAME`. --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9200/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9199

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9199/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9199/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9199/events

https://github.com/kubeflow/pipelines/issues/9199

I_kwDOB-71UM5kCQ7j

[sdk] Cannot conncect with MLflow Tracking Server from KFP

closed

NONE

### Environment * KFP version: ``` kfp==1.8.19 kfp-pipeline-spec==0.1.16 kfp-server-api==1.8.5 ``` * All dependencies version: ``` mlflow==2.1.1 ``` ### Steps to reproduce I have Kubeflow 1.6 installed in Openshift. I ran Python code in my Jupyter Notebook Server: ``` import boto3 import os import mlflow os.environ['AWS_ACCESS_KEY_ID'] = 'XXX' os.environ['AWS_SECRET_ACCESS_KEY'] = 'YYY' os.environ['MLFLOW_S3_ENDPOINT_URL'] = 'http://ZZZ:9000' remote_server_uri='http://mlflow:5000' mlflow.set_tracking_uri(remote_server_uri) client = mlflow.MlflowClient() data = client.get_run('cb75f28b1c25479e8101d5ce52f8c482').data print(data) ``` and I got ok response: ``` <RunData: metrics={'accuracy': 1.0, 'precision': 1.0, 'recall': 1.0}, params={'C': '10', 'algorithm': 'logistic_regression'}, tags={'mlflow.log-model.history': '[{"run_id": "cb75f28b1c25479e8101d5ce52f8c482", ' '"artifact_path": "model_logistic_regression", ' '"utc_time_created": "2023-02-22 ' '11:44:51.822321", "flavors": {"python_function": ' '{"model_path": "model.pkl", "predict_fn": ' '"predict", "loader_module": "mlflow.sklearn", ' '"python_version": "3.8.10", "env": {"conda": ' '"conda.yaml", "virtualenv": "python_env.yaml"}}, ' '"sklearn": {"pickled_model": "model.pkl", ' '"sklearn_version": "0.24.2", ' '"serialization_format": "cloudpickle", "code": ' 'null}}, "model_uuid": ' '"ba88d0550c2640848d5fb5ba158028df", ' '"mlflow_version": "2.1.1"}]', 'mlflow.source.name': '/opt/conda/lib/python3.8/site-packages/ipykernel_launcher.py', 'mlflow.source.type': 'LOCAL', 'mlflow.user': 'jovyan'}> ``` However, I created modified code for KFP pipeline: ``` import kfp from kfp import dsl import kfp.components as components from typing import NamedTuple def get_iris_data() -> NamedTuple('Outputs', [('acc', float)]): import boto3 import os import mlflow import logging logging.basicConfig( level = logging.INFO, format = '%(asctime)s %(levelname)s: %(message)s') os.environ['AWS_ACCESS_KEY_ID'] = 'XXX' os.environ['AWS_SECRET_ACCESS_KEY'] = 'YYY' os.environ['MLFLOW_S3_ENDPOINT_URL'] = 'http://ZZZ:9000' remote_server_uri='http://mlflow:5000' mlflow.set_tracking_uri(remote_server_uri) client = mlflow.MlflowClient() data = client.get_run('cb75f28b1c25479e8101d5ce52f8c482').data logging.info(data) acc = data.metrics['accuracy'] from collections import namedtuple divmod_output = namedtuple('Outputs', ['acc']) return [float(acc)] comp_get_iris_data = components.create_component_from_func(get_iris_data, base_image="python:3.9.16-slim-buster", packages_to_install=['mlflow==2.1.1', 'boto3']) @kfp.dsl.pipeline( name = 'mlflow_test', description = 'test' ) def mlflow_test(): step1 = comp_get_iris_data() if __name__ == "__main__": kfp.compiler.Compiler().compile(pipeline_func=mlflow_test,package_path='mlflow_test.yaml') ``` and I got error in pod: ``` time="2023-04-21T07:37:11.897Z" level=info msg="capturing logs" argo=true WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv Traceback (most recent call last): File "/tmp/tmp.upL4JRaEG8", line 43, in <module> _outputs = get_iris_data(**_parsed_args) File "/tmp/tmp.upL4JRaEG8", line 21, in get_iris_data data = client.get_run('cb75f28b1c25479e8101d5ce52f8c482').data File "/usr/local/lib/python3.9/site-packages/mlflow/tracking/client.py", line 150, in get_run return self._tracking_client.get_run(run_id) File "/usr/local/lib/python3.9/site-packages/mlflow/tracking/_tracking_service/client.py", line 72, in get_run return self.store.get_run(run_id) File "/usr/local/lib/python3.9/site-packages/mlflow/store/tracking/rest_store.py", line 134, in get_run response_proto = self._call_endpoint(GetRun, req_body) File "/usr/local/lib/python3.9/site-packages/mlflow/store/tracking/rest_store.py", line 56, in _call_endpoint return call_endpoint(self.get_host_creds(), endpoint, method, json_body, response_proto) File "/usr/local/lib/python3.9/site-packages/mlflow/utils/rest_utils.py", line 281, in call_endpoint response = verify_rest_response(response, endpoint) File "/usr/local/lib/python3.9/site-packages/mlflow/utils/rest_utils.py", line 213, in verify_rest_response raise MlflowException( mlflow.exceptions.MlflowException: API request to endpoint /api/2.0/mlflow/runs/get failed with error code 403 != 200. Response body: 'RBAC: access denied' time="2023-04-21T07:41:34.482Z" level=error msg="cannot save artifact /tmp/outputs/acc/data" argo=true error="stat /tmp/outputs/acc/data: no such file or directory" Error: exit status 1 ``` Why did I get 403 status running similar code? How can I fix it? ### Expected result Successfull connection with MLflow Tracking Server and recieve data from MLflow experiment

https://api.github.com/repos/kubeflow/pipelines/issues/9199/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9196

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9196/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9196/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9196/events

https://github.com/kubeflow/pipelines/issues/9196

I_kwDOB-71UM5kBjs8

[sdk] Can't submit recurring run with pipeline version reference

closed

NONE

### Environment * KFP version: 1.6 (running 2.0.0-beta.1 pipeline API & UI) * KFP SDK version: `kfp==2.0.0b14` ### Steps to reproduce Submit recurring run using latest SDK ``` client.create_recurring_run( experiment_id = experiment.experiment_id, job_name=job_name, cron_expression=cron_expression, pipeline_id=pipeline_id, version_id=pipeline_version_id, params=json.loads(params) ) ``` Got error ``` File "/opt/lib/python3.9/site-packages/kfp/client/client.py", line 979, in _create_job_config pipeline_spec=pipeline_doc.to_dict(), AttributeError: 'NoneType' object has no attribute 'to_dict' ``` https://github.com/kubeflow/pipelines/blob/2.0.0b14/sdk/python/kfp/client/client.py#L979 mitigated by replacing that line with `pipeline_spec=pipeline_doc.to_dict() if pipeline_doc else {},` ### Expected result Submit recurring run to Kubeflow ### Materials and reference https://github.com/kubeflow/pipelines/blob/2.0.0b14/sdk/python/kfp/client/client.py#L979 ### Labels /area sdk --- Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9196/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9195

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9195/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9195/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9195/events

https://github.com/kubeflow/pipelines/issues/9195

I_kwDOB-71UM5kBjeZ

[sdk] `get_pipeline_version` returns empty `pipeline_spec`

closed

COLLABORATOR

### Environment * KFP version: 2.0.0-beta.1 * KFP SDK version: 2.0.0-beta.14 * All dependencies version: kfp-server-api==2.0.0b1 ... ### Steps to reproduce ```python import kfp client = kfp.Client( host=<host_url>, ) print( client.get_pipeline_version('e9b08c24-7ea5-46f9-b06e-b4a9c47c4202', '57e4044d-38ca-4be3-b25b-cf61f7b5d3c2')) ``` ### Expected result Returned response should show non-empty `pipeline_spec` ### Actual behavior `pipeline_spec` is empty: ``` {'code_source_url': None, 'created_at': datetime.datetime(2023, 4, 20, 17, 4, 51, tzinfo=tzlocal()), 'description': None, 'display_name': 'add-numbers_version_at_2023-04-20T17:04:49.322Z', 'error': None, 'package_url': {'pipeline_url': 'https://raw.githubusercontent.com/kubeflow/pipelines/master/sdk/python/test_data/components/add_numbers.yaml'}, 'pipeline_id': 'e9b08c24-7ea5-46f9-b06e-b4a9c47c4202', 'pipeline_spec': {'fields': None}, 'pipeline_version_id': '57e4044d-38ca-4be3-b25b-cf61f7b5d3c2'} ```

https://api.github.com/repos/kubeflow/pipelines/issues/9195/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9172

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9172/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9172/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9172/events

https://github.com/kubeflow/pipelines/issues/9172

I_kwDOB-71UM5j1BGm

kubernetes_resource_delete_op - Remove hardcoded dependency on gcr.io

open

NONE

### Feature Area <!-- Uncomment the labels below which are relevant to this feature: --> <!-- /area frontend --> <!-- /area backend --> /area sdk <!-- /area samples --> <!-- /area components --> ### What feature would you like to see? Allow SDK users to override the default image used in `VolumeOp.delete()`. ### What is the use case or pain point? Hardcoded dependency on gcr.io makes it impossible to use this feature in an enterprise setting where access to public image repos are usually locked down. Related issue: https://github.com/kubeflow/pipelines/issues/9032 <!-- It helps us understand the benefit of this feature for your use case. --> ### Is there a workaround currently? No. A cleanup job outside KFP is probably the best we could do. <!-- Without this feature, how do you accomplish your task today? --> --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9172/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9171

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9171/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9171/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9171/events

https://github.com/kubeflow/pipelines/issues/9171

I_kwDOB-71UM5j0Wfg

[sdk] Unable to use custom params when using kfp.Client().run_pipeline()

open

NONE

### Environment * KFP version: <!-- For more information, see an overview of KFP installation options: https://www.kubeflow.org/docs/pipelines/installation/overview/. --> [2.0.0-alpha.5](https://github.com/kubeflow/pipelines/tree/2.0.0-alpha.5/manifests/kustomize) * KFP SDK version: <!-- Specify the version of Kubeflow Pipelines that you are using. The version number appears in the left side navigation of user interface. To find the version number, See version number shows on bottom of KFP UI left sidenav. --> kfp==2.0.0b13 * All dependencies version: <!-- Specify the output of the following shell command: $pip list | grep kfp --> ``` kfp 2.0.0b13 kfp-pipeline-spec 0.2.0 kfp-server-api 2.0.0a6 ``` ### Steps to reproduce <!-- Specify how to reproduce the problem. This may include information such as: a description of the process, code snippets, log output, or screenshots. --> ``` client = kfp.Client() run = client.run_pipeline( experiment_id=experiment_id, job_name=run_name, params=params, pipeline_id=pipeline_id, version_id=version_id, ) ``` pipeline runs with default params and not custom ones specified below: tried 2 formats for params: * {"x": "206", "seed": "3007"} * {"parameters": {"x": "206", "seed": "3007"}} ### Expected result <!-- What should the correct behavior be? --> Pipeline should run with custom params ### Materials and Reference <!-- Help us debug this issue by providing resources such as: sample code, background context, or links to references. --> Followed docs: https://kubeflow-pipelines.readthedocs.io/en/2.0.0b13/source/client.html#kfp.client.Client.run_pipeline --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9171/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9162

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9162/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9162/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9162/events

https://github.com/kubeflow/pipelines/issues/9162

I_kwDOB-71UM5jpKw6

[chore] Migrate tests that use deprecated namespace in KFP SDK

open

MEMBER

### What feature would you like to see? - [ ] We need to migrate v2 tests that use `kfp.deprecated` namespace to use v2 KFP SDK - [ ] Enable separate testing for v1 API using v1 KFP SDK ### Is there a workaround currently? We currently enforce installing compatible `kfp` and `kfp-server-api` packages in order to use the test cases that use `deprecated` namespace of v2 `kfp` package. --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9162/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9160

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9160/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9160/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9160/events

https://github.com/kubeflow/pipelines/issues/9160

I_kwDOB-71UM5jnm3A

[sdk] Volumes with kfp-kubernetes - no such file or directory

closed

CONTRIBUTOR

### Environment * KFP version: 2.0.0.beta.1 * KFP SDK version: 2.0.0b14 * All dependencies version: kfp 2.0.0b14 kfp-kubernetes 0.0.1 kfp-pipeline-spec 0.2.2 kfp-server-api 2.0.0b1 ### Steps to reproduce Following the [readme](https://github.com/kubeflow/pipelines/blob/master/kubernetes_platform/python/README.md) of the new functionality, I can create and mount the volume in 2 tasks (one depends on the other), however the second one cannot read (no such file or directory) what the first one wrote. I noticed in yaml file of the task pod the following: ```yaml volumes: - emptyDir: {} name: 59943aa7-e26a-4714-9d8e-51f3aafc20d8-my-pvc ``` Reading [this](https://kubernetes.io/docs/concepts/storage/volumes/#emptydir), the `emptyDir` just seems wrong and it doesn't look like I can control that (in sdk). Is this a bug or am I doing something wrong? ### Expected result One task can see what previous task wrote to pv. ### Materials and Reference <!-- Help us debug this issue by providing resources such as: sample code, background context, or links to references. --> --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9160/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9159

