Skip to content

Latest commit

 

History

History
2248 lines (1680 loc) · 95.3 KB

CHANGELOG.md

File metadata and controls

2248 lines (1680 loc) · 95.3 KB
Raw

Changelog

Upcoming

3.10.0

What's New

  • Introduce the new dataset v2.0.
  • Redesign the job page. Merge jobs and schedules pages.
  • Add more PrimeHub SDK/CLI commands

Notable Changes

  • From v3.10, PHFS is enabled by default
  • The original 'dataset' is renamed to 'volume'
  • The original 'Schedule' is renamed to 'Recurring Job'

Dataset 2.0

  • Add search file function in dataset and shared file (sc-22796)
  • Calculate and show the dataset size (sc-22789)
  • Create Dataset 2.0 from a directory on PHFS (sc-22422)
  • Create Dataset 2.0 from a directory on PHFS - implement new dataset uploader (sc-22784)
  • Dataset 2.0 - Refinement (sc-23053)
  • Dataset v2.0 browser/upload (sc-22787)
  • Dataset v2.0 CRUD (sc-22788)
  • Doc Site - Rename old "Datasets" to "Volumes" (sc-22665)
  • Jupyterhub - Rename old "Datasets" to "Volumes" (sc-22747)
  • PrimeHub Console - Rename old "Datasets" to "Volumes" (sc-22421)
  • Rename old "Git" to "Git Sync" (sc-22871)
  • SDK/CLI - Rename old "Datasets" to "Volumes" (sc-22666)
  • Upload Files to PHFS via SDK (sc-22600)
  • [Bugfix] Dataset 2.0 - Failed to upload file if the group name contains capital characters or underscore (sc-23352)
  • [Bugfix] Error listObjectsV2 shown in a newly created server, without PHFS enabled (sc-23273)

Merge Jobs and Schedule pages

  • Refine the "schedule naming" in the job list recurrence (sc-23211)
  • SDK/CLI - Rename old "Schedule" after the merge of jobs and schedule pages (sc-23065)
  • UI/UX refinement: Merge Jobs and Schedule pages (sc-22039)
  • [Bugfix] Schedule Jobs (instead of Recurring Jobs) shown in CE (sc-23311)

PrimeHub SDK/CLI

  • Add admin command in the available command list (sc-22500)
  • Add Admin Groups support in PrimeHub SDK/CLI (sc-21469)
  • Add Admin Images support in PrimeHub SDK/CLI (sc-21470)
  • Add InstanceTypes support in PrimeHub SDK/CLI (sc-22118)
  • Add Users support in PrimeHub SDK/CLI (sc-22111)
  • Add Models support in PrimeHub SDK/CLI (sc-22117)
  • SDK enhancements: instancetypes: able to list detailed info (sc-23015)
  • SDK enhancements: make some values optional (sc-22571)
  • SDK enhancements: query result output should be json formatted (sc-23027)

Complete MLOps workflow within a single notebook

  • increase logging in the demo cluster (everything now logged) (sc-21396)
  • [showcase][PHApp][blog post] End-to-end ML with Open Source tools (sc-20758)
  • Allow PHApps to control PrimeHub Resources (sc-21117)

UI/UX Improvement

  • [Admin][InstanceType] Add CTA button under every tab (sc-22235)
  • fix wording in create deployment tooltip (sc-22127)
  • Weird undo action in adding applications (sc-23354)
  • Cascading deletion for group related resources (sc-21284)
  • [Bugfix] Base Image URL is acting weird when multiple dropdown items contain the same image URL (sc-22691)
  • [Bugfix] Blank screen shown while logging in onto Demo (sc-22119)
  • [Bugfix] Dash to indicate unlimited in user resource at home page (sc-22353)
  • [Bugfix] Link in Group Settings -> Deployments does not link to correct group (sc-22298)
  • [Bugfix] [FE] Duplicate download buttons appear in the menu (sc-22822)
  • [Bugfix] Blank screen shown when switching between group while in group -> members page (sc-21757)
  • [Bugfix] Show the deployment usage info on the create deployment page only (sc-22523)
  • [Bugfix] The default phadmin user's name is "null null" (sc-19035)

1-Click Installation

  • Ask user to agree our EULA when using 1-click install (sc-22853)
  • Tweak layout for the app card (sc-22341)
  • Tweak UI of Notebooks Admin page (sc-22507)

Miscellaneous

  • step-by-step tutorial on how to convert an existing app into PHApp - use label studio as an example (sc-20746)
  • [Doc Site] Restructure markdown sourcefiles (sc-21947)
  • [Doc Site] Upgrade to Docusaurus v2 (sc-21949)
  • [PH Apps] Ready for GA (sc-19187)
  • [Refactoring] Refactor the graphql ce/ce app.ts (sc-22105)
  • Break example notebook by features (sc-22403)
  • Change the submodule of awesome primehub app to https (sc-22614)
  • mlflow broken caused by bodyParser (sc-22129)
  • No confirmation dialog was shown while deleting a group image (sc-22037)
  • Rewrite PrimeHub Quickstart (sc-21900)
  • Should not be able to set User quota that larger than Group quota (sc-22038)
  • Show group usage data in license info (sc-21270)
  • Update ce-metrics-fetcher (sc-23281)
  • Upgrade of Streamlit App (sc-22790)
  • implement model management related metrics (sc-20763)
  • [Bugfix] Build custom image is not enabled in CE (sc-22334)
  • [Bugfix] Cannot import PHApp template from URL (sc-23135)
  • [Bugfix] Group settings: job default timeout should be taken by the new job (sc-23180)
  • [Bugfix] Hub wouldn't be restarted when PrimeHub upgraded with primehub-configmap changed (sc-21590)
  • [Bugfix] If the MLFLow register model name contains space, it may cause deployments to mount folder error. (sc-21545)
  • [Bugfix] Invitation is not working in CE mode (sc-22332)
  • [Bugfix] Limit of Deployment is empty in Group resource (sc-22320)
  • [Bugfix] New AppCard incompatible with e2e tests (sc-23300)
  • [Bugfix] PrimeHub Airgap image tarball fix image pause:3.2 (sc-22772)
  • [Bugfix] primehub-examples doesn't contained in release image tarball (sc-22573)
  • [Bugfix] Should not show app settings in PrimeHub deploy (sc-22329)
  • [Bugfix] Sorting in dataset list is not working (sc-23319)
  • [Bugfix] SSH hostname in SSH Server Instruction does not match (sc-22340)
  • [Bugfix] Still show group image page when user doesn't have permission to access (sc-21874)
  • [Bugfix] Tool tips not correct for Image in Admin portal (sc-22301)
  • [Bugfix] Unable to display second page of deployment list (sc-22346)
  • [Bugfix][UI] The light gray Background of title row spans outside table in Dataset; Text too long (sc-23309)
  • [Bugfix] [normal user] got "Not Authorised" error while accessing job creation page (sc-23197)
  • [Bugfix] [User Portal][SysAdmin] Failed to create a new image (sc-23255)
  • [Bugfix] Cannot change items per page for graphql API response (sc-22098)
  • [Bugfix] Got blank page with error while switching to Job Monitoring tab (sc-22376)
  • [Bugfix] GraphQL Slow down when there's a large amount of jobs (running or completed) in the system (sc-22097)
  • [Bugfix] Hint string appears inside Text field of Group search in Admin portal (sc-18965)

3.9.0

What's New

  • Admin could invite users by an invitation link.

