Deployment of ml-pipeline with Kubeflow pipeline version 1.0.3 with IAM role is not working.
amit043 opened this issue · 2 comments
when I run the command .
kubectl kustomize overlay/${VARIANT} > kubeflow-pipelines-aws.yaml
I got the config merge issue.
I1027 20:05:10.994651 40204 log.go:172] well-defined vars that were never replaced: mlmdDb,pipelinePath,awsIAMAnnotationKey,dbHost,dbPort,kfp-app-name,kfp-app-version,artifactRepositoryKeyPrefix,groupConcatMaxLen,kfp-namespace,mysqlStorageType,cacheDb,kfp-container-runtime-executor,mysqlPvcName,mysqlStorageSize,pipelineDb,artifactRepositoryBucket,awsIAMRole,awsRegion,kfp-artifact-bucket-name,mysqlStorageClass
Error: 80 is expected to be string
ok let me have a look
Seems to be an issue with kustomize
. I fixed it by remove the permission to patch service/ports/port
(i.e. u can't modify the ports for service resources in the downstream overlay).
It has been quite some time since I written the manifest, I am asking my friend to review the change before merging to master.
Meanwhile u can refer to my PR #19 branch.