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9159/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9159/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9159/events

https://github.com/kubeflow/pipelines/issues/9159

I_kwDOB-71UM5jlDfE

[help wanted] use PipelineParam list for packages_to_install

open

NONE

Hello, I'm using python sdk `kfp==1.8.20`. I want to pass in list of packages to install for container as pipeline parameter. Here is the core version of my code. ``` @dsl.pipeline(name="pipeline") def pipeline(mysql_packages_to_install: list = ["mysql-connector-python==8.0.31"]): op_task = create_component_from_func(task_func, base_image="python:3.9", packages_to_install=mysql_packages_to_install) ``` I get the following error ``` Traceback (most recent call last): File "/Users/user/Workspace/kg-embedding-pipeline/main.py", line 75, in <module> run_pipeline(namespace, pipeline_input.input_class, pipeline_input.input_property) File "/Users/user/Workspace/kg-embedding-pipeline/main.py", line 44, in run_pipeline kfp.compiler.Compiler().compile(pipeline.embedding_pipeline, pipeline_tar_path) File "/Users/user/Workspace/kg-embedding-pipeline/venv/lib/python3.9/site-packages/kfp/compiler/compiler.py", line 1175, in compile self._create_and_write_workflow( File "/Users/user/Workspace/kg-embedding-pipeline/venv/lib/python3.9/site-packages/kfp/compiler/compiler.py", line 1227, in _create_and_write_workflow workflow = self._create_workflow(pipeline_func, pipeline_name, File "/Users/user/Workspace/kg-embedding-pipeline/venv/lib/python3.9/site-packages/kfp/compiler/compiler.py", line 1005, in _create_workflow pipeline_func(*args_list, **kwargs_dict) File "/Users/user/Workspace/kg-embedding-pipeline/core/pipeline.py", line 27, in embedding_pipeline op_mysql_generate_hrt = create_component_from_func(mysql.mysql_generate_hrt, base_image="python:3.9", packages_to_install=mysql_packages_to_install) File "/Users/user/Workspace/kg-embedding-pipeline/venv/lib/python3.9/site-packages/kfp/components/_python_op.py", line 1098, in create_component_from_func component_spec = _func_to_component_spec( File "/Users/user/Workspace/kg-embedding-pipeline/venv/lib/python3.9/site-packages/kfp/components/_python_op.py", line 768, in _func_to_component_spec ' '.join([repr(str(package)) for package in packages_to_install])) TypeError: 'PipelineParam' object is not iterable ``` I specify that PipelineParam "mysql_packages_to_install" is type list and even give it a default value as an iterable but it doesn't seem to notice the type. Is there a workaround? --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9159/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9158

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9158/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9158/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9158/events

https://github.com/kubeflow/pipelines/issues/9158

I_kwDOB-71UM5jetPR

[backend] <Sample Pipeline not consuming input IAM role parameter>

closed

CONTRIBUTOR

### Environment * How did you deploy Kubeflow Pipelines (KFP)? EKS 1.25 Kustomize * KFP version: v1.7.0 * KFP SDK version: v1 ### Steps to reproduce configure example pipeline per https://github.com/kubeflow/pipelines/tree/master/samples/contrib/aws-samples/simple_train_pipeline Replace these parameters with aws resources created: https://github.com/kubeflow/pipelines/blob/master/samples/contrib/aws-samples/simple_train_pipeline/training-pipeline.py#L28 https://github.com/kubeflow/pipelines/blob/master/samples/contrib/aws-samples/simple_train_pipeline/training-pipeline.py#L41 https://github.com/kubeflow/pipelines/blob/master/samples/contrib/aws-samples/simple_train_pipeline/training-pipeline.py#L57 https://github.com/kubeflow/pipelines/blob/master/samples/contrib/aws-samples/simple_train_pipeline/training-pipeline.py#L64 ### Expected result Pipeline can be run and complete. ### Error See this in the pipeline UI log ``` raise error_class(parsed_response, operation_name) botocore.exceptions.ClientError: An error occurred (AccessDenied) when calling the AssumeRoleWithWebIdentity operation: Not authorized to perform sts:AssumeRoleWithWebIdentity time="2023-04-14T21:40:33.669Z" level=error msg="cannot save artifact /tmp/outputs/job_name/data" argo=true error="stat /tmp/outputs/job_name/data: no such file or directory" time="2023-04-14T21:40:33.669Z" level=error msg="cannot save artifact /tmp/outputs/model_artifact_url/data" argo=true error="stat /tmp/outputs/model_artifact_url/data: no such file or directory" time="2023-04-14T21:40:33.669Z" level=error msg="cannot save artifact /tmp/outputs/training_image/data" argo=true error="stat /tmp/outputs/training_image/data: no such file or directory" Error: exit status 1 ``` Basically the pipeline didn't seem to use the role I specified which has the required S3 and Sagemaker Full Access Permission

https://api.github.com/repos/kubeflow/pipelines/issues/9158/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9155

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9155/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9155/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9155/events

https://github.com/kubeflow/pipelines/issues/9155

I_kwDOB-71UM5jawIi

Tensorboard pods unable to launch

open

NONE

Tensorboard is unable to launch in user profile it throwing below error .do we need to give any permission to default service Account or any configmap changes we need to make . Presently we PSP in cluster default-editor has the permissions ownerReferences: - apiVersion: tensorboard.kubeflow.org/v1alpha1 blockOwnerDeletion: true controller: true kind: Tensorboard name: t1 uid: 8f415f73-30f6-4674-9dc9-3356be6f75fa resourceVersion: "216338025" uid: 03da80a3-8903-4368-a440-e6d87e749fce spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: app: t1 strategy: rollingUpdate: maxSurge: 25% maxUnavailable: 25% type: RollingUpdate template: metadata: creationTimestamp: null labels: app: t1 spec: affinity: {} containers: - args: - --logdir=/tensorboard_logs/ - --bind_all command: - /usr/local/bin/tensorboard image: tensorflow/tensorflow:2.1.0 imagePullPolicy: IfNotPresent name: tensorboard ports: - containerPort: 6006 protocol: TCP resources: {} terminationMessagePath: /dev/termination-log terminationMessagePolicy: File volumeMounts: - mountPath: /tensorboard_logs/ name: tbpd readOnly: true workingDir: / dnsPolicy: ClusterFirst restartPolicy: Always schedulerName: default-scheduler securityContext: {} terminationGracePeriodSeconds: 30 volumes: - name: tbpd persistentVolumeClaim: claimName: test-volume status: conditions: - lastTransitionTime: "2023-01-31T14:48:53Z" lastUpdateTime: "2023-01-31T14:48:53Z" message: Deployment does not have minimum availability. reason: MinimumReplicasUnavailable status: "False" type: Available - lastTransitionTime: "2023-01-31T14:48:54Z" lastUpdateTime: "2023-01-31T14:48:54Z" ### **m**essage: 'pods "t1-79d5c94587-" is forbidden: PodSecurityPolicy: unable to admit pod: [spec.initContainers[0].securityContext.runAsNonRoot: Invalid value: false: must be true spec.initContainers[0].securityContext: Invalid value: []int64{0}: group 0 must be in the ranges: [{1 65535}] spec.initContainers[0].securityContext.capabilities.add: Invalid value: "NET_ADMIN": capability may not be added spec.initContainers[0].securityContext.capabilities.add: Invalid value: "NET_RAW": capability may not be added]'**** reason: FailedCreate status: "True" type: ReplicaFailure - lastTransitionTime: "2023-01-31T14:58:54Z" lastUpdateTime: "2023-01-31T14:58:54Z" message: ReplicaSet "t1-79d5c94587" has timed out progressing. reason: ProgressDeadlineExceeded status: "False" type: Progressing observedGeneration: 1 unavailableReplicas: 1 ](url)

https://api.github.com/repos/kubeflow/pipelines/issues/9155/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9154

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9154/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9154/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9154/events

https://github.com/kubeflow/pipelines/issues/9154

I_kwDOB-71UM5jYFWG

[backend] Validate pipeline arguments when submitting run from client

open

MEMBER

Currently, it is possible to submit a pipeline run without the required pipeline arguments and there is no backend validation. Reported by: @connor-mccarthy /cc @chensun @Linchin --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9154/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9147

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9147/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9147/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9147/events

https://github.com/kubeflow/pipelines/issues/9147

I_kwDOB-71UM5jVkHT

[bug] NCCL Error configurations in tasks, anyway to expose shm options in a task?

open

MEMBER

### Environment * How do you deploy Kubeflow Pipelines (KFP)? Vertex AI Pipelines * KFP SDK version: $ kfp --version kfp 2.0.0-beta.13 ### Steps to reproduce Hi kubeflow pipelines team, I am using Vertex AI Pipelines to train detection models, and a since today my pipelines are failing due to a NCCL error on GPU machines that have more than one GPU, I am creating the task as follows: ```python @dsl.container_component def train_model( container_trainer_uri: str, train_dataset: Input[Dataset], test_dataset: Input[Dataset], categories_json: Input[Artifact], pretrained_model: Input[Artifact], num_iters: int, batch_size: int, lr: str, num_epochs: int, lr_step: str, gpus: str, num_workers: int, val_intervals: int, model_output: Output[Artifact], model_metrics: Output[Markdown], ): import os python_cmd = ( "python3 main.py ctdet " " --dataset vertex_db " f"--train_path {str(train_dataset.path)} " f"--val_path {str(test_dataset.path)} " f"--categories_json {categories_json.path} " f"--load_model {pretrained_model.path} " "--exp_id vertex_training " "--input_res 512 " f"--num_iters {num_iters} " f"--batch_size {batch_size} " f"--lr {lr} " f"--num_epochs {num_epochs} " f"--lr_step {lr_step} " f"--gpus {gpus} " f"--num_workers {num_workers} " f"--val_intervals {val_intervals} " f"--markdown_path {model_metrics.path} " f"--output_path {model_output.path} " ) return dsl.ContainerSpec( image=str(container_trainer_uri), command=["sh", "-c", python_cmd], ) train_task = train_model( container_trainer_uri=container_uri.output, train_dataset=train_importer_task.output, test_dataset=test_importer_task.output, categories_json=categories_importer_task.output, pretrained_model=pretrained_model.output, num_iters=num_iters, batch_size=batch_size, lr=lr, num_epochs=num_epochs, lr_step=lr_step, gpus=gpus, num_workers=num_workers, val_intervals=val_intervals, ) train_task.set_display_name("Train model").set_cpu_limit("12").set_memory_limit( "170G" ).add_node_selector_constraint("NVIDIA_TESLA_A100").set_gpu_limit( "2" ).set_env_variable( name="NCCL_SHM_DISABLE", value="1" # Set or unset does not work ) ``` Then since April 12th, all my pipelines start to fail with: ``` NCCL Error 2: unhandled system error ``` This I solved usually by augmenting the shared memory or allowing `ipc=host` on the host, the thing is that with kubeflow pipeline v2. `ContainerSpec` does not have control over this, neither that like mounting a volume to shm. The whole thing also is that probably something in the infrastructure on Google changed, and it broke my training steps. So any help in this? ### Expected result Container runs without issues. ### Labels /area components --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9147/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9137

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9137/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9137/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9137/events

https://github.com/kubeflow/pipelines/issues/9137

I_kwDOB-71UM5jH6Si

[CVE] Resolve critical CVEs in KFP v1.8

closed

MEMBER

List of CVEs: - [x] #8153 - [x] #8154 - [x] #8155 - [ ] --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9137/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9136

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9136/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9136/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9136/events

https://github.com/kubeflow/pipelines/issues/9136

I_kwDOB-71UM5jH34G

[CVE] Resolve critical CVEs in KFP v2

closed

MEMBER

The list of discovered CVEs: - [x] #8786 - [ ] --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9136/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9127

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9127/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9127/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9127/events

https://github.com/kubeflow/pipelines/issues/9127

I_kwDOB-71UM5i8e8T

Report a potential vulnerability via GitHub feature

closed

NONE

I may have identified a vulnerability in the latest codebase of kubeflow/pipelines. I would like to submit the security report privately using the [GitHub Private Vulnerability Reporting](https://docs.github.com/en/code-security/security-advisories/repository-security-advisories/configuring-private-vulnerability-reporting-for-a-repository) feature. Which permits the possibility to accept/refuse the security report privately and request a CVE accordingly. Could you please consider enabling the feature for the kubeflow/pipelines repository?