Available in CE

UI/UX Improvement

  • Weird icons in Shared files + Weird action of adding new folder (ch19095)
  • Some dialogs can't be closed by tapping other areas (ch19100)
  • Inconsistent Action Button from Table (ch19099)
  • General searching function improvement (ch19096)
  • Handle overflow situations in UI (ch19098)

PrimeHub CLI/SDK

  • Download API config file (ch20079)
  • CRUD for Datasets with PrimeHub SDK (ch21471)
  • Support App CRUD in SDK/CLI (ch21133)
  • Make the job stream follow print carriage return correctly (ch21679)
  • OAuth to SDK token UI (ch21397)
  • [Bugfix] User can't download artifact by SDK (ch21490)

Extend PHFS file browser to support rendering notebooks

  • Share/unshare a notebook in primehub (ch20617)
  • Improve the UX of in-group notebook render page (ch20651)

1-click installer

  • 1-click ask for email and send credentials directly to the user (ch20755)
  • 1-click credential email content (ch20756)
  • Opt-in to subscribe to newsletter when setting up a 1-click (ch22011)
  • Improve 1-Click Setup - Authentication Experience (ch21678)
  • increase 1-click node group down scale timeout to 48hr (ch21154)
  • Convert Standalone App traffics to PrimeHub (ch21127)
  • Easier way for 1-click's first user to invite more people into the server (ch20765)
  • 1-click install FAQ in README.md (ch20556)
  • [Bugfix] invalid parameter: redirect_uri in notebook panel (ch20826)
  • [AWS-CDK] Modify the document of Setup AWS environment for PrimeHub on doc-site (ch18733)

PrimeHub Applications

  • Extract PHApp related doc and spec to the community repo (ch20744)
  • Simplify the process of adding new apps to an existing primehub instance (ch20745)

Drop Canner Dependencies

  • Migrate to new admin protal from legacy canner obj: Remove unnecessary code and configuration (ch16899)
  • Drop Canner: Image (ch16903)
  • Drop Old Canner Image Builder (ch20425)

Contents for PrimeHub Application and ML Orchestration

  • Refine getting started doc (ch21210)
  • Update the document for storage mount points (ch19080)

Miscellaneous

  • Users can know which instanceType/image they are using to start notebook (ch21280)
  • Check storage resource before installing microk8s by PrimeHub-install (ch20607)
  • Unable to disable telemetry popup in CE (ch21162)
  • Enable sanity, regression in primehub after UI change (ch18986)
  • [admin portal] the tip of global option is not consistent (ch21232)
  • Can't display Scalar with TF 2.3, 2.4, 2.5 Tensorboard (ch20817)
  • URL is missing in the The API Token example (ch20288)
  • Prevent the license install on different editions of primehub (ch21296)
  • [Bugfix] Unable to enable SharedVolume if not change the default capacity (ch22154)
  • [Bugfix] unable to clean up GPU URL (ch20895)
  • [Bugfix] Can not enable group share volume after the group created. (ch21283)
  • [Bugfix] Request 0.5 CPU in instance type, show 0.6 in usage (ch21096)
  • [Bugfix] primehub-install version don't work on Linux (ch21168)
  • [Bugfix] Error shown when removing app from ph (ch21871)
  • [Bugfix] System settings/smtp settings should be required field and only updated the modified fields (ch21659)
  • [Bugfix] User cannot update if the user is imported from LDAP (ch22041)
  • [Bugfix] "Check off" don't display correctly after page 10th (ch17606)
  • [Bugfix] Admin Group: Failed to update the shared volume (ch21923)

EE only

Model Deployment

  • Allow user to know the limitation of model deployment (ch19510)
  • Ease effort for getting the model management (mlflow setup) to work (ch20238)

Complete MLOps workflow within a single notebook

  • Demonstrate key values of PrimeHub within a single notebook with PrimeHub SDK (ch20748)
  • Reset demo cluster every 24hr (ch21395)
  • Create Demo PrimeHub Instance for notebook showcase (ch20749)
  • Revisit Tutorial notebook in PrimeHub to utilize SDK (ch21202)

Miscellaneous

  • [Bugfix] Don't show model deployment information in group settings (ch20718)
  • [Bugfix] The "Build Custom Image" is not disabled if no registry is configured (ch19426)

3.8.2

What's New

  • [Bugfix] list_namespaced_pod hang issue.

3.8.1

What's New

  • Patched serveral bugs.

3.8.0

What's New

  • Support 1-click install PrimeHub on AWS.
  • Built-in image customize tool: Right now we can manual build image in image creating page.
    • Previously, our custom image builder is scattered around multiple pages.
    • In 3.8.0, we've consolidated the function into one single page for ease of use.
    • For users of the old custom image builder: please use the command kubectl -n hub get imagespec -o yaml > /tmp/imagespecs.yaml to export your images
  • Python SDK/CLI wrapper for PrimeHub GraphQL API.

Deprecated

  • Maintenance notebook will remove next version.

