kubeflow

First release of kubeflow v0.6

Features in Kubeflow v0.6:

• introduces a new Metadata component, along with a Metadata API and initial corresponding clients.
• introduces multi-user authentication & authorization with SSO.
• extends the Kubeflow Pipelines’ DSL to seamlessly support the use of Persistent Volumes and Volume Snapshots as distinct Kubeflow Pipelines resources.
• includes a number of new features and bug fixes within the Pipelines system, such as preemptible VMs, streamlined run creation, improved visualization of pipeline metadata, and support for default experiments.

Pickup new changes on master

The start of the 0.6 branch.

kfctl uses 0.5 (was master)

0.5.0

Some fixes

Fixes for notebook and fairing

Cherrypick fix until 4/1

With Cherrypick https://github.com/kubeflow/kubeflow/pull/2868

Pipeline version updated to: 0.1.14

First RC for 0.5

kubeflow registry fix
pipeline update
And other cherry picks

pick up pipeline fixes

latest cherry picks

remove ghost registries from 0.4 branch
Adding pytorch operator to default apply

Closed issues:

• Issue installing Kubeflow 0.3 #2128
• Cloud Endpoints Controller not working on master #2120
• Issue while deploying kubeflow on GKE using command line- killed message in Cloud Shell #2075
• Ambassador crashing in kops cluster #2074
• deploy.sh should not assume uuidgen is present #2072
• setup-minikube.sh doesn't install ksonnet if missing #2068
• New Jupyter spawner Ui doesn't allow entering a custom image #2060
• add a jobs component #2058
• Upgrade ks to 0.13.1 #2031
• Auto-scaling for Seldon serving? #2029
• Upgrade bootstrapper to use go modules #2023
• Grant kubeflow user service account CMLE permission #2012
• Seems to be an issue in kfctl.sh re existence of dir ${DEPLOYMENT_NAME} #2009
• trying to deploy a component using ks after initial deployment fails #2006
• Move components in core into a GCP specific ksonnet package #1996
• Installation woes - ./kfctl.sh: line 189: env.sh: No such file or directory #1993
• TF-Serving http-proxy error #1979
• TFServing template needs to convert numGpus from string to int #1972
• Expose Istio's grafana dashboard #1969
• Integrate pipelines into Kubeflow click to deploy #1967
• Figure out Istio installation in Kubeflow #1909
• upgrade cloud-endpoints-controller to use metacontroller/metacontroller:v0.3.0 #1824
• Jupyter image with NVIDIA Rapids #1806
• [GCP] Upgrade envoy used for JWT validation #1696
• On Premises deployment of Kubeflow fails unless ks flags are set v0.3 #1615
• Test Flake git clone fails RPC error #1178
• [Test Flake] GCP deployer script resource conflict setting IAM policy #1140
• Document Serving for PyTorch models #1117

Merged pull requests:

• Create cloud sql database for ml pipeline #2123 (IronPan)
• fixes "Cloud Endpoints Controller not working on mmaster" #2122 (kkasravi)
• Fix the modeldb ambassador route. #2117 (jlewi)
• Fix the check if directory named ${DEPLOYMENT_NAME} exists #2115 (jlewi)
• Update kubeflow components to v0.4.0 #2112 (richardsliu)
• Add yebrahim to pipeline ksonnet owner #2108 (IronPan)
• add r2d4 to approvers #2106 (r2d4)
• Replace with std.asciiUpper with supported by required version of jsonnet #2104 (Jeffwan)
• Don't grant cloudservices account IAM admin priveleges. #2101 (jlewi)
• ks init: --skip-default-registries + append ${KS_INIT_EXTRA_ARGS} #2100 (doodlesbykumbi)
• Update aws parameters in serving and dashboard #2097 (Jeffwan)
• Update private IP configuration with new GKE API #2085 (IronPan)
• Correct typo, "for ever image" should be "for every image" #2084 (suigh)
• Fix variable in setup-minikube.sh #2082 (andreyvelich)
• Owners file for jupyter ksonnet #2081 (pdmack)
• fixes 'upgrade cloud-endpoints-controller to use metacontroller/metacontroller:v0.3.0 #2080 (kkasravi)
• JH Spawner Enhancements - Fixes #2060 #2079 (ioandr)
• Add job to load pipeline samples #2071 (IronPan)
• Enforce use string type param for S3_USE_HTTPS and S3_VERIFY_SSL #2067 (Jeffwan)
• Temp workaround for RAPIDS shared lib issues #2066 (pdmack)
• Update katib components #2064 (richardsliu)
• update kf version to v0.3.4 #2063 (kunmingg)
• Fix a version number bug and add tf-batch-prediction into kubeflow registry #2061 (yixinshi)
• unify central dashboard layout #2056 (kunmingg)
• Updates for RAPIDS AI v0.4.0 image #2053 (pdmack)
• Add modeldb as a package for kubeflow #2050 (mpvartak)
• refactor monitoring metrics #2048 (kunmingg)
• upgrade central dashboard image to v0.3.4 to include pipeline ui link #2047 (kunmingg)
• Update Pipeline SDK version to v0.1.3 #2046 (IronPan)
• fixes 'Move components in core into a GCP specific ksonnet package' #2043 (kkasravi)
• Easy for code reading #2042 (wangkeqiang123)
• always convert numGpu to int #2034 (lluunn)
• Support running batch prediction by launching a Dataflow job on GCP #2026 (yixinshi)
• Give jupyter-notebook pods/log permission for fairing #2015 (r2d4)

Latest cherry-picks

Pull in latest cherry-picks.

Include pipeline release 0.1.6

• Pipelines Component