https://api.github.com/repos/kubeflow/pipelines/issues/9127/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9126

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9126/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9126/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9126/events

https://github.com/kubeflow/pipelines/issues/9126

I_kwDOB-71UM5i8XbJ

[frontend] Newly created v2beta1 recurring run has status DISABLED

closed

MEMBER

Frontend uses `Status` instead of `Mode` when [creating a new recurring run](https://github.com/kubeflow/pipelines/blob/063494343232814058aa354f14d48cfaa3e6c736/frontend/src/pages/NewRunV2.tsx#L332). We need to use `Mode` [according to the API](https://github.com/kubeflow/pipelines/blob/master/backend/api/v2beta1/recurring_run.proto#L110-L133). Relevant BE issue: #9125 ### Environment * How did you deploy Kubeflow Pipelines (KFP)? Standalone * KFP version: master branch ### Steps to reproduce 1. Create a `v2beta1` recurring run and set Mode = ENABLED via UI 3. See the status of the recurring run in Recurring Runs page 4. Status = DISABLED ### Expected result Status = ENABLED --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9126/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9125

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9125/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9125/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9125/events

https://github.com/kubeflow/pipelines/issues/9125

I_kwDOB-71UM5i8UgA

[backend] Newly created v2beta1 recurring run has status DISABLED

closed

MEMBER

Backend does not set the `Status` properly when creating a new recurring run. Relevant FE issue: #9126 ### Environment * How did you deploy Kubeflow Pipelines (KFP)? Standalone * KFP version: master branch ### Steps to reproduce 1. Create a `v2beta1` recurring run and set Mode = ENABLED via UI 3. See the status of the recurring run in Recurring Runs page 4. Status = DISABLED ### Expected result Status = ENABLED --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9125/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9119

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9119/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9119/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9119/events

https://github.com/kubeflow/pipelines/issues/9119

I_kwDOB-71UM5i4llQ

[backend] failure to run pipeline OCI runtime create failed: runc create failed: unable to start container process: exec: "/var/run/argo/argoexec": stat /var/run/argo/argoexec: no such file or

open

NONE

All pipeline runs fail with the following error message: ``` Error: failed to start container "main": Error response from daemon: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: exec: "/var/run/argo/argoexec": stat /var/run/argo/argoexec: no such file or directory: unknown ``` ### Environment * KFP Standalone * KFP version: 2.0.0b13 * KFP SDK version: kfp 2.0.0b13 kfp-pipeline-spec 0.2.0 kfp-server-api 2.0.0a6 ### Steps to reproduce 1. compile pipeline from hello_world sample https://github.com/kubeflow/pipelines/blob/2.0.0b13/samples/v2/hello_world.py 2. use the UI to create a pipeline run after uploading the pipeline yaml from the UI ### Expected result pipeline should run successfully ### Materials and Reference <!-- Help us debug this issue by providing resources such as: sample code, background context, or links to references. --> --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9119/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9113

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9113/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9113/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9113/events

https://github.com/kubeflow/pipelines/issues/9113

I_kwDOB-71UM5i0OTQ

[sdk] UnboundLocalError: local variable 'artifact_class' referenced before assignment

open

NONE

### Environment * KFP version: <!-- For more information, see an overview of KFP installation options: https://www.kubeflow.org/docs/pipelines/installation/overview/. --> * KFP SDK version: 2.0.0-beta.13 <!-- Specify the version of Kubeflow Pipelines that you are using. The version number appears in the left side navigation of user interface. To find the version number, See version number shows on bottom of KFP UI left sidenav. --> * All dependencies version: <!-- Specify the output of the following shell command: $pip list | grep kfp --> ### Steps to reproduce In the attached line, the variable `artifact_class` is undefined within the else block. A unit test where the condition `type_annotations.is_artifact_class(annotation)` triggers should be sufficient to reproduce I think. https://github.com/kubeflow/pipelines/blame/e7c82c05933e52a5184d20a1e255939d5302189a/sdk/python/kfp/components/types/custom_artifact_types.py#L58 <!-- Specify how to reproduce the problem. This may include information such as: a description of the process, code snippets, log output, or screenshots. --> ### Expected result <!-- What should the correct behavior be? --> ### Materials and Reference <!-- Help us debug this issue by providing resources such as: sample code, background context, or links to references. --> Partial stacktrace: ``` File "/usr/local/lib/python3.7/site-packages/kfp/components/component_decorator.py", line 127, in component kfp_package_path=kfp_package_path) File "/usr/local/lib/python3.7/site-packages/kfp/components/component_factory.py", line 448, in create_component_from_func func=func) File "/usr/local/lib/python3.7/site-packages/kfp/components/component_factory.py", line 361, in _get_command_and_args_for_lightweight_component ] + custom_artifact_types.get_custom_artifact_type_import_statements(func) File "/usr/local/lib/python3.7/site-packages/kfp/components/types/custom_artifact_types.py", line 29, in get_custom_artifact_type_import_statements artifact_imports = get_custom_artifact_import_items_from_function(func) File "/usr/local/lib/python3.7/site-packages/kfp/components/types/custom_artifact_types.py", line 173, in get_custom_artifact_import_items_from_function param_to_ann_obj = get_param_to_custom_artifact_class(func) File "/usr/local/lib/python3.7/site-packages/kfp/components/types/custom_artifact_types.py", line 58, in get_param_to_custom_artifact_class if artifact_class not in kfp_artifact_classes: UnboundLocalError: local variable 'artifact_class' referenced before assignment ``` --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9113/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9110

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9110/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9110/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9110/events

https://github.com/kubeflow/pipelines/issues/9110

I_kwDOB-71UM5iw0oU

[backend] Argo Workflows is Using Legacy Pod Patches

open

NONE

When executing any pipeline, the following lines are logged in the `wait` container: ``` # ... time="2023-04-06T07:58:47.339Z" level=info msg="Create workflowtaskresults 403" time="2023-04-06T07:58:47.339Z" level=warning msg="failed to patch task set, falling back to legacy/insecure pod patch, see https://argoproj.github.io/argo-workflows/workflow-rbac/" error="workflowtaskresults.argoproj.io is forbidden: User \"system:serviceaccount:kubeflow-user-example-com:default-editor\" cannot create resource \"workflowtaskresults\" in API group \"argoproj.io\" in the namespace \"kubeflow-user-example-com\"" time="2023-04-06T07:58:47.356Z" level=info msg="Patch pods 200" time="2023-04-06T07:58:47.362Z" level=info msg="Killing sidecars []" time="2023-04-06T07:58:47.362Z" level=info msg="Alloc=7530 TotalAlloc=12844 Sys=24530 NumGC=4 Goroutines=12" ``` In addition to Argo considering this method to be legacy / insecure, we've had issue with pipeline steps being stuck / not being displayed as succeeded in the UI, depending on the files that were written to the OutputPath, as the resulting pod's annotation was malformed. Fixing this RBAC issue is easily done by adding the resource `workflowtaskresults` to the [aggregate-to-kubeflow-pipelines-edit ClusterRole](https://github.com/kubeflow/pipelines/blob/1caa8e0d6889e50cde217a18abddb2c17df6ef6e/manifests/kustomize/base/installs/multi-user/view-edit-cluster-roles.yaml#L100). However, this will cause Argo Workflow to no longer write the annotation `workflows.argoproj.io/outputs` to the pipeline's pods, which several Kubeflow components rely on, e.g.: - [metadata_writer](https://github.com/kubeflow/pipelines/blob/3f4d686f6a04d0e2c2bda551af82981e329ed95c/backend/metadata_writer/src/metadata_writer.py#L72) - [cache mutation](https://github.com/kubeflow/pipelines/blob/3f4d686f6a04d0e2c2bda551af82981e329ed95c/backend/src/cache/server/mutation.go#L44)

https://api.github.com/repos/kubeflow/pipelines/issues/9110/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9108

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9108/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9108/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9108/events

https://github.com/kubeflow/pipelines/issues/9108

I_kwDOB-71UM5iuJ36

[sdk] set_timeout ignored in component in kfp pipeline

open

NONE

### Environment * KFP version: 1.8.19 * KFP SDK version: We are using with GCP Vertex AI Interface * All dependencies version: kfp 1.8.19 pypi_0 pypi kfp-pipeline-spec 0.1.16 pypi_0 pypi kfp-server-api 1.8.4 pypi_0 pypi ### Steps to reproduce I'm trying to use the `set_timeout(...)` method in a component but it seems to be ignored in the compile phase. The idea of using the `set_timeout(...)` is to enforce the maximum runtime of a step in our pipeline. The json file generated by the `compiler.Compiler().compile(pipeline_func=pipeline, package_path=build_path)` does not change at all if I use the set_timeout or not in the component. Examples: Test Component: ``` @component def test_comp() -> str: import time print("going to sleep") time.sleep(360) # Sleep print("waking up") return "Hello There" ``` Calling it inside our pipeline and trying to use a 100 seconds timeout: ``` @dsl.pipeline(name="test-pipeline") def pipeline(): test_comp_op = test_comp().set_timeout(100) ``` Generates the same json file as: ``` @dsl.pipeline(name="test-pipeline") def pipeline(): test_comp_op = test_comp() ``` ### Expected result The pipeline step is terminated after reaching n seconds in the set_timeout method. ### Materials and Reference JSON generated Before: ``` { "pipelineSpec": { "components": { "comp-test-comp": { "executorLabel": "exec-test-comp", "outputDefinitions": { "parameters": { "Output": { "type": "STRING" } } } } }, "deploymentSpec": { "executors": { "exec-test-comp": { "container": { "args": [ "--executor_input", "{{$}}", "--function_to_execute", "test_comp" ], "command": [ "sh", "-c", "\nif ! [ -x \"$(command -v pip)\" ]; then\n python3 -m ensurepip || python3 -m ensurepip --user || apt-get install python3-pip\nfi\n\nPIP_DISABLE_PIP_VERSION_CHECK=1 python3 -m pip install --quiet --no-warn-script-location 'kfp==1.8.19' && \"$0\" \"$@\"\n", "sh", "-ec", "program_path=$(mktemp -d)\nprintf \"%s\" \"$0\" > \"$program_path/ephemeral_component.py\"\npython3 -m kfp.v2.components.executor_main --component_module_path \"$program_path/ephemeral_component.py\" \"$@\"\n", "\nimport kfp\nfrom kfp.v2 import dsl\nfrom kfp.v2.dsl import *\nfrom typing import *\n\ndef test_comp() -> str:\n import time\n print(\"going to sleep\")\n time.sleep(360) # Sleep \n print(\"waking up\")\n return \"Hello There\"\n\n" ], "image": "python:3.7" } } } }, "pipelineInfo": { "name": "teste-pipeline" }, "root": { "dag": { "tasks": { "test-comp": { "cachingOptions": { "enableCache": true }, "componentRef": { "name": "comp-test-comp" }, "taskInfo": { "name": "test-comp" } } } } }, "schemaVersion": "2.0.0", "sdkVersion": "kfp-1.8.19" }, "runtimeConfig": {} } ``` JSON generated After: ``` { "pipelineSpec": { "components": { "comp-test-comp": { "executorLabel": "exec-test-comp", "outputDefinitions": { "parameters": { "Output": { "type": "STRING" } } } } }, "deploymentSpec": { "executors": { "exec-test-comp": { "container": { "args": [ "--executor_input", "{{$}}", "--function_to_execute", "test_comp" ], "command": [ "sh", "-c", "\nif ! [ -x \"$(command -v pip)\" ]; then\n python3 -m ensurepip || python3 -m ensurepip --user || apt-get install python3-pip\nfi\n\nPIP_DISABLE_PIP_VERSION_CHECK=1 python3 -m pip install --quiet --no-warn-script-location 'kfp==1.8.19' && \"$0\" \"$@\"\n", "sh", "-ec", "program_path=$(mktemp -d)\nprintf \"%s\" \"$0\" > \"$program_path/ephemeral_component.py\"\npython3 -m kfp.v2.components.executor_main --component_module_path \"$program_path/ephemeral_component.py\" \"$@\"\n", "\nimport kfp\nfrom kfp.v2 import dsl\nfrom kfp.v2.dsl import *\nfrom typing import *\n\ndef test_comp() -> str:\n import time\n print(\"going to sleep\")\n time.sleep(360) # Sleep \n print(\"waking up\")\n return \"Hello There\"\n\n" ], "image": "python:3.7" } } } }, "pipelineInfo": { "name": "teste-pipeline" }, "root": { "dag": { "tasks": { "test-comp": { "cachingOptions": { "enableCache": true }, "componentRef": { "name": "comp-test-comp" }, "taskInfo": { "name": "test-comp" } } } } }, "schemaVersion": "2.0.0", "sdkVersion": "kfp-1.8.19" }, "runtimeConfig": {} } ``` --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9108/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9102