Available in CE

Extend PHFS file browser to support rendering notebooks

  • Jupyter notebook viewer (ch19097)

Drop Canner Dependencies

  • Make image, instancetype, dataset default global in the creating page (ch19533)
  • Drop Canner: Groups (ch16908)
  • Drop Canner: Instance Types (ch16901)
  • Drop Canner: Datasets (ch16902)
  • Drop Canner: Users (ch16909)
  • Drop Canner: usage report (ch16906)
  • Drop Canner: secret (ch16904)
  • Prevent to show fullscreen error if something error in admin portal page. (ch20171)
  • Secret requires to check the name is DNS-compatible (ch19790)
  • [Feature] variable basename for breadcrumbs share component (ch19710)
  • [Bugfix] Drop Canner: System Settings (ch19618)
  • [Bugfix] Default page of Admin portal changed from Groups to Image (ch20613)
  • [Bugfix] Drop Canner: Secrets (ch20013)

AWS one-click install web action

  • Use cloudfront to solve the untrusted certificate problem (ch19658)
  • Provide custom PrimeHub instanceType for AWS env (ch20215)
  • Fix JupyterHub SSH Server show the ELB domain instead of using cloudFront domain. (ch20507)
  • Allows to override the default instancetypes (ch20257)
  • Change default node pool for cpu from 0 to 1 (ch20265)
  • Make EKS CDK can eat primehub chart version by config option (ch20142)
  • Disable group quota for default group (phusers) (ch19875)
  • Support ECR as registry (ch19723)
  • Design and implement CloudFormation template for our production use of one click button (ch18670)
  • Support to customize the different cpu/gpu instance types in cf template (ch19882)
  • Enlarge the instance disk size (ch19884)
  • Investigate why the cpu request/limit raise since late jul (ch19703)
  • [AWS-CDK] Enable EFS CSI dynamic provisioning in EKS cluster, and configure it on group volume storage class (ch18728)
  • [AWS-CDK] Enable PHStore and rclone to use AWS S3 endpoint directly use IAM role instead of using account/secret key to access S3 API (ch18729)
  • [AWS-CDK] Support all the AWS region, currently we only support Tokyo region (ch18727)

PrimeHub CLI/SDK

  • CLI/SDK notebook (ch19232)
  • CLI/SDK Basic Functions (ch19227)
  • CLI/SDK print human friendly format by default (ch19455)
  • Command primehub group doesn't need to show images, instancetypes, datasets. (ch19676)
  • Show resources not found (ch19902)
  • CLI support no argument option (ch19722)
  • Python CLI and SDK config search logic (ch19231)

Polishing Apps

  • [PHApp] K8s native configurability - Init container (ch18974)
  • [PHApp] Persistence with Customize Group Volume Mount Path (ch18972)

PrimeHub App

  • PrimeHub App: Streamlit supports external package dependency (ch19646)

Build Image: Single Image Repository Support

  • Image Builder: Single Repository Support (ch19650)

UI/UX Inconsistencies - 2021Q3

  • Different Cell/Block/Card Style (ch19094)
  • Different ways of showing datasets (ch19102)

Miscellaneous

  • Allow SysAdmin to do what GroupAdmin can do (ch19375)
  • Modify example notebook to be cpu/gpu agnostic (ch20357)
  • Label-studio app: use fixed login credentials (ch19946)
  • Email Collector for self-served user (ch20220)
  • Global dataset cannot be found in spawner/job page (ch20019)
  • Make the type of secret to be read-only in edit mode (ch19519)
  • Fix cudnn version to match cuda 11 (ch20193)
  • Clarify the Kubernetes support version as PrimeHub v3.6+ (ch18968)
  • [Bugfix] PrimeHub Deploy: There should be no "Shared Volume" in the group admin setting (ch19577)
  • [Bugfix] primehub-aws-cdk one click can not create primehub ce cluster (ch20188)
  • [Bugfix] the layout of monitoring tab is broken (ch19927)
  • [Bugfix] Incorrect display of license usage when max limit is -1 (ch19870)
  • [Bugfix] System setting "Default User Volume Capacity" change issue (ch19832)
  • [Bugfix] In PrimeHub Deploy, the admin portal should not show image, image buidler, datasets, admin notebook. (ch19580)
  • [Bugfix] PrimeHub CE: When i click admin > systems, it will show error message (ch19584)
  • [Bugfix] Miss Intl module in cms (ch19331)

EE Only

PrimeHub CLI/SDK

  • Enhance the download behavior for files and job-artifact (ch19672)
  • CLI/SDK for Deployments (ch19230)
  • CLI/SDK Job artifact and files (ch19229)
  • CLI/SDK for Jobs/Schedules (ch19228)

Submit Notebook as Job

  • [Bugfix] Notebook JSON is invalid: Additional properties are not allowed ('id' was unexpected) (ch19515)

PrimeHub Install

  • [Bugfix] primehub-install script failed to process PRIMEHUB_STORAGE_CLASS when user's customer don't have default storage class. (ch20737)

Model Deployment

  • Enhancement of "model deployment # limit per group" (ch19183)
  • the update action can bypass the model deployment maximum constraint (ch19901)
  • [Bugfix] Server error while deleting a deployment (ch19855)
  • [Bugfix] Failed to update deployment when the number of deployments is equal to maximum deployment limitation (ch20480)
  • [Bugfix] PriemHub EE unlicensed can support to create 1 deployment (ch19235)

Miscellaneous

  • Don't show “GraphQL error: ….” when creating group in EE-Trial (ch20255)
  • [Bugfix] the 'maximum deployment' should not include the stopped deployment (ch19290)
  • [Bugfix] Primehub-install version don't show alpha release version (ch20648)

3.7.2

What's New

  • Fix cudnn version to match cuda 11.
  • Fix JupyterHub ssh server domain display.
  • Allows to override the default instancetypes in 1 click installation.

3.7.1

What's New

  • ImageBuilder support single repo.
  • Several bugfix.
  • Support 1-Click install.

3.7.0

What's New

  • Add in-notebook tutorial for getting started

Breaking Changes

Upgrade helm version from 3.3.4 to 3.6.2 Upgrade helm-diff version to 3.1.3 Upgrade helmfile version to 0.139.9

3.6.2

What's New

  • Bugfix: patch airgap image issue

3.6.1

What's New

  • UI Bugfix
  • Update Grafarna

3.6.0

What's New

  • Model Management(Beta): Users can manage and version their models. And the versioned models can be deployed in the model deployment.
  • In-App Product Guide: Add action tooltip in portal pages.

Available in CE

Pluggability (PrimeHub App)

  • [Enhancement] The PrimeHub App list should not sort by app-id (ch16854)
  • PhApplication design document (ch15956)
  • PhApplication tutorial documentation (by mlflow) (ch15955)
  • PhApplication tutorial documentation (label-studio): use the labeled results to re-train a model (ch17329)
  • Document about how to customize a PrimeHub App (ch17390)
  • Do NOT install ph-app-templates in primehub-deploy (ch17457)
  • PhApplication tutorial documentation (label-studio) (ch17324)
  • Add label-studio phapptemplate (ch17323)
  • PrimeHub usage should have phapplication usage data (ch15954)

Consolidate install-helper and install script

  • Update multi-node installation path (ch17451)
  • Update single-node installation path (ch17428)
  • Verify if all the dev-cluster-bootstrap work (ch17427)
  • Consolidate install-helper and install script (Implementation) (ch17426)

Next-generation Multi-node Solution

  • PrimeHub Installation and Configuration (ch16011)
  • Multi-node Solution for airgap (ch15998)

EE Only

Model Management

  • GraphQL for models and model versions (ch16761)
  • Inject envs in jupyter and jobs (ch16762)
  • Add model management related configuration (ch16758)
  • Change the routing and menuitem for model deployment (ch16766)
  • MLflow: register model: update custom pre-packaged server (ch17517)
  • Model management tutorial (ch16780)
  • The parsed source run info should be sorted as MLflow (alphabetical order) (ch17539)
  • Model Management: add in-app product guide (ch16779)
  • Support mlflow model in pytorch pre-packaged server (ch16769)
  • Model management configuration document (ch16781)
  • Model Managment FInal check (ch16772)
  • Model description always shows "Invalid date" (ch17525)
  • Support mlflow model in tensorflow pre-packaged server (ch16768)
  • Implement the mlflow model downloader in model deployment (ch16771)
  • On-demand Run info of model versions (ch17338)
  • Add mlflow in docker-stacks (ch16767)
  • Deploy model in model management UI (ch16765)
  • Model Management UI (ch16764)

Miscellaneous

  • shared files - the pagination is gone while switching view option (ch16752)
  • Phapplication deployment will create two replicaset (ch17267)
  • JupyterHub - Disable Timeout When Pulling Image (ch16839)
  • UI/UX enhancement: Label/capital case/minor changes (ch16829)
  • Tutorial: User can compare ML model outcome in multiple experiments/runs (ch17020)
  • the “expand” right-arrow icon of Event log is disappeared (ch17465)
  • Refine front-end code infrastructure (ch17694)
  • Disable telemetry in CI (ch17642)
  • add e2e tests of resource validation (ch15924)
  • Only save and output valid data from the mlflow-artifact-envs attribute (ch17473)
  • Fix primehub-admission security update (ch15042)
  • Users can’t scale GPU instances in demo clusters (ch17540)
  • phJob cannot use GPU (ch17567)
  • binder-based IDE with repo2docker: add doc of Spyder and ROS (ch17230)
  • Remove user portal config and images to free up space (ch17480)
  • Click on "Deploy Model" shortcut in "Home" under "Open" doesn't open Deployments page (ch17425)

Bugfix

  • [Bugfix[ Add missing pagination to Notebooks Admin
  • [Bugfix] When user uploaded a large file in Shared Files, it doesn't appear. (ch17097)
  • [Bugfix] Image builder fail to skip tls verify (ch17236)
  • [Bugfix] The group resource dashboard is not correct in certain scenarios (ch16849)
  • [Bugfix] Should not allows to inject non-relevant commands in build image package (ch16075)
  • [Bugfix] Group setting typo (ch17889)
  • [Bugfix] Pytorch 1.7 with CUDA 11 image doesn't work because of wrong Pytorch version (ch17391)
  • [Bugfix] Locale changed after logging out, reproduced with automation on Demo A (ch17723)
  • [Bugfix] JupyterHub Admin page has no pagination (ch17510)
  • [Bugfix] GraphQL error shown when switching Global flag On of images(pytorch-1.7.0, tf-1.15.4, tf-2.4.1) (ch17622)
  • [Bugfix] Shared file upload ui is broken in primehub-ce (ch17469)
  • [Bugfix] Cannot delete a image builder item (ch17372)
  • [Bugfix] phapp Env is not set correctly with env dataset (ch16973)
  • [Bugfix] Primehub app does not mount phfs in primehub ce (ch17471)

3.5.3

What's New

  • Bugfix: Patch for missing grafana link in admin portal on deploy version.

3.5.2

What's New

  • Bugfix: Fix the shared file uploader broken in primehub-ce
  • Bugfix: Fix GPU resources setting

3.5.1

What's New

  • Patch PrimeHub Deploy assets config

3.5.0

What's New

  • PrimeHub Apps(Alpha): Users can visit the PrimeHub Apps to start 3rd-party applications in PrimeHub.
  • Group Settings UX Refinement: Group admins can now access detailed group information and adjust per-group job setting on user portal.

Action Required

  • PrimeHub Apps(Alpha): [Optional] Setup PRIMEHUB_FEATURE_APP flag or app.enabled: true in the value file to enable PrimeHub App feature.

Avaliable in CE

Group Dashboard - currently available data

  • implement UI of new landing page (ch16087)

Reuse installed package in jobs or notebook sessions

  • Provide environment document in primehub site (ch15692)
  • Update the customize job runtime document (ch15693)
  • Source primehub profile files in job (ch15689)
  • Source primehub profile files in jupterhub (ch15685)

EE Only

Pluggability (PrimeHub App)

  • Support rewrite option for code-server (ch16532)
  • Add CRDs and MLFlow app templates to primehub chart (ch15922)
  • [Console] Implement proxy to primehub app (ch15910)
  • [Controller] Implement the phapplication controller - abnormal cases (ch15951)
  • PrimeHub APP UI (without log) (ch15913)
  • [Console] Implement proxy to a dummy service (ch15826)
  • [GraphQL] PrimehHub App GraphQL Implementation (ch15912)
  • [Bug] Front-end env tooltip doesn't show up (ch16858)
  • Final check for alpha feature (ch15949)
  • [Bug] Cannot create Primehub App if the group volume is not enabled (ch16847)
  • Discovery: Pluggability ecosystem/showcase (ch16212)
  • Make the PRIMEHUB_APP_ROOT with attribute 777 for all app container image (ch16669)
  • [Bug] PRIMEHUB_APP_ROOT is not set correctly (ch16737)
  • [Bug] PHFS and dataset is not mounted correctly (ch16655)
  • [Bug] The update app scope is not working (ch16654)
  • NetPol for jupyter pods: egress allows to primehub app pods (ch15908)
  • [Controller] Implement the phapplication controller - happy path (ch15919)
  • Define the PrimeHub App CRD and MLFlow example (ch15915)

Group settings UX refinement

  • implement UI of group settings page: info and members tab (ch16089)
  • group settings: final check (ch16092)
  • Tweak group settings page in user/admin portal (ch16410)