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9102/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9102/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9102/events

https://github.com/kubeflow/pipelines/issues/9102

I_kwDOB-71UM5imwvh

[feature] Add environment variables support for ContainerSpec when definining Container Component in v2

closed

NONE

### Feature Area area sdk ### What feature would you like to see? Add environment variables support to the ContainerSpec class when defining a container_component in v2 code improving reusability on the container components. Is odd that is not implemented because is a fairly easy change of [code](https://github.com/kubeflow/pipelines/blob/master/sdk/python/kfp/components/structures.py#LL324C16-L324C16),I open a PR that should did the trick. ### What is the use case or pain point? Can input the env variables that you want especially when using a container image that is used on other projects and is a problem to modify. ### Is there a workaround currently? <!-- Without this feature, how do you accomplish your task today? --> I need to define an environment variable direct on the Dockerfile or put them on args and add an export command to the container image, but this is a lot less elegant --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9102/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9100

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9100/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9100/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9100/events

https://github.com/kubeflow/pipelines/issues/9100

I_kwDOB-71UM5ijkgf

[backend] <Artifacts are not accepted as pipeline Parameters>

closed

NONE

### Environment google-cloud-aiplatform==1.11.0 google-cloud-bigquery==1.0.20 google-cloud-pipeline-components==1.0.20 google-api-core==1.32.0 * How did you deploy Kubeflow Pipelines (KFP)? <!-- For more information, see an overview of KFP installation options: https://www.kubeflow.org/docs/pipelines/installation/overview/. --> Deployed on google cloud platform * KFP version: <!-- Specify the version of Kubeflow Pipelines that you are using. The version number appears in the left side navigation of user interface. To find the version number, See version number shows on bottom of KFP UI left sidenav. --> * KFP SDK version: <!-- Specify the output of the following shell command: $pip list | grep kfp --> kfp 1.8.13 kfp-pipeline-spec 0.1.16 kfp-server-api 1.8.5 I am running kfp on Pycharm not on the UI, so I just past on the packages related to kfp here. ### Steps to reproduce <!-- Specify how to reproduce the problem. This may include information such as: a description of the process, code snippets, log output, or screenshots. --> 1. I created a python function component using @component decorator, and I used the Input[Dataset], for the input. the whole specification of function looks like this: ```python def training_auto_arima_forecasting( data_in_y_train: Input[Dataset], data_in_y_test: Input[Dataset], feature_processor: Input[Artifact], metrics: Output[Metrics], model: Output[Model] ): ``` 2. When I compile the pipeline, I got the following error: TypeError: The pipeline argument "data_in_y_train" is viewed as an artifact due to its type "Dataset". And we currently do not support passing artifacts as pipeline inputs. Consider type annotating the argument with a primitive type, such as "str", "int", "float", "bool", "dict", and "list". ### Expected result <!-- What should the correct behavior be? --> I expect the annotation is correct according to the type specification given by documentation for kfp v2. ### Materials and Reference <!-- Help us debug this issue by providing resources such as: sample code, background context, or links to references. --> this is implementation for the pipeline ``` Python def auto_arima_pipeline(project_id: str, data_set_id: str, view_name: str, model_name: str, deploy_endpoint: str, metric_name: str, threshold: str, version: str): with dsl.ExitHandler(notify_email_task): dataset_ingestion = ingest_data(project_id=project_id, data_set_id=data_set_id, view_name=view_name) preprocessed_data = preprocessing(dataset_ingestion.outputs['data_out_train']) features_data = time_series_transformation(preprocessed_data.outputs['data_out']) model_build = training_auto_arima_forecasting( features_data.outputs['data_out_train'], features_data.outputs['data_out_test'], features_data.outputs['transformation_pipeline'] ) ``` This is time seies feature transformation component ```python def time_series_transformation( data_out_train: Output[Dataset], data_out_test: Output[Dataset], transformation_pipeline: Output[Artifact], data_path: Input[Dataset] ) ->NamedTuple("Outputs", [("column_names", list)]): data = pd.read_csv(data_path.path + ".csv", index_col=False) y = data.copy() # Temperal Data Split y_train, y_test = temporal_train_test_split(y, test_size=36) ### Export data as artifact ### pd.DataFrame(y_train).to_csv(data_out_train.path + ".csv") pd.DataFrame(y_test).to_csv(data_out_test.path + ".csv") full_processor = ColumnTransformer( transformers=[] ) file_name = transformation_pipeline.path + f".pkl" with open(file_name, 'wb') as file: pickle.dump(full_processor, file) column_names = list(y.columns) return (column_names,) ``` --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9100/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9092

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9092/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9092/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9092/events

https://github.com/kubeflow/pipelines/issues/9092

I_kwDOB-71UM5idQin

[bug] Missing backticks in BigQuery table reference

open

NONE

### Environment Error is related to how a query is formatted for passing to BigQuery. ### Steps to reproduce Use the prebuilt pipeline at: https://github.com/kubeflow/pipelines/blob/master/components/google-cloud/google_cloud_pipeline_components/experimental/automl/forecasting/prophet_predict_pipeline.yaml Where the input table name requires a [quoted identifier](https://cloud.google.com/bigquery/docs/reference/standard-sql/lexical). An error is returned by the component `bigquery-query-job` with `taskInfo: name: remove-feature-columns` ### Expected result Accept any allowable BigQuery table name as input. Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9092/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9087

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9087/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9087/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9087/events

https://github.com/kubeflow/pipelines/issues/9087

I_kwDOB-71UM5iaS4G

[backend] Resource limits and requests not working in KFP v2

closed

MEMBER

### Environment * How did you deploy Kubeflow Pipelines (KFP)? Standalone KFP from master * KFP SDK version: 2.0.0-beta.13 ### Steps to reproduce 1. Create a pipeline version with `.set_memory_limit(A).set_memory_request(B).set_cpu_limit(C).set_cpu_request(D)` 2. Create a run from the pipeline 3. Check the corresponding pod's `spec.requests` ### Expected result ```yaml resources: limits: cpu: C memory: A requests: cpu: D memory: B ``` ### Materials and Reference ```yaml resources: limits: cpu: 500m memory: 512Mi requests: cpu: 10m memory: 32Mi ``` The values are hardcoded and do not depend on what we set as `A`, `B`, `C`, and `D`. They seem to come from `workflow-controller-configmap`. The change was probably introduced in #5695. --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9087/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9081

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9081/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9081/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9081/events

https://github.com/kubeflow/pipelines/issues/9081

I_kwDOB-71UM5iV1xZ

[sdk] Python SDK incompatible with other kubeflow operators

closed

CONTRIBUTOR

### Environment * KFP version: 2.0.0-alpha.7 <!-- For more information, see an overview of KFP installation options: https://www.kubeflow.org/docs/pipelines/installation/overview/. --> * KFP SDK version: 1.8.19 <!-- Specify the version of Kubeflow Pipelines that you are using. The version number appears in the left side navigation of user interface. To find the version number, See version number shows on bottom of KFP UI left sidenav. --> * All dependencies version: <!-- Specify the output of the following shell command: $pip list | grep kfp --> ### Steps to reproduce pip install kfp 1.8.19 and attempt to pip install ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts ``` the conflict is caused by: kfp 1.8.19 depends on kubernetes<20 and >=8.0.0 kserve 0.10.0 depends on kubernetes>=18.20.0 kubeflow-training 1.6.0 depends on kubernetes>=23.6.0 ``` <!-- Specify how to reproduce the problem. This may include information such as: a description of the process, code snippets, log output, or screenshots. --> ### Expected result <!-- What should the correct behavior be? --> We expect to be able to use different kubeflow operators together or in the same environment ### Materials and Reference <!-- Help us debug this issue by providing resources such as: sample code, background context, or links to references. --> --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9081/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9078

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9078/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9078/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9078/events

https://github.com/kubeflow/pipelines/issues/9078

I_kwDOB-71UM5iVqzs

[feature] Support TASK_FINAL_STATUS field

open

MEMBER

### Feature Area <!-- Uncomment the labels below which are relevant to this feature: --> <!-- /area frontend --> /area backend <!-- /area sdk --> <!-- /area samples --> <!-- /area components --> ### What feature would you like to see? Support TASK_FINAL_STATUS as added in https://github.com/kubeflow/pipelines/pull/9061. The BE should permit both the new TASK_FINAL_STATUS type and STRUCT type on components that use the [InputParameterSpec.task_final_status](https://github.com/kubeflow/pipelines/blob/ffd3a75d638b54b144f56df8be2e0b54bcc14f71/api/v2alpha1/pipeline_spec.proto#L322) input. ### What is the use case or pain point? Enables compiling/loading components/pipelines with PipelineTaskFinalStatus as root input type. Needed to fix https://github.com/kubeflow/pipelines/issues/8649. --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9078/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9074

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9074/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9074/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9074/events

https://github.com/kubeflow/pipelines/issues/9074

I_kwDOB-71UM5iSSkI

[sdk] Can't filter pipelines by namespace/profile

closed

NONE

### Environment * KFP version: $pip list | grep kfp --> kfp 1.8.19 kfp-pipeline-spec 0.1.16 kfp-server-api 1.6.0 ### Steps to reproduce <!-- Specify how to reproduce the problem. This may include information such as: a description of the process, code snippets, log output, or screenshots. --> 1. Log in Kubeflow and select your namespace/profile and go to Pipelines component 2. I can see the pipelines uploaded by other teams(irrespective of profile) and others can view my pipelines as well ### Expected result <!-- What should the correct behavior be? --> Behaviour should be same as Runs tab. Others pipelines should be hidden as the yaml might hold confidential information ### Materials and Reference <!-- Help us debug this issue by providing resources such as: sample code, background context, or links to references. --> --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9074/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9065

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9065/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9065/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9065/events

https://github.com/kubeflow/pipelines/issues/9065

I_kwDOB-71UM5iNprO

[feature] Change gcpc model_deploy input param 'Endpoint' to a required field

open

NONE

### Feature Area /area sdk ### What feature would you like to see? The value for gcpc's model_deploy's input param endpoint is [true for `optional`](https://github.com/kubeflow/pipelines/blob/c32868c2fd9fd2a5ce375484408503a56d7fdf50/components/google-cloud/google_cloud_pipeline_components/v1/endpoint/deploy_model/component.yaml#LL114C32-L114C32). Based on the description, it should be [set to `false`](https://github.com/kubeflow/pipelines/blob/c32868c2fd9fd2a5ce375484408503a56d7fdf50/components/google-cloud/google_cloud_pipeline_components/v1/endpoint/deploy_model/component.yaml#L10) to indicate that it is a required field. ### What is the use case or pain point? User application reads the optional flags in inputs to render UI correctly. ### Is there a workaround currently? User application will have to dig into the file to find the optional inputs. --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9065/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9062

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9062/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9062/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9062/events

https://github.com/kubeflow/pipelines/issues/9062

I_kwDOB-71UM5iKO2m

[frontend] Flaky integration test for changing input parameter

closed

COLLABORATOR

https://github.com/kubeflow/pipelines/blob/master/test/frontend-integration-test/helloworld.spec.js#L181 The test should be always passed. However, It keep failing currently. Change the parameter from `testValue` to `defaultValue` first to make the test pass, will investigate later on. Part of #9001

https://api.github.com/repos/kubeflow/pipelines/issues/9062/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9055

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9055/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9055/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9055/events

https://github.com/kubeflow/pipelines/issues/9055

I_kwDOB-71UM5h9QrG

[feature] Setting name for worker in dsl.ParallelFor

open

NONE

### Feature Area <!-- Uncomment the labels below which are relevant to this feature: --> <!-- /area frontend --> <!-- /area backend --> <!-- /area sdk --> <!-- /area samples --> area components ### What feature would you like to see? <!-- Provide a description of this feature and the user experience. --> Possibility to set a suffix for workers in dsl.ParallelFor component ### What is the use case or pain point? I need to parallelize two different model and merge their prediction. Now I cannot see which worker works on which model unless opening a component and looking its argument. <!-- It helps us understand the benefit of this feature for your use case. --> ### Is there a workaround currently? No workaround because LoopArgument are initialized inside ParallelFor <!-- Without this feature, how do you accomplish your task today? --> --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9055/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9049

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9049/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9049/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9049/events

https://github.com/kubeflow/pipelines/issues/9049

I_kwDOB-71UM5h7Ckp

docs(migration): Update deployment doc with callout that users can continue use v1 SDK with v2 deployment.

closed

COLLABORATOR

https://api.github.com/repos/kubeflow/pipelines/issues/9049/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9034

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9034/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9034/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9034/events

https://github.com/kubeflow/pipelines/issues/9034

I_kwDOB-71UM5hoARo

[feature] Python SDK function to generate executor input

open

NONE

### Feature Area <!-- Uncomment the labels below which are relevant to this feature: --> <!-- /area frontend --> <!-- /area backend --> sdk <!-- /area samples --> <!-- /area components --> ### What feature would you like to see? A Python function that generates the executor input, given the component yaml, or the component function, and a dictionary of (specific) function parameter values to pass. For example, looking at [this test](https://github.com/kubeflow/pipelines/blob/707916413554be053726bc32f0140021910b0844/sdk/python/kfp/components/executor_test.py#L359) with [this test function](https://github.com/kubeflow/pipelines/blob/707916413554be053726bc32f0140021910b0844/sdk/python/kfp/components/executor_test.py#L379), the new input generator function could look like this: ``` input_gen(test_func, values={"first": 0.0, "second": 1.2}) ``` and the output to `input_gen` would be the correctly formatted executor input string: ``` """\ { "inputs": { "parameterValues": { "first": 0.0, "second": 1.2 } }, "outputs": { "parameters": { "Output": { "outputFile": "gs://some-bucket/output" } }, "outputFile": "%(test_dir)s/output_metadata.json" } } """ ``` ### What is the use case or pain point? The file `pipelines/sdk/python/kfp/components/executor_test.py` creates hard-coded executor inputs to test the executor [like here](https://github.com/kubeflow/pipelines/blob/707916413554be053726bc32f0140021910b0844/sdk/python/kfp/components/executor_test.py#L63). Specifying the executor input JSON is arguably one of the most difficult and tedious tasks of creating these tests, especially with complex data types. My team writes similar tests, but has found it difficult to wade through the nuances (for example when OutputPath(str) is a parameter, but OutputPath("Model") is an artifact) in hard coding executor input for testing and debugging. A Python function that generates the executor input, given the component yaml, or the component function would save time, and it would be less error prone. It would improve KFP component testing functionality, which has been a persistent need. <!-- It helps us understand the benefit of this feature for your use case. --> ### Is there a workaround currently? Currently, we hard code executor input per test, which doesn't scale well. There may be a way to generate the executor input using the backend go scripts, but it's not easily accessible for Python testing. <!-- Without this feature, how do you accomplish your task today? --> --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9034/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9033

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9033/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9033/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9033/events

https://github.com/kubeflow/pipelines/issues/9033

I_kwDOB-71UM5hnItp

[sdk] High Risk CVE with Kubernetes library inside KFP SDK

closed

NONE

### Environment * KFP version: 1.8.2 * KFP SDK version: 1.8.19 * All dependencies version: * kfp 1.8.19 * kfp-pipeline-spec 0.1.16 * kfp-server-api 1.8.5 ### Steps to reproduce The library kubernetes version <20 (specifically 19.15.0) was detected in KFP SDK 1.8.19 and is vulnerable to CVE-2021-29923, which exists in kubernetes versions <25.3.0. The vulnerability was found in the Pyup.io Safety DB with vendor severity: High ([NVD](https://nvd.nist.gov/vuln/detail/CVE-2021-29923) severity: High). This vulnerability has a known exploit available. Source: Github [[1](https://github.com/golang/go/issues/30999), [2](https://github.com/sickcodes/security/blob/master/advisories/SICK-2021-016.md)]. The vulnerability can be remediated by updating the library to version 25.3.0 or higher currently there is a hard requirement for Kubernetes library version to be below 20 based on the [requirements.txt](https://github.com/kubeflow/pipelines/blob/1.8.19/sdk/python/requirements.in) in v1.8.19 ### Expected result THE CVE's are resolved ### Materials and Reference <!-- Help us debug this issue by providing resources such as: sample code, background context, or links to references. --> --- Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9033/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9032

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9032/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9032/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9032/events

https://github.com/kubeflow/pipelines/issues/9032

I_kwDOB-71UM5hl27N

Hardcoded image makes it impossible to run pipelines that 1) require custom dependencies 2) have no open internet access due to Organisational policies.

open

NONE

As per the title, hardcoding images URIs make it impossible to run components that require custom dependencies. For instance, [the DataflowPythonJobOp](https://github.com/kubeflow/pipelines/blob/707916413554be053726bc32f0140021910b0844/components/google-cloud/google_cloud_pipeline_components/v1/dataflow/python_job/component.py#L44) uses the default image and accepts a ref to a `requirements.txt` file to pull dependencies at runtime. If the function that instantiates the Dataflow Pipeline uses any dependency not included in the provided image the job will fail. Example to reproduce (assuming an environment under VPS-SC or similar that limits networking with the open internet): ```python # pipeline.py , the code containing the actual dataflow pipeline. # uploaded to gs://my-bucket/pipeline.py for example import apache_beam as beam import tensorflow as tf # Fails here from apache_beam.options.pipeline_options import PipelineOptions, SetupOptions parser = argparse.ArgumentParser() known_args, pipeline_args = parser.parse_known_args() pipeline_options = PipelineOptions(pipeline_args) pipeline_options.view_as(SetupOptions).save_main_session = True with beam.Pipeline(options=pipeline_options) as p: data = p | beam.Create(iter(tf.random.normal(shape=(100, 3, 3)).numpy())) rounded = ( data | beam.Map(lambda x: tf.squeeze(tf.reshape(tf.math.round(x), shape=(1, -1)))) ) rounded | beam.Map(print) ``` And the code for starting the pipeline: ```python @dsl.pipeline(name=PIPELINE_NAME, description="Dataflow launch python pipeline") def pipeline( python_file_path: str = "gs://my-bucket/pipeline.py", project_id: str = PROJECT_ID, location: str = LOCATION, staging_dir: str = os.path.join(PIPELINE_ROOT, "staging"), requirements_file_path: str = "", ): args = [ "--runner", "DataflowRunner", "--project", PROJECT_ID, "--region", LOCATION, "--temp_location", staging_dir, "--stating_location", staging_dir, "--experiments", "use_runner_v2", # Dataflow workers have access to everything since you can specify a given image with # the dependencies installed. "--sdk_container_image", f"europe-west4-docker.pkg.dev/{PROJECT_ID}/mlops/beam_image/data_preparation_pipeline:latest", ] dataflow_python_op = DataflowPythonJobOp( project=project_id, location=location, python_module_path=python_file_path, temp_location=staging_dir, requirements_file_path=requirements_file_path, args=args, ) dataflow_wait_op = WaitGcpResourcesOp( gcp_resources=dataflow_python_op.outputs["gcp_resources"] ) if __name__ == "__main__": compiler.Compiler().compile(pipeline_func=pipeline, package_path="pipeline.json") aip.init( project=PROJECT_ID, location=LOCATION, ) # Prepare the pipeline job job = aip.PipelineJob( display_name="automl-image-training-v2", template_path="pipeline.json", pipeline_root=PIPELINE_ROOT, parameter_values={"project_id": PROJECT_ID}, ) job.submit() ``` I suggest moving the image declaration to a function argument with the default parameter as the current image like so: ```python @container_component def dataflow_python( project: str, python_module_path: str, temp_location: str, gcp_resources: OutputPath(str), location: str = 'us-central1', requirements_file_path: str = '', args: List[str] = [], image: str = 'gcr.io/ml-pipeline/google-cloud-pipeline-components:latest' ): """Launch a self-executing beam python file on Google Cloud using the DataflowRunner. Args: project (str): Required. Project to create the Dataflow job in. location (Optional[str]): Location for creating the Dataflow job. If not set, default to us-central1. python_module_path (str): The GCS path to the python file to run. temp_location (str): A GCS path for Dataflow to stage temporary job files created during the execution of the pipeline. requirements_file_path (Optional[str]): The GCS path to the pip requirements file. args(Optional[List[str]]): The list of args to pass to the python file. Can include additional parameters for the beam runner. image (str): The image used by the Container invoking the Dataflow job. Returns: gcp_resources (str): Serialized gcp_resources proto tracking the Dataflow job. For more details, see https://github.com/kubeflow/pipelines/blob/master/components/google-cloud/google_cloud_pipeline_components/proto/README.md. """ return ContainerSpec( image=image, command=[ 'python3', '-u', '-m', 'google_cloud_pipeline_components.container.v1.dataflow.dataflow_launcher' ], args=[ '--project', project, '--location', location, '--python_module_path', python_module_path, '--temp_location', temp_location, '--requirements_file_path', requirements_file_path, '--args', args, '--gcp_resources', gcp_resources, ]) ``` This way users can bake their images using the provided one as base image AND install custom dependencies without relying on internet access. Similar considerations can be made with other components (e.g. [HyperParameterTuningJob](https://github.com/kubeflow/pipelines/blob/707916413554be053726bc32f0140021910b0844/components/google-cloud/google_cloud_pipeline_components/v1/hyperparameter_tuning_job/component.py#L166) when users might want to use another optimiser). Happy to open a MR once this is accepted.

https://api.github.com/repos/kubeflow/pipelines/issues/9032/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9031

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9031/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9031/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9031/events

https://github.com/kubeflow/pipelines/issues/9031

I_kwDOB-71UM5hlZIY

[feature] How to save kubeflow logs in GCP Stackdriver Monitoring for specific interval of time (like 3 months, 6 months) etc ?

open

NONE

Hi folks , I need to save Kubeflow pipeline logs in GCP Stackdriver Monitoring for specific interval of time (like 3 months, 6 months) etc. How can I can I achieve something like that ? Any configurations that I need to change ? As of now I get the below error prompt: <img width="788" alt="image" src="https://user-images.githubusercontent.com/42741300/227166189-1b5bc4c4-a96e-433e-a8c0-584c3f85d1cc.png"> Also can we store Kubeflow pipeline logs to GCS buckets ?

https://api.github.com/repos/kubeflow/pipelines/issues/9031/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9019

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9019/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9019/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9019/events

https://github.com/kubeflow/pipelines/issues/9019

I_kwDOB-71UM5ha7cV

docs(samples): Update/Migrate v1 samples to be v2-compliant.

open

COLLABORATOR

https://api.github.com/repos/kubeflow/pipelines/issues/9019/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/9016

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9016/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9016/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9016/events

https://github.com/kubeflow/pipelines/issues/9016

I_kwDOB-71UM5hVLWR

[backend] Backend api should accept pipelineSpec in both current format and a nested struct

closed

COLLABORATOR

### Environment * How did you deploy Kubeflow Pipelines (KFP)? N/A * KFP version: Beta 0 * KFP SDK version: N/A Currently, the backend API V2 accepts pipeline spec as a struct (see [here](https://github.com/kubeflow/pipelines/blob/cc30d1aa85eb25fb20a6d2531956b99e5a9ec04d/backend/api/v2beta1/run.proto#L167-L168)). Since we want to support platform-specific features, we will turn this generic struct into a nested format: `Struct{pipeline_spec: <Struct>, kubernetes_config: <Struct>, ...}`. Related: #8983 --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9016/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9012

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9012/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9012/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9012/events

https://github.com/kubeflow/pipelines/issues/9012

I_kwDOB-71UM5hTVde

[backend] ApiParameter in v1 recurring run is missing.

closed

COLLABORATOR

V1 one-off run with `ApiParameter` <img width="1497" alt="Screenshot 2023-03-17 at 2 00 55 PM" src="https://user-images.githubusercontent.com/56132941/226404076-8a18b9fc-9c90-4987-90c3-18def876a96d.png"> V1 recurring run from same pipeline is missing `ApiParameter` 

https://api.github.com/repos/kubeflow/pipelines/issues/9012/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9011