Replace S2I with Dockerfile in the model deployment

  • update primehub-site to replace S2I with Dockerfile (ch15840)
  • model deployment can't show the logs while requesting the correct data (ch16252)
  • update model-deployment-examples repo to replace S2I with Dockerfile (ch15858)
  • update primehub-seldon-servers repo to replace S2I with Dockerfile (ch15857)

Next-generation Multi-node Solution

  • Allows to grafana oidc-auth from helm-based installation (ch16054)
  • Update primehub site and ensure users can install primehub ee (ch16009)
  • Test and document integrate helm install prometheus-operator (ch15997)

Solution of PrimeHub Operation

  • [Installer] Provide instructions to use install script to install PrimeHub to a primehub-ready kubernetes (ch14932)

Miscellaneous

  • [Patch] CVE-2021-21362 MinIO (ch16297)
  • binder-based IDE with repo2docker (ch16067)
  • Fix primehub chart image CRD (ch16301)
  • [Bug] PhApplication doesn't mount global dataset (ch16623)
  • [Bug] user can't see artifacts in online demo environments (ch15307)
  • [Bug] phapplication run-app script does not quote properly (ch16494)
  • Remove legacy billing annotation from jupyter-profiles (ch16612)
  • Add graphql primehub store related integration test (ch13418)
  • [CSE] Virtualbox image for PrimeHub trial with Packer (ch16113)
  • KC_PASSWORD and PH_PASSWORD should have format checking or hint for digital input only (ch16807)
  • [Bug] TensorFlow 2.4 image doesn't work because of wrong cuda version (ch16290)
  • Refinement of "Run an Example" in deployment detail page (ch15515)
  • [Bug] PrimeHub graphql crash when access the job artifacts tab when the job doesn't enable artifacts (ch16379)
  • [Bug] The log request is disconnected if the log has no message more than 1 min (ch11588)
  • [Bug] docker-stack doesn't build tensorflow v2.4.1 GPU image (ch16229)
  • Remove "beta" from model deployment (ch16065)
  • [Bug] Cert-manager OOM (ch15827)

3.4.2

Bugfix

  • Patch tls verify in custom image building part.

3.4.1

Bugfix

  • Fix minio client

3.4.0

What's New

  • Group Admin Image Management: Group admins now can create custom group images.
  • Shared Files: Browse: Users can visit the Shared Files tab in PrimeHub then browse through, view, and download your groups' project files in one place.

Action Required

Avaliable in CE

Shared Space: Browse

  • Implement the PHFS browser (ch14726)
  • Make the phfs files endpoint support both view and download behavior (ch14725)
  • Implement the phfs browse API (ch14724)
  • Shared file Browse: Final Check (ch15755)
  • shared files menuitem issues (ch16064)
  • PHFS Browser followups (ch15550)
  • [Bug] Deleting file name or folder with chinese name would fail (ch15561)

Group Admin: Image Management

  • Allow group admins to build group images with custom APT, conda and pip packages.
  • [GraphQL] Support to create image with custom image type (ch14896)
  • [Controller] Implement the image controller (ch14894)
  • Document how to configure image builder (ch14899)
  • Revisit the configuration of image builder (ch16104)
  • Allows to fill in the existing image as base image in the custom build image (ch15810)
  • Build custom group image final check (ch14906)
  • [Bug] As a group admin, i cannot select the pull secret checkbox (ch15984)
  • make package settings are same as image builder (ch16007)
  • Implement the image job log API (ch15467)
  • Add primehub controller to ce in primehub chart (ch15813)
  • [Console] supports to create a custom image. (ch14897)
  • [Console] Handle not ready image in image list and spawner page. (ch14898)

Reuse package in jobs

  • User can source primehub profile in jobs and jupyterhub.

Solution of K8S Infrastructure

  • [Bug] Cannot create groupVolume and dataset by other storage class rather than "" (ch15208)
  • [logging] Provide external logging solution (ch14925)

Architecuture Backlog

  • Remove the admission job (ch14442)

DX Misc

  • e2e test to access every single page (ch15224)

Other

  • [Bug] PHFS Browser is not working in demo.a (ch15630)
  • image creating failed, make ci broken (ch15638)
  • z2jh 0.10.6 doesn't work on demo.a (ch15540)
  • Upgrade z2jh to support k8s 1.17 (ch11091)
  • [Discovery] primehub model deploy UX (ch15220)
  • Failed to install rook-operator while provisioning a ec2-rke cluster (ch15385)
  • Fix eks autoscaler failed (ch15351)
  • e2e tests to cover 'group admin: image management' (ch14661)
  • Remove unused static page from primehub-console. (ch14443)
  • Update notebook spawning timeouts (ch14718)
  • [Bug] job artifacts can not be accessed from PrimeHub console on demo.a (ch14888)
  • e2e enhancement: reorganize output artifacts (ch14976)
  • [Model Deployment] TensorFlow model hangs when .predict is called (ch14774)
  • CE DEVELOP.md doc & flow (ch14570)
  • [Bug] User and group quota limit doesn't work (ch15892)
  • Implement grafana dashboard and update nvidia-gpu-exporter charts (ch16077)
  • [Chore] Polish dev cluster setup script and document (ch15988)
  • Prevent timeout for model deployment prediction endpoint (ch14805)
  • k8s 1.20 compat check (ch15348)
  • [Bug] primehub-watcher default memory limit is too low (ch15210)
  • [Bug] PrimeHub-controller will throw error when imageSepcJob create pod (ch15657)
  • Review/Update docker-stack after v3.3 released (ch15238)
  • [Request from III] image of pytorch 1.7, python3.7, cuda10.x (ch14292)
  • [Bug] When user starts jupyter, the jupyterlab is started, but the original window shows "400 Bad Request" (ch15673)
  • update doc of tensorflow model server (ch15768)

EE only

License Constraints: Node & Deploy

  • Primehub-console modification for max_node & max_model_deploy (ch14880)
  • Primehub-controller support max_node & max_model_deploy in license (ch14867)
  • Add graphQL model deployment API validation (ch14877)
  • GraphQL License API modification (ch14872)

Solution of PrimeHub Operation

  • [MISC] Document for customer/SI to collect PrimeHub log by diagnostic tool (ch14943)
  • [Installer] Provide PrimeHub build-in diagnostic tool for collection issue log (ch14931)
  • [Installer] Single node install script can install Prometheus + Primehub dashboard (ch14930)
  • [logging] Review PrimeHub code to remove EFK from PrimeHub (ch14942)

Miscellaneous

  • Add per node GPUs allocation in Grafana dashboard (ch16077)

3.3.0

What's New

  • Notebook Log: Users can see logs for the notebook server. It is especially useful to investigate the reason why the jupyter server cannot launch successfully.
  • Group Admin Image Management: The system admin can assign a group member with group admin permission, and the group admin can manage their group images. It will improve work productivity across a group/team through managing images for a group.
  • Shared space upload: Users can upload files to PHFS from the sidebar menu Share Files.