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9011/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9011/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9011/events

https://github.com/kubeflow/pipelines/issues/9011

I_kwDOB-71UM5hQjUA

[sdk] Not able to install GCP artifact registry python package using pip_index_urls parameter in component

closed

NONE

* KFP SDK version: 1.8.19 Hi folks I need to install a private package from gcp artifact registry using pip_index_urls of the @component decorator in Kubeflow pipeline which is on GCP cluster. My index url looks like this => pip_index_urls = ["https://location-python.pkg.dev/project-id/repo-name/simple/ package_name==1.0.0"] ``` @component(pip_index_urls = ["https://location-python.pkg.dev/project-id/repo-name/simple/ package_name==1.0.0"],base_image="python:3.9") def process_data( output_data : Output[Dataset]): from private_package import some_function some_function() ``` GCP Artifact python index url example: ``` pip install --index-url https://us-central1-python.pkg.dev/PROJECT_ID/quickstart-python-repo/simple/ sampleproject ``` Reference : [GCP Documentation](https://cloud.google.com/artifact-registry/docs/python/store-python#download) When I use this. It gives me a weird error. ``` time="2023-03-20T08:03:19.432Z" level=info msg="capturing logs" argo=true I0320 08:03:19.455992 21 cache.go:120] Connecting to cache endpoint --------- ERROR: Exception: Traceback (most recent call last): File "/usr/local/lib/python3.9/site-packages/pip/_internal/cli/base_command.py", line 167, in exc_logging_wrapper status = run_func(*args) File "/usr/local/lib/python3.9/site-packages/pip/_internal/cli/req_command.py", line 205, in wrapper return func(self, options, args) File "/usr/local/lib/python3.9/site-packages/pip/_internal/commands/install.py", line 339, in run requirement_set = resolver.resolve( File "/usr/local/lib/python3.9/site-packages/pip/_internal/resolution/resolvelib/resolver.py", line 94, in resolve result = self._result = resolver.resolve( File "/usr/local/lib/python3.9/site-packages/pip/_vendor/resolvelib/resolvers.py", line 481, in resolve state = resolution.resolve(requirements, max_rounds=max_rounds) File "/usr/local/lib/python3.9/site-packages/pip/_vendor/resolvelib/resolvers.py", line 348, in resolve self._add_to_criteria(self.state.criteria, r, parent=None) File "/usr/local/lib/python3.9/site-packages/pip/_vendor/resolvelib/resolvers.py", line 172, in _add_to_criteria if not criterion.candidates: File "/usr/local/lib/python3.9/site-packages/pip/_vendor/resolvelib/structs.py", line 151, in __bool__ return bool(self._sequence) File "/usr/local/lib/python3.9/site-packages/pip/_internal/resolution/resolvelib/found_candidates.py", line 155, in __bool__ return any(self) File "/usr/local/lib/python3.9/site-packages/pip/_internal/resolution/resolvelib/found_candidates.py", line 143, in <genexpr> return (c for c in iterator if id(c) not in self._incompatible_ids) File "/usr/local/lib/python3.9/site-packages/pip/_internal/resolution/resolvelib/found_candidates.py", line 44, in _iter_built for version, func in infos: File "/usr/local/lib/python3.9/site-packages/pip/_internal/resolution/resolvelib/factory.py", line 294, in iter_index_candidate_infos result = self._finder.find_best_candidate( File "/usr/local/lib/python3.9/site-packages/pip/_internal/index/package_finder.py", line 868, in find_best_candidate candidates = self.find_all_candidates(project_name) File "/usr/local/lib/python3.9/site-packages/pip/_internal/index/package_finder.py", line 809, in find_all_candidates page_candidates = list(page_candidates_it) File "/usr/local/lib/python3.9/site-packages/pip/_internal/index/sources.py", line 134, in page_candidates yield from self._candidates_from_page(self._link) File "/usr/local/lib/python3.9/site-packages/pip/_internal/index/package_finder.py", line 769, in process_project_url html_page = self._link_collector.fetch_page(project_url) File "/usr/local/lib/python3.9/site-packages/pip/_internal/index/collector.py", line 566, in fetch_page return _get_html_page(location, session=self.session) File "/usr/local/lib/python3.9/site-packages/pip/_internal/index/collector.py", line 471, in _get_html_page resp = _get_html_response(url, session=session) File "/usr/local/lib/python3.9/site-packages/pip/_internal/index/collector.py", line 123, in _get_html_response resp = session.get( File "/usr/local/lib/python3.9/site-packages/pip/_vendor/requests/sessions.py", line 542, in get return self.request('GET', url, **kwargs) File "/usr/local/lib/python3.9/site-packages/pip/_internal/network/session.py", line 454, in request return super().request(method, url, *args, **kwargs) File "/usr/local/lib/python3.9/site-packages/pip/_vendor/requests/sessions.py", line 529, in request resp = self.send(prep, **send_kwargs) File "/usr/local/lib/python3.9/site-packages/pip/_vendor/requests/sessions.py", line 652, in send r = dispatch_hook('response', hooks, r, **kwargs) File "/usr/local/lib/python3.9/site-packages/pip/_vendor/requests/hooks.py", line 31, in dispatch_hook _hook_data = hook(hook_data, **kwargs) File "/usr/local/lib/python3.9/site-packages/pip/_internal/network/auth.py", line 270, in handle_401 username, password, save = self._prompt_for_password(parsed.netloc) File "/usr/local/lib/python3.9/site-packages/pip/_internal/network/auth.py", line 233, in _prompt_for_password username = ask_input(f"User for {netloc}: ") File "/usr/local/lib/python3.9/site-packages/pip/_internal/utils/misc.py", line 200, in ask_input return input(message) EOFError: EOF when reading a line F0320 08:03:22.196366 21 main.go:50] Failed to execute component: exit status 2 Error: exit status 1 User for asia-south1-python.pkg.dev: ``` Any workaround for this ?

https://api.github.com/repos/kubeflow/pipelines/issues/9011/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9008

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9008/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9008/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9008/events

https://github.com/kubeflow/pipelines/issues/9008

I_kwDOB-71UM5hJxcd

[backend] The ApiParameter in v1 run should store in v2 runtime config.

closed

COLLABORATOR

In v1 run, we use `APIParameter` in `pipeline_spec` to store the parameters. After introducing runtime_config concept in v2 run, we have to store the parameters into `RuntimeParameter` in `runtime_config`. Note: The change is also needed in recurring run. <img width="1497" alt="Screenshot 2023-03-17 at 2 00 55 PM" src="https://user-images.githubusercontent.com/56132941/226057903-7d674db1-5790-4359-a4cc-368486f1d269.png"> <img width="1497" alt="Screenshot 2023-03-17 at 2 02 30 PM" src="https://user-images.githubusercontent.com/56132941/226057936-f7a0ce0f-17e3-429d-92d7-9efc93c3d75c.png">

https://api.github.com/repos/kubeflow/pipelines/issues/9008/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9005

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9005/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9005/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9005/events

https://github.com/kubeflow/pipelines/issues/9005

I_kwDOB-71UM5hH_b3

[sdk] One can't set environment on v2 YamlComponent

closed

NONE

### Environment * Python Version: 3.10.6 * PIP Version: 22.3.1 (with Pipenv 2022.12.19) * KFP version: kfp 2.0.0b12 <!-- For more information, see an overview of KFP installation options: https://www.kubeflow.org/docs/pipelines/installation/overview/. --> * KFP SDK version: kfp 2.0.0b12 <!-- Specify the version of Kubeflow Pipelines that you are using. The version number appears in the left side navigation of user interface. To find the version number, See version number shows on bottom of KFP UI left sidenav. --> * All dependencies version: <!-- Specify the output of the following shell command: $pip list | grep kfp --> ``` kfp 2.0.0b12 kfp-pipeline-spec 0.2.0 kfp-server-api 2.0.0a6 ``` ### Steps to reproduce <!-- Specify how to reproduce the problem. This may include information such as: a description of the process, code snippets, log output, or screenshots. --> Given the V1 Yaml ComponentSpec: ```yaml name: Write Lines description: Writes text to a file. inputs: - {name: text, type: String} outputs: - {name: data, type: Data} implementation: container: image: busybox command: - sh - -c - | mkdir -p "$(dirname "$1")" echo "$GREETING $0" > "$1" args: - {inputValue: text} - {outputPath: data} env: GREETING: "Hello," ``` The generated artifact only contain the input "text" value. ### Expected result <!-- What should the correct behavior be? --> The generated artifact should have both Environment and Input values. ### Materials and Reference <!-- Help us debug this issue by providing resources such as: sample code, background context, or links to references. --> The code used to load YamlComponent: ```python from kfp import components yaml_component = components.load_component_from_file("filewriter.yml") ``` And the pipeline is as simple as it could be: ```python @dsl.pipeline( name='filewriter-pipeline', pipeline_root=PIPELINE_ROOT ) def workflow(text: str): filewriter_op = yaml_component(text=text) ``` Should be noticed that ComponenSpec [does create environment variables definitions](https://github.com/kubeflow/pipelines/blob/2.0.0b12/sdk/python/kfp/components/structures.py#L632) from file: ```python ... env = { key: placeholders.maybe_convert_v1_yaml_placeholder_to_v2_placeholder( command, component_dict=component_dict) for key, command in container.get('env', {}).items() } container_spec = ContainerSpecImplementation.from_container_dict({ 'image': container['image'], 'command': command, 'args': args, 'env': env }) ``` But ContainerSpecImplementation.from_container_dict [just ignores env existence](https://github.com/kubeflow/pipelines/blob/2.0.0b12/sdk/python/kfp/components/structures.py#L329): ```python return ContainerSpecImplementation( image=container_dict['image'], command=container_dict.get('command'), args=container_dict.get('args'), env=None, # can only be set on tasks resources=None) # can only be set on tasks ``` Is there any reason to do so? Is there any other means to set environment in runtime? --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9005/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/9003

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/9003/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/9003/comments

https://api.github.com/repos/kubeflow/pipelines/issues/9003/events

https://github.com/kubeflow/pipelines/issues/9003

I_kwDOB-71UM5hGCMA

[backend] Getting read_artifact from kfp.Client().runs results in internal server error

closed

NONE

### Environment * How did you deploy Kubeflow Pipelines (KFP)? Kubeflow deployed using Kubeflow manifests repo to Azure Kubernetes cluster * KFP version: 2.0.0-alpha3 * KFP SDK version: 1.8.19 ### Steps to reproduce Calling `read_artifact` on `client.runs`, similar to [this issue](https://github.com/kubeflow/pipelines/issues/4327) results in a 500 Internal Server error. Top of error message says `InternalServerError: failed to unmarshal workflow` and at bottom `Invalid input error: Unknown execution spec`. <img width="706" alt="image" src="https://user-images.githubusercontent.com/74452047/225863091-48ac11ca-111d-430a-9153-b6c413ff47de.png"> ### Expected result Previously, this has returned artifact as dictionary where `data` can be picked out to retrieve run output. ### Materials and Reference Following steps in [this issue](https://github.com/kubeflow/pipelines/issues/4327) for a run created with `create_run_from_pipeline_func` should reproduce error. --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/9003/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/8997

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8997/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8997/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8997/events

https://github.com/kubeflow/pipelines/issues/8997

I_kwDOB-71UM5hBkiP

[feature] Add support for ad hoc triggering of recurring runs to the KFP SDK

open

CONTRIBUTOR

### Feature Area <!-- Uncomment the labels below which are relevant to this feature: --> <!-- /area frontend --> <!-- /area backend --> /area sdk <!-- /area samples --> <!-- /area components --> ### What feature would you like to see? <!-- Provide a description of this feature and the user experience. --> Ad hoc trigger support for recurring runs in the KFP SDK (and eventually the UI). ### What is the use case or pain point? <!-- It helps us understand the benefit of this feature for your use case. --> Currently, there is no way to ad hoc trigger a recurring run in KFP. A common user experience is creating a recurring run then validating it works as expected by creating a run. If the cron expression is set to once a day, it's not reasonable to wait until the cron expression is satisfied. Argo Workflows makes this turn key in its UI and CLI. At the moment, this is not possible in KFP. ### Is there a workaround currently? <!-- Without this feature, how do you accomplish your task today? --> The only workaround currently is to take some of the parameters passed to `create_recurring_run` and manually pass them to `create_run_from_pipeline_package`. Part of the problem with this approach, in addition to the fact that it's not obvious or intuitive, is that there parameters of these functions vary in subtle ways. For example, `create_recurring_run` accepts an `experiment_id`, whereas `create_run_from_pipeline_func` accepts an `experiment_name`. --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/8997/timeline

reopened

https://api.github.com/repos/kubeflow/pipelines/issues/8996

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8996/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8996/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8996/events

https://github.com/kubeflow/pipelines/issues/8996

I_kwDOB-71UM5g-uUf

[sdk] No support for filtering on unrecognized field finished_at

closed

NONE

### Environment * KFP version: <!-- For more information, see an overview of KFP installation options: https://www.kubeflow.org/docs/pipelines/installation/overview/. --> 2.0.0-alpha.5 * KFP SDK version: Python SDK version: `2.0.0b13` * All dependencies version: <!-- Specify the output of the following shell command: $pip list | grep kfp --> ``` kfp 2.0.0b13 kfp-pipeline-spec 0.2.0 kfp-server-api 2.0.0a6 ``` ### Steps to reproduce Using the protobuf filter below to fitler KFP `list_runs()` in version `2.0.0b13` of the Python Package, ``` archive_filter = json.dumps({ "key": "finished_at", "op": "LESS_THAN_EQUALS", "timestamp_value": (datetime.now(timezone.utc) - timedelta(days=90)) .isoformat() .replace("+00:00", "Z") }] }) ``` I get the following exception: ``` kfp_server_api.exceptions.ApiException: (400) Reason: Bad Request HTTP response headers: HTTPHeaderDict({'Audit-Id': '042ba586-9789-4a1f-9065-d1f781f3ecc5', 'Cache-Control': 'no-cache, no-store, must-revalidate, no-cache, no-store, must-revalidate, no-cache, private', 'Content-Length': '530', 'Content-Type': 'application/json', 'Date': 'Wed, 15 Mar 2023 19:41:27 GMT', 'Expires': '0, 0', 'Pragma': 'no-cache, no-cache', 'Referrer-Policy': 'strict-origin, strict-origin', 'Server': 'istio-envoy', 'Strict-Transport-Security': 'max-age=31536000; includeSubDomains, max-age=31536000; includeSubDomains', 'X-Content-Type-Options': 'nosniff, nosniff', 'X-Envoy-Decorator-Operation': 'ml-pipeline.kubeflow.svc.cluster.local:8888/*', 'X-Envoy-Upstream-Service-Time': '1', 'X-Frame-Options': 'SAMEORIGIN, SAMEORIGIN', 'X-Xss-Protection': '1; mode=block, 1; mode=block'}) HTTP response body: {"error":"Failed to create list options: Invalid input error: no support for filtering on unrecognized field \"finished_at\"","code":3,"message":"Failed to create list options: Invalid input error: no support for filtering on unrecognized field \"finished_at\"","details":[{"@type":"type.googleapis.com/api.Error","error_message":"no support for filtering on unrecognized field \"finished_at\"","error_details":"Failed to create list options: Invalid input error: no support for filtering on unrecognized field \"finished_at\""}]} ``` ### Expected result This should work as this feature was merged in version `2.0.0-beta.0` and I am using a newer version. --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/8996/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/8994

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8994/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8994/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8994/events

https://github.com/kubeflow/pipelines/issues/8994

I_kwDOB-71UM5g8afg

[backend] Unable to Build APIServer image locally

closed

CONTRIBUTOR

### Environment Branch used `2.0.0-alpha.7` * How did you deploy Kubeflow Pipelines (KFP)? AWS Kubeflow Distro v1.6.1 k8s v1.22 * KFP version: 1.6.1 ### Steps to reproduce I was following this to build the image: https://github.com/kubeflow/pipelines/tree/master/backend#building-apiserver-image-locally I have moved my sample to the `/sample` folder from https://github.com/kubeflow/pipelines/blob/master/components/aws/sagemaker/TrainingJob/samples/mnist-kmeans-training/mnist-kmeans-training-pipeline.py error message: ``` docker build -f backend/Dockerfile . --tag $API_SERVER_IMAGE [+] Building 186.1s (22/36) => [internal] load build definition from Dockerfile 0.0s => => transferring dockerfile: 3.84kB 0.0s => [internal] load .dockerignore 0.0s => => transferring context: 34B 0.0s => [internal] load metadata for docker.io/library/debian:stable 1.2s => [internal] load metadata for docker.io/library/golang:1.17.6-stretch 1.3s => [internal] load metadata for docker.io/library/python:3.7 1.3s => [auth] library/debian:pull token for registry-1.docker.io 0.0s => [auth] library/python:pull token for registry-1.docker.io 0.0s => [auth] library/golang:pull token for registry-1.docker.io 0.0s => [compiler 1/10] FROM docker.io/library/python:3.7@sha256:f3887bc228fb58b0883e945d9af4bf8bac8fba34481ebfb9b3fee6d569e4f97c 0.0s => [stage-2 1/10] FROM docker.io/library/debian:stable@sha256:c1c4bb9bb5914a9a220a9d130966922b177ab552efdc019d7372936c32cc1e2e 0.0s => [builder 1/8] FROM docker.io/library/golang:1.17.6-stretch@sha256:83a2dd5267864cb4c15a78f8722a6bc0d401d4baec358695f5dc5dd1067d15d1 0.0s => [internal] load build context 0.8s => => transferring context: 9.50MB 0.8s => CACHED [stage-2 2/10] WORKDIR /bin 0.0s => CACHED [stage-2 3/10] COPY backend/src/apiserver/config/ /config 0.0s => CACHED [compiler 2/10] RUN apt-get update -y && apt-get install --no-install-recommends -y -q default-jdk python3-setuptools python3-dev jq 0.0s => CACHED [compiler 3/10] RUN wget https://bootstrap.pypa.io/get-pip.py && python3 get-pip.py 0.0s => CACHED [compiler 4/10] COPY backend/requirements.txt . 0.0s => CANCELED [compiler 5/10] RUN python3 -m pip install -r requirements.txt --no-cache-dir 183.9s => CACHED [builder 2/8] RUN apt-get update && apt-get install -y cmake clang musl-dev openssl 0.0s => CACHED [builder 3/8] WORKDIR /go/src/github.com/kubeflow/pipelines 0.0s => [builder 4/8] COPY . . 0.9s => ERROR [builder 5/8] RUN GO111MODULE=on go build -o /bin/apiserver backend/src/apiserver/*.go 180.9s ------ > [builder 5/8] RUN GO111MODULE=on go build -o /bin/apiserver backend/src/apiserver/*.go: #17 90.31 go: github.com/Masterminds/squirrel@v0.0.0-20190107164353-fa735ea14f09: Get "https://proxy.golang.org/github.com/%21masterminds/squirrel/@v/v0.0.0-20190107164353-fa735ea14f09.mod": dial tcp: lookup proxy.golang.org on 192.168.65.5:53: read udp 172.17.0.3:49039->192.168.65.5:53: i/o timeout #17 90.31 go: downloading github.com/cenkalti/backoff v2.2.1+incompatible #17 90.31 go: downloading github.com/fsnotify/fsnotify v1.5.1 #17 90.31 go: downloading github.com/golang/glog v1.0.0 #17 90.32 go: downloading github.com/gorilla/mux v1.8.0 #17 90.32 go: downloading github.com/grpc-ecosystem/grpc-gateway v1.16.0 #17 90.32 go: downloading github.com/jinzhu/gorm v1.9.1 #17 100.3 go: downloading github.com/minio/minio-go/v6 v6.0.57 #17 100.3 go: downloading github.com/prometheus/client_golang v1.12.1 #17 100.3 go: downloading google.golang.org/grpc v1.44.0 #17 100.3 go: downloading github.com/spf13/viper v1.10.1 #17 100.4 go: downloading github.com/pkg/errors v0.9.1 #17 110.3 go: downloading k8s.io/api v0.23.3 #17 110.3 go: downloading github.com/argoproj/argo-workflows/v3 v3.3.8 #17 110.3 go: downloading k8s.io/apimachinery v0.23.3 #17 120.3 go: downloading github.com/go-sql-driver/mysql v1.6.0 #17 120.3 go: downloading k8s.io/client-go v0.23.3 #17 130.3 go: downloading github.com/ghodss/yaml v1.0.1-0.20190212211648-25d852aebe32 #17 130.3 go: downloading github.com/golang/protobuf v1.5.2 #17 130.3 go: downloading github.com/kubeflow/pipelines/api v0.0.0-20220311022801-11635101d944 #17 130.3 go: downloading github.com/robfig/cron v1.2.0 #17 140.3 go: downloading github.com/stretchr/testify v1.7.0 #17 140.3 go: downloading google.golang.org/protobuf v1.27.1 #17 140.3 go: downloading google.golang.org/genproto v0.0.0-20220310185008-1973136f34c6 #17 150.3 go: downloading github.com/go-openapi/runtime v0.21.1 #17 150.3 go: downloading github.com/go-openapi/strfmt v0.21.1 #17 150.3 go: downloading github.com/Masterminds/squirrel v0.0.0-20190107164353-fa735ea14f09 #17 150.3 go: downloading github.com/google/uuid v1.3.0 #17 150.3 go: downloading github.com/sirupsen/logrus v1.8.1 #17 150.3 go: downloading github.com/lestrrat-go/strftime v1.0.4 #17 160.3 go: downloading github.com/mattn/go-sqlite3 v1.9.0 #17 160.3 go: downloading github.com/VividCortex/mysqlerr v0.0.0-20170204212430-6c6b55f8796f #17 160.4 go: downloading k8s.io/kubernetes v1.11.1 #17 160.4 go: downloading github.com/kubeflow/pipelines/third_party/ml-metadata v0.0.0-20220118175555-e78ed557ddcb #17 160.4 go: downloading gocloud.dev v0.22.0 #17 160.4 go: downloading github.com/grpc-ecosystem/go-grpc-middleware v1.3.0 #17 170.3 go: downloading gopkg.in/yaml.v2 v2.4.0 #17 170.3 go: downloading github.com/aws/aws-sdk-go v1.42.50 #17 180.3 go: github.com/Masterminds/squirrel@v0.0.0-20190107164353-fa735ea14f09: Get "https://proxy.golang.org/github.com/%21masterminds/squirrel/@v/v0.0.0-20190107164353-fa735ea14f09.mod": dial tcp: lookup proxy.golang.org on 192.168.65.5:53: read udp 172.17.0.3:49039->192.168.65.5:53: i/o timeout ------ executor failed running [/bin/sh -c GO111MODULE=on go build -o /bin/apiserver backend/src/apiserver/*.go]: exit code: 1 ``` <!-- Specify how to reproduce the problem. This may include information such as: a description of the process, code snippets, log output, or screenshots. --> ### Expected result image get built

https://api.github.com/repos/kubeflow/pipelines/issues/8994/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/8993

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8993/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8993/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8993/events

https://github.com/kubeflow/pipelines/issues/8993

I_kwDOB-71UM5g7pvP

[bug] Error parsing yaml from GCS when using `load_component_from_url`

open

NONE

### Environment <!-- Please fill in those that seem relevant. --> * How do you deploy Kubeflow Pipelines (KFP)? * On vertex AI pipelines <!-- For more information, see an overview of KFP installation options: https://www.kubeflow.org/docs/pipelines/installation/overview/. --> * KFP version: 1.8.19 <!-- Specify the version of Kubeflow Pipelines that you are using. The version number appears in the left side navigation of user interface. To find the version number, See version number shows on bottom of KFP UI left sidenav. --> * KFP SDK version: SDK V2 <!-- Specify the output of the following shell command: $pip list | grep kfp --> ### Steps to reproduce <!-- Specify how to reproduce the problem. This may include information such as: a description of the process, code snippets, log output, or screenshots. --> - upload a component.yaml file in a gcs bucket - Try to load that component using url in the `load_from_component_url` function ### Expected result <!-- What should the correct behavior be? --> ### Materials and reference <!-- Help us debug this issue by providing resources such as: sample code, background context, or links to references. --> ### Labels <!-- Please include labels below by uncommenting them to help us better triage issues --> <!-- /area frontend --> <!-- /area backend --> <!-- /area sdk --> <!-- /area testing --> <!-- /area samples --> <!-- /area components --> --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/8993/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/8991

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8991/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8991/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8991/events

https://github.com/kubeflow/pipelines/issues/8991

I_kwDOB-71UM5g5JX5

CustomTrainingJobOp: Failed to parse the container spec json payload to requested prototype.

closed

NONE

### Environment <!-- Please fill in those that seem relevant. --> * How do you deploy Kubeflow Pipelines (KFP)? I am using Vertex AI with the v2 python SDK (kfp==2.0.0b12), together with the `google-cloud-pipeline-components==1.0.33` library, compile the pipeline and run it with aiplatform.run(). * KFP version: v2 * KFP SDK version: 2.0.0b12 ### Steps to reproduce I am running a simple pipeline with only one component from the `google-cloud-pipeline-components`: google_cloud_pipeline_components.experimental.custom_job.CustomTrainingJobOp to start a vertex ai custom training job. I am using the `containerSpec` option for the Job, which links to a container in gcr (google container registry). This worked without any problems until last week and no changes have been made to the code (it was just running in production). At the moment, the pipeline component fails immediately with the error `Failed to parse the container spec json payload to requested prototype. Payload: {...}`. I already double checked the formatting of the specs that are passed with the CustomTrainingJobOp and they match the google documentation https://cloud.google.com/vertex-ai/docs/reference/rest/v1/CustomJobSpec#ContainerSpec. ### Expected result The pipeline component should start a training job on vertex training. ### Materials and reference ### Labels <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/8991/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/8990