Avaliable in CE

Notebook Logs

  • Implement the jupyter log endpoint
  • Implement the notebook log retry logic
  • Implement the notebook page

Group Admin: Image Management

  • Add group Image support in API
  • Group admin support
  • Group Admin: Image Management Final Check
  • Allow group admins to manage group images in user portal
  • Change the list items of images in spawner pages

Shared Space: Upload

  • [Final Check] Shared space upload
  • Implement the shared space upload UI
  • Add network policy for tusd
  • tusd supports to upload file to minio
  • Implements the tus compatible endpoint
  • dataset upload and phfs upload use the same tus image
  • [Bugfix] The upload group name should follow the notebook mount logic

EE only

Action Required

  • In this version, we start to collect anonymous usage behavior by default to ensure that we can deliver the best PrimeHub experience. To disable tracking, please set the value as follows.

    telemetry:
  enabled: false

  • Please delete {ReleaseName}-usage-db StatefulSet before executing helm upgrade, because ch14237 changes the value of storageClass.

[Model Deployment] Deploy by Model File

  • Model Deployment: Advanced document
  • Add pre-packaged servers documents
  • Implement the pytorch model server and document
  • Update model deployment tutorials for pre-packaged image suggestion
  • model image input field suggestion per supported pre-packaged servers
  • [Bugfix] Cannot remove model URI in model deployment update
  • [Tutorial] Deploy a Model by Pre-packaged Server (PHFS)

User Behavior Telemetry

  • Telemetry supports EE relative metrics
  • Telemetry Support

PrimeHub Usage Refactory

  • Usage DB storage class should be configurable

Miscellaneous

  • Update default image with job submit extension on demo cluster
  • Use 24hr instead of 12hr in metrics chart
  • Add cancel button when notebook is spawning
  • update image list in docs sites
  • scheduled automated test of GPU feature
  • Merge the three admission webhook pod to single pod
  • Merge metacontroller jsonnetd webhook to single pods
  • remove graphql workspace relative code
  • simplify all user portal headings/breadcrumb
  • basic e2e test to cover notebook logs feature
  • Add information icon to Monitoring timespan
  • reorganize scheduled e2e tests
  • z2jh vendor upgrade poc
  • long execution time while using installed package in PHFS
  • Provide model deployment example with pre-processing, post-processing, and third party library
  • Make Buildah support self-signed certificate
  • tensorflow2 model server enhancement: support image input
  • tensorflow2 model server: upgrade to 2.4.0
  • e2e test: deployed by model file
  • e2e test: launch tensorboard
  • [Documentation] install specific common custom software without reinstalling for jobs
  • [Bugfix] PhDeploymentController crashloopback if can not find group in response
  • [Bugfix] graphql should compare group names in the lower case
  • [Bugfix] job artifacts can not be accessed from PrimeHub console on demo.c
  • [Bugfix] InstanceTypes/Images are not shown in spawner when a group set "zero" of any type of quota
  • [Bugfix] Usage DB storage class should be configurable

3.2.2

  • [Bugfix] job artifacts can not be accessed from PrimeHub console

3.2.1

  • [Bugfix] Fix PhDeploymentController crashloopback if can not find group in response.

3.2.0

What's New

Avaliable in CE

  • Support canceling a spawning notebook.

SSH Server

  • The default user volume permission is too open for SSH on microk8s singlenode cluster (ch13675)

Miscellaneous

  • PrimeHub-graphql can configure liveness and readness timeout second by helm value (ch13382)
  • [Bugfix] Group name duplicate verification should case insensitive. (ch13672)
  • Group selector - saved the last selected group (ch12426)

EE only

  • [Feature] Support Job Resources Monitoring
  • [Feature] Support Model URI in model deployment. It allows users to deploy a model from model URI with a pre-packaged model server. It is no longer necessary to re-build a model image in order to deploy a new model.

Jobs Monitoring

  • Implement Hardware usage UI in the Job component (ch13364)
  • Implement GraphQL to output metrics (ch13362)
  • Update controller to inject job-agent at init-container (ch13360)
  • Implement metric collector agent (ch13357)
  • design doc for job monitoring (ch13432)
  • job resource monitoring final check (ch13433)
  • [job submission monitoring] switch to long interval will direct to the blank page (ch13885)

Model Deployment

  • PhDeployment controller handle model loading error case (ch13588)
  • PhDeployment controller support model uri (ch13132)
  • Prepare the tensorflow2 framework image (ch13129)
  • Deploy by model file - UI tweak (ch13997)
  • Model file final check (ch13138)
  • Reorganize the model deployment document (ch13597)
  • Design document for model file support (ch13134)
  • Support env variables for model deployment (ch13731)
  • Refactor the airgap image tarball logic (ch12996)
  • Provide a way to change the PRIMEHUB_AIRGAPPED image prefix (ch13000)
  • [Bugfix] Cannot download files of model uri in GKE (ch14009)

Job Submission Artifacts

  • Implement the artifact retention (ch12303)

PrimeHub Usage Refactory

  • Shrink the primehub-usage image size (ch13516)

Miscellaneous

  • Handle the error situation while number of jobs are increased (ch13416)
  • All pods should have cpu/memory limits - Job Artifact / primehub-store (ch13707)
  • [Bugfix] fix airgap issues in v3.1.1 (ch13832)
  • [Bugfix] model deploymet should support all gpu resource (ch13562)
  • [Bugfix] Cannot run job when artifact and phfs are not enabled (ch13811)
  • [Bugfix] Group name should transfer to lowercase when fetch artifacts info from minio (ch13665)
  • [Bugfix] Schedule job's timeout is not working (ch13584)
  • [Bugfix] failed to view artifacts (ch13745)
  • [Bugfix] Add primehub store configuration document (ch13525)
  • [Bugfix] Phfs in notebook is not editable when file is empty (ch13160)
  • [Bugfix] Failed to access GPU in job submission (ch13904)
  • Update primehub-install script to support feature of PrimeHub 3.1 (ch13386)
  • Image Error Handling (ch13328)
  • Custom software apt-install for notebooks and jobs (ch13330)
  • Allow users to search Jobs/Schedule tables by name (ch11741)
  • Primehub-store should have tolerations for GPU node (ch13249)
  • Artifact open in new tab (ch13409)
  • Cleanup completed Job (ch13340)
  • Modify to cd to the path and submit a notebook from jupyterlab (ch13396)
  • PrimeHub EE can be installed by Makefile only provide domain name (ch13564)
  • Provide the instructions to install our jupyterlab extension (ch12803)
  • Primehub Postflight Checklist (ch12622)

3.1.1

  • [Bugfix] Failed to view artifacts (ch13745)
  • [Bugfix] Missing controller part of artifact retention (ch12303)
  • [Bugfix] Fix job's timout not working issue (ch13584)
  • [Bugfix] Fix airgap issue (ch13759)
  • [Bugfix] Phfs in notebook is not editable when file is empty (ch13160)
  • [Bugfix] Group name duplicate verification should case insensitive (ch13672)
  • [Bugfix] Group name should transfer to lowercase when fetch artifacts info from minio

Upgrade Notes

  • The image version of csi-rclone image changes in this upgrade. The pods with phfs mounted need to restart after upgrading or they cannot access the phfs data.

3.1.0

What's new

Available in CE

Spawner

  • Disable JupyterHub consecutive_failure_limit due to it will auto restart hub process (ch12897)
  • [Bugfix] when user is spawning a jupyter pod, and user clicks "my server" in jupyterhub page, he will see the spawning status without header (ch12374)

Server-to-Server Connection

  • Modify CI and e2e testing to use non-nip endpoint (ch12072)
  • Update the doc for s2s connection (ch13029)
  • Admin-notebook oidc supports both ex/internal keycloak (ch12039)
  • JupyterHub oidc support both ex/internal URL (ch12037)
  • Grafana oidc supports both ex/internal URL (ch12038)
  • Review and modify primehub-console to support default use internal URL to connect keycloak for oidc (ch12109)
  • [Bug] Cannot load page group-context ui page for non-admin user in in-cluster keycloak settings (ch12915)

Group-Context Sidebar Menu

  • Remove group in list view column for job/schedule/model (ch12287)
  • [Bugfix] when admin click "access server" in jupyterhub admin in new UI, it will open jupyterlab in iframe (ch12364)

Admin Portal

  • Update Group setup ui in admin portal. (ch12940)
  • Admin portal to user portal sidebar UX enhancement (ch12330)
  • [Bugfix] Failed to create group (ch13273)

Users

  • [Bugfix] Failed to connect user to group (ch12594)
  • [Bugfix] Should not show the everyone group in the user page (ch12254)

Datasets

  • Remove GPU/CPU quota columns from create/edit dataset tables (ch12469)
  • [Bugfix] Datasets page is broken by clicking "Add" then clicking "Back" (ch12512)

Miscellaneous

  • [CE][Popularity] README revamp (ch11958)
  • [CE DX] Auto generate secrets & tokens if not provided (ch11353)

EE Only

  • Support Job Artifact: Users can output files generated from a job. The generated files can be downloaded later from the job UI.
  • Submit Job from Jupyter Notebook: User can submit job by our notebook plugin on notebook directly.
  • On-Prem Usage Reporting.

Job Submission

  • Make job timeout configurable based on group and job (ch10336)
  • Show resource information in job submission form (ch12332)
  • PhJobs should change to Failed state if group is not found. (ch12092)
  • Streamline headings/titles in Job Submission (ch10276)
  • [Bugfix] cannot clone job when job contain specific char or newline char. (ch12179)
  • [Bugfix] Submitted job by general users shows "{"code":"INTERNAL_ERROR","message":"Request failed with status code 403"}" in log (ch13189)

Job Submission Artifacts

  • Design document Job submission artifacts (ch12305)
  • Implement the artifact retention (ch12303)
  • Implement the proxy to download API (ch12614)
  • Job Submission UI: Implement Artifact Tab (ch12299)
  • Implement graphql PHFS list API (ch12292)
  • Implement artifacts copy logic (ch12294)
  • Job submission UI: Move basic information from tab to upper pane. (ch12298)
  • Implement REST download api (ch12302)

Submit Job from Jupyter Notebook

  • Publish the jupyterlab extension to NPM (ch12799)
  • Papermill: check installed and run notebook as a job (ch12911)
  • Update base image to install extension and papermill (ch12800)
  • Submit a notebook as a job in jupyterlab (ch11908)
  • Improve the error message in graphql (ch12592)
  • Add necessary environment variables in job submission controller for primehub-job package (ch11901)
  • Using the ipywidgets to show the job status (ch11910)

PrimeHub Store

  • [Bugfix] PrimeHub store settings would be missing if primehub store changes to enabled. (ch12644)

On-Prem Usage Reporting

  • On-prem usage design document (ch11426)
  • Tool to patch non usage-annotated pods (ch11823)
  • Detailed usage report (ch12665)
  • Change download file name of usage report (ch13121)

Miscellaneous

  • Update the ceph default version to 13.2.7 (ch12885)
  • Update base image's JL to 1.2.7 (ch12823)
  • Simplified EE installation using public chart (ch10393)
  • [Bugfix] The csi-nodeplugin-rclone DaemonSet need to add tolerations (ch13118)
  • [Bugfix] Applying CRD yaml creates an item with wrong ID on Keycloak (ch12994)
  • [Bugfix] A scheduled job is failed because pod already exists (ch11074)
  • [Bugfix] PrimeHub 3.0 internal keycloak chart issues (ch12435)

3.0.1

What's New

  • [Bugfix] Fix spawning page issue (ch12374)

3.0.0

What's New

Available in CE

New Primehub UI

  • Group Context: Add a global group selector on top-right corner.
  • Integrate our components in new entry portal.

SSH Server

  • Merge ssh-proxy-server chart into PrimeHub chart #ch11274
  • Install sshd should not block start-notebook.d #ch11962

Dataset

  • Deprecate the launch group only features for dataset #ch11020

Miscellaneous

  • Support default storageClass #ch11886
  • Primehub Chart includes metacontroller and keycloak #ch11128

EE Only

  • Primehub Usage(Alpha): PrimeHub Usage provides administrators a overall insight of the usage of the PrimeHub.

Primehub Usage

  • Verify usage data with stackdriver's log #ch11822
  • Implement the usage report generator #ch6613
  • Implement usage query API #ch6614
  • Add big-query importer !1055
  • Usage component docker/helm integration. #ch6616

Miscellaneous

  • Add rook rgw support !1040
  • Add legacy pod migration helper !1059
  • Add pre upgrade script for primehub 3.0 !1049
  • Remove kibana from primehub-upgrade script. #ch11710
  • Add running pods monitor !1054
  • Add pre-upgrade for v3.0, auto add KEYCLOAK_DEPLOY and METACONTROLLER_DEPLOY #ch11779
  • [Feature] ch11886 Set rook-block storageclass as default storageclass !1057

Breaking Changes

Action Required

Deprecated

  • The annotation dataset.primehub.io/launchGroupOnly in dataset is deprecated now.