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8990/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8990/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8990/events

https://github.com/kubeflow/pipelines/issues/8990

I_kwDOB-71UM5g4mQD

[backend] ParallelFor does not work on beta.0 - multiple issues

open

CONTRIBUTOR

### Environment * How did you deploy Kubeflow Pipelines (KFP)? Manifest installation replacing alpha.6 with beta.0 * KFP version: 2.0.0-beta.0 * KFP SDK version: kfp 2.0.0b13 kfp-pipeline-spec 0.2.0 kfp-server-api 2.0.0a6 ### Steps to reproduce Simple pipeline defined like this: ```python @dsl.component def addition_component(num1: int, num2: int) -> int: print('hello from component') return num1 + num2 @dsl.pipeline(name='addition-pipeline') def my_pipeline(a: int, b: int): add_task_1 = addition_component(num1=a, num2=b) with dsl.ParallelFor(items=[1, 2], parallelism=2) as item: add_task_2 = addition_component(num1=add_task_1.output, num2=item) ``` When running the pipeline, getting the following error: ``` Failed to create a new v1beta1 run: InternalServerError: Failed to validate workflow for (): spec.templates[5].name: 'comp-tasksgrouptype-for-loop-2-TasksGroupType.FOR-LOOP-2' is invalid: name must consist of alpha-numeric characters or '-', and must start with an alpha-numeric character (e.g. My-name1-2, 123-NAME) ``` As a workaround, I can only compile the pipeline, change the `TasksGroupType.FOR-LOOP-2` to `TasksGroupType-FOR-LOOP-2` in yaml and upload the pipeline. However when I run it, I get other error from kfp-driver: ``` F0315 08:09:30.260370 19 main.go:74] KFP driver: failed to unmarshal component spec, error: unknown field "iteratorPolicy" in ml_pipelines.PipelineTaskSpec ``` As a workaround, I can remove the parallelism - respectively the `iteratorPolicy` from compiled pipeline yaml (meaning all will run at once), upload and run again, the pipeline run successfully. However, my other pipeline, where the tasks in parallel context are getting the input not as parameters, but as an artifact, like this: ```python @dsl.pipeline(name='train-pipeline') def train_pipeline(filename: str): data_reader_task = read_data(filename=filename) with dsl.ParallelFor( items=list(range(8, 18)), parallelism=2 ) as bucket: task = train_bucket(data=data_reader_task.outputs['output'], bucket=bucket) ``` there is another error in kfp-driver: ``` F0315 12:02:29.585512 18 main.go:74] KFP driver: driver.Container(pipelineName=pipeline/train-pipeline, runID=39e035a8-baac-4a4c-a5cf-65360d66665f, task="train_bucket", component="comp-train-bucket", dagExecutionID=115, componentSpec) failed: failed to resolve inputs: failed to resolve input artifact data with spec component_input_artifact:"pipelinechannel--read-data-output": component input artifact not implemented yet ``` Does that mean that for now tasks in parallel context cannot get artifacts from other tasks, only parameters? Is there a plan to fix this? ### Expected result Pipeline with ParallelFor should run out of the box, respecting parallelism parameter and tasks in parallel context should be able to accept artifacts. --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/8990/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/8989

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8989/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8989/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8989/events

https://github.com/kubeflow/pipelines/issues/8989

I_kwDOB-71UM5g16Q_

[backend] Minio failed to put file

closed

NONE

I am getting this error : ``` This step is in Error state with this message: Error (exit code 1): failed to put file: Get "http://minio-service.kubeflow:9000/mlpipeline/?location=": dial tcp 100.68.6.113:9000: i/o timeout ``` Restarting the Kfp pods under kubeflow namespace -> did not work [ same error ] i also check volume mounts | Everything is okay from that side. Hence i am confused now with next steps. Need your help regarding the same.

https://api.github.com/repos/kubeflow/pipelines/issues/8989/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/8987

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8987/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8987/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8987/events

https://github.com/kubeflow/pipelines/issues/8987

I_kwDOB-71UM5gzcFW

Execution list pagination

closed

COLLABORATOR

https://api.github.com/repos/kubeflow/pipelines/issues/8987/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/8986

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8986/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8986/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8986/events

https://github.com/kubeflow/pipelines/issues/8986

I_kwDOB-71UM5gzcDY

Artifact list pagination

closed

COLLABORATOR

https://api.github.com/repos/kubeflow/pipelines/issues/8986/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/8985

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8985/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8985/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8985/events

https://github.com/kubeflow/pipelines/issues/8985

I_kwDOB-71UM5gwAlw

[BUG] ml-pipeline pod CrashLoopBackOff: Config DBConfig.ExtraParams not specified, skipping

open

NONE

### Environment * How did you deploy Kubeflow Pipelines (KFP)? => Using Kustomize method <!-- For more information, see an overview of KFP installation options: https://www.kubeflow.org/docs/pipelines/installation/overview/. --> * KFP version: <!-- Specify the version of Kubeflow Pipelines that you are using. The version number appears in the left side navigation of user interface. To find the version number, See version number shows on bottom of KFP UI left sidenav. --> => Kubeflow 1.6.1 * KFP SDK version: <!-- Specify the output of the following shell command: $pip list | grep kfp --> ### Steps to reproduce <!-- Specify how to reproduce the problem. This may include information such as: a description of the process, code snippets, log output, or screenshots. --> The only pods that have health issue are ml-pipeline & ml-pipeline-persistenceagent when I did deploy kubeflow on my kubernetes cluster 1.25 Here's the logs f **ml-pipeline**  For **ml-pipeline-persistenceagent**  And by looking at different issues and solutions, I tried to investigate the MySQL pod since these pods are related to storage and I did find the communication error between the ml-pipeline pods and mysql pod  ### Expected result <!-- What should the correct behavior be? --> ### Materials and Reference <!-- Help us debug this issue by providing resources such as: sample code, background context, or links to references. --> --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/8985/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/8977

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8977/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8977/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8977/events

https://github.com/kubeflow/pipelines/issues/8977

I_kwDOB-71UM5gnDoQ

[feature] Add enable caching option while running pipeline form kubeflow UI

open

NONE

Add enable caching option while running pipeline form kubeflow UI. Currently enable caching is not present in kubeflow UI while running pipeline.

https://api.github.com/repos/kubeflow/pipelines/issues/8977/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/8969

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8969/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8969/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8969/events

https://github.com/kubeflow/pipelines/issues/8969

I_kwDOB-71UM5geowT

[bug] ml_metadata.errors.InternalError: mysql_real_connect on KFP Pipelines GCP Marketplace Deployment

open

NONE

In the past few days, our TFX pipelines that run on, our long running KFP deployment on GCP, started to get intermittent failures related to `connecting to MLMD` in various different components during execution. Even the default TFX taxi pipeline can have a component that fails (non-deterministically, it can be `statisticsGen`, or `model training`, or any other), which usually with a few `Retry`s will work and go beyond that component, sometimes failing yet again in a downstream component, but other times successfully finishing execution. The issue seems to be from when the TFX Driver wants to publish a metadata to MLMD, either at the beginning of execution, or at the end. I'm not sure if it's a bug, or a misconfiguration that suddenly has happened due to infrastructure changes hidden to our team in the GKE cluster backing the KFP deployment, but most importantly, I'd want to know how to approach solving this problem. It seems the `grpc metadata service` in the GKE cluster has been running for around 6 months, and it's still RUNNING (not EXITED/TERMINATED/etc). Not sure what's going on and any help, links on what this piece is doing and what can be causes of failures would be super welcome, as this suddenly started impacting many nodes. ```bash Traceback (most recent call last): File "/usr/local/lib/python3.8/site-packages/ml_metadata/metadata_store/metadata_store.py", line 213, in _call_method response.CopyFrom(grpc_method(request, timeout=self._grpc_timeout_sec)) File "/usr/local/lib/python3.8/site-packages/grpc/_channel.py", line 946, in __call__ return _end_unary_response_blocking(state, call, False, None) File "/usr/local/lib/python3.8/site-packages/grpc/_channel.py", line 849, in _end_unary_response_blocking raise _InactiveRpcError(state) grpc._channel._InactiveRpcError: <_InactiveRpcError of RPC that terminated with: status = StatusCode.INTERNAL details = "mysql_real_connect failed: errno: , error: " debug_error_string = "UNKNOWN:Error received from peer ipv4:10.42.6.72:8080 {created_time:"2023-03-09T18:11:55.333836586+00:00", grpc_status:13, grpc_message:"mysql_real_connect failed: errno: , error: "}" > During handling of the above exception, another exception occurred: Traceback (most recent call last): File "/usr/local/lib/python3.8/runpy.py", line 194, in _run_module_as_main return _run_code(code, main_globals, None, File "/usr/local/lib/python3.8/runpy.py", line 87, in _run_code exec(code, run_globals) File "/usr/local/lib/python3.8/site-packages/tfx/orchestration/kubeflow/container_entrypoint.py", line 545, in <module> main(sys.argv[1:]) File "/usr/local/lib/python3.8/site-packages/tfx/orchestration/kubeflow/container_entrypoint.py", line 526, in main execution_info = component_launcher.launch() File "/usr/local/lib/python3.8/site-packages/tfx/orchestration/portable/launcher.py", line 528, in launch execution_preparation_result = self._prepare_execution() File "/usr/local/lib/python3.8/site-packages/tfx/orchestration/portable/launcher.py", line 315, in _prepare_execution execution = self._register_or_reuse_execution( File "/usr/local/lib/python3.8/site-packages/tfx/orchestration/portable/launcher.py", line 241, in _register_or_reuse_execution return _register_execution( File "/usr/local/lib/python3.8/site-packages/tfx/orchestration/kubeflow/container_entrypoint.py", line 69, in _register_execution return execution_publish_utils.register_execution( File "/usr/local/lib/python3.8/site-packages/tfx/orchestration/portable/execution_publish_utils.py", line 264, in register_execution execution = execution_lib.prepare_execution( File "/usr/local/lib/python3.8/site-packages/tfx/orchestration/portable/mlmd/execution_lib.py", line 119, in prepare_execution execution.type_id = common_utils.register_type_if_not_exist( File "/usr/local/lib/python3.8/site-packages/tfx/orchestration/portable/mlmd/common_utils.py", line 74, in register_type_if_not_exist metadata_type = get_type_handler(metadata_type.name) File "/usr/local/lib/python3.8/site-packages/ml_metadata/metadata_store/metadata_store.py", line 765, in get_execution_type self._call('GetExecutionType', request, response) File "/usr/local/lib/python3.8/site-packages/ml_metadata/metadata_store/metadata_store.py", line 188, in _call return self._call_method(method_name, request, response) File "/usr/local/lib/python3.8/site-packages/ml_metadata/metadata_store/metadata_store.py", line 218, in _call_method raise _make_exception(e.details(), e.code().value[0]) # pytype: disable=attribute-error ml_metadata.errors.InternalError: mysql_real_connect failed: errno: , error: ``` ### Environment * How do you deploy Kubeflow Pipelines (KFP)? GCP Marketplace deployment of KFP only TFX Pipelines using tfx==1.7.* * KFP version: 1.8.5 * KFP SDK version: Not relevant, this happens as part of the pipeline execution ### Steps to reproduce Not sure, as this is GKE cluster related it seems. ### Expected result Nodes should deterministically pass ### Materials and reference <!-- /area frontend --> <!-- /area backend --> <!-- /area sdk --> <!-- /area testing --> <!-- /area samples --> <!-- /area components --> --- <!-- Don't delete message below to encourage users to support your issue! --> Impacted by this bug? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/8969/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/8966

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8966/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8966/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8966/events

https://github.com/kubeflow/pipelines/issues/8966

I_kwDOB-71UM5gc-RF

feature(frontend) Support Status-IR and remove MLMD client in KFP Frontend

open

MEMBER

When status-IR will be fully implemented and ready to be consumed by the Frontend we need to: 1. Implement visualization using Status-IR 2. Remove MLMD client and migrate to calling api-server --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/8966/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/8965

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8965/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8965/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8965/events

https://github.com/kubeflow/pipelines/issues/8965

I_kwDOB-71UM5gc9ve

feature(backend) Provide interface for calling MLMD via api-server

open

MEMBER

KFP Frontend calls MLMD directly to retrieve information such as execution status, context, artifacts, etc. As status-IR provides status information and includes MLMD-related keys, we can provide an interface to retrieve the same information via api-server. --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/8965/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/8964

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8964/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8964/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8964/events

https://github.com/kubeflow/pipelines/issues/8964

I_kwDOB-71UM5gc9Cg

feature(backend) Status-IR: add MLMD client to persistence agent and support for Status-IR

open

MEMBER

KFP Frontend server currently calls MLMD to retrieve status information (as well as artifacts and other information). As status-IR includes DAG and status information, persistence agent needs to be able to call MLMD and update the status via api-server's call. --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/8964/timeline

https://api.github.com/repos/kubeflow/pipelines/issues/8961

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8961/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8961/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8961/events

https://github.com/kubeflow/pipelines/issues/8961

I_kwDOB-71UM5gc6OR

feature(backend) Status-IR: support status information for run messages in api-server

closed

MEMBER

Status information has beed added to `run.proto` in KFP v2 API (ref. #8960) We need to implement functionality to fill the information and store the data in the backend database. --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/8961/timeline

completed

https://api.github.com/repos/kubeflow/pipelines/issues/8960

https://api.github.com/repos/kubeflow/pipelines

https://api.github.com/repos/kubeflow/pipelines/issues/8960/labels{/name}

https://api.github.com/repos/kubeflow/pipelines/issues/8960/comments

https://api.github.com/repos/kubeflow/pipelines/issues/8960/events

https://github.com/kubeflow/pipelines/issues/8960

I_kwDOB-71UM5gc5qK

feature(backend) Status-IR: provide API message to store status information of a pipeline run

closed

MEMBER

KFP UI currently uses workflow and pipeline manifests of string types to obtain status of each node in the execution graph. While status information is outside of scope of the `pipelinespec` package, we need to provide a native support for this feature in KFP v2 API. --- <!-- Don't delete message below to encourage users to support your feature request! --> Love this idea? Give it a 👍.

https://api.github.com/repos/kubeflow/pipelines/issues/8960/timeline