2.8.1

Bugfixes

  • Fix env dataset contain invalid character.
  • Fix log component doesn't follow newest logs.

2.8.0

What's New

  • Admission Webhook now will ignore the pvcs that does not exist for users' jupyter pod and submitted job.
  • Object Store: support csi-rclone installation

Breaking Changes

  • Upgrade helm version from 2.16.1 to 3.2.4
  • Upgrade helm-diff version to 3.1.2
  • Upgrade helmfile version to 0.125.0

Action Required

Before primehub upgrade:

After primehub upgrade:

  • If helm-diff doesn't show anything but failed to apply, you might encounter this issue: helm/helm#6642 You might run into this if prometheus is installed: 
    Error: UPGRADE FAILED: cannot patch "prometheus-operator-grafana" with kind PersistentVolumeClaim: PersistentVolumeClaim "prometheus-operator-grafana" is invalid: spec: Forbidden: is immutable after creation except resources.requests for bound claims
    To solve this, you have to specify grafana.persistence.storageClassName in etc/helm_override/prometheus.yaml

Spawner

  • SSH Server - controller for jupyter (ssh) service. #ch10291
  • [Bugfix] Fix spawner when launching latest jupyter/base-notebook will go to classic notebook view #ch10914
  • [Bugfix] SSH Server may make the jupyter server not accessible #ch10786

PrimeHub Store

  • Make minio pod managed by statefulset rather than deployment #ch11177
  • Refine the installation experience for csi-rclone #ch10951
  • Refactor the minio templates in primehub chart #ch10994

Model Deployment

  • Add effectiveGroups field in the User type of graphql #ch10618
  • csi-rclone installation #ch10603
  • Model Deployment Installation - Install end-to-end check #ch9976

Admission

  • [Bugfix] admission pvc-check failed to add an annotation to user submitted jobs #ch10754

Miscellaneous

  • Check AWS env work with auto scale #ch11026
  • Refactor cluster shutdown script #ch10812
  • Ignore any pvc that does not exist #ch10235
  • Refine single node PrimeHub install script #ch10351
  • Review and update full cluster shutdown document #ch10501

v2.7.0

What's New

  • Model Deployment(Alpha): [Optional] Setup PRIMEHUB_MODE=deploy flag to install model deployment feature only.
  • Image Builder: It is general available now.
  • Model Deployment: Enable model deployment by default when it is installed in deploy mode
  • JupyterHub: Support non-standard images (not jupyter/base-notebook compatible)
  • API Auth: User can get self api token for using graphql api on API auth page now.
  • Model deployment client token.

Action Required

  • Before upgrading primehub-grafana-dashboard-basic chart, please touch $(phenv --path)/helm_override/primehub-grafana-dashboard-basic.yaml.

Job Submission

  • Job submission controller supports dataset with type nfs and hostpath #ch9630
  • [Bugfix] Fix the issue of job list is out of sync #ch10219

Model Deployment

  • Model deployment client token - controller create ingress and secret according to clients spec #ch9884
  • Move model deployment podmonitor from primehub chart to primehub-grafana-dasboard-basic chart. #ch9523
  • Model deployment - upgrade dev-gce's ingress (rke) greater than or equals to 0.25.1 #ch9083
  • Model deployment controller - check group is deployment enabled #ch8888
  • Model deployment installation - enable model deployment for groups by default #ch9994
  • Model Deployment Installation - Add CI test for model deployment only. #ch9975
  • Replace seldon engine to executor #ch9511
  • [Bugfix] Model Deployment Controller: The endpoint should use the scheme the helm value defined #ch9212

Spawner

  • Jupyter spawner support dataset with type nfs and hostpath #ch9629
  • [Bugfix] JupyterHub Spawner page Show advanced settings and Datasets button don't work #ch9576
  • [Bugfix] User can't spawn latest jupyter/base-notebook on version master@f6331bae #ch9792
  • [Bugfix] jupyterhub shows wrong GPU "-" when user limit is 0 #ch9688
  • [Bugfix]chown_extra.sh should not break notebook when fail #ch9372
  • [Bugfix] using instanceType of floating point RAM will cause jupyter spawner page to show 'No group is configured for you to launch a server' #ch9578

Image Builder

  • Modify primehub-controller logic: allow empty image builder configs #ch9264

Miscellaneous

  • Add simple CI for PrimeHub CE chart #ch9011
  • Basic checking when we prepare k8s upgrade #ch9619
  • Confirm model deployment airgap ready #ch8993
  • Create datasets test cases to cover more e2e coverage on spawner #ch9794
  • Better PrimeHub default configuration with auto script #ch9667
  • Refactor the primehub grafana dashboard basic #ch9447
  • Update CRD dataset schema for nfs, hostPath, and pv manual provisioning #ch9628
  • Update the model deployment dashboard for seldon engine 1.x #ch9282
  • Add model deployment icon to user portal #ch9522
  • [Bugfix] Hub redirect loop after switch user via primehub-console #ch9480
  • [Bugfix] failed to pull image from private repo #ch9691
  • [Bugfix] On AWS EKS cluster we can not cross zone to access pvc volume #ch8995
  • [Bugfix] dataset upload is broken in http environment #ch9110

v2.6.3

  • [Bugfix] Fix zero display issue on Spawner page.

v2.6.2

  • [Bugfix] Spawner assets upgrade for click issue.

v2.6.1

What's New

  • [Bugfix] Fix hub redirect loop after logout in portal.

v2.6.0

What's New

  • Model Deployment(Alpha): [Optional] Setup PRIMEHUB_FEATURE_MODEL_DEPLOYMENT flag to enable model deployment feature.

v2.6.0

What's New

  • Job Submission(Beta) Job Submission functions are beta now. It is enabled by default. You can disable it by setting PRIMEHUB_FEATURE_JOB_SUBMISSION flag to false.
  • Support K8S 1.17

Breaking Changes

  • Job Submission(Beta): The mount path of the working directory is changed from /workingdir to /home/jovyan. This makes the path consistent with jupyter notebook.

Job Submission

  • Change the workingdir mount path #ch9042
  • Remove job scheduler from user portal #ch8896
  • [Bugfix] Job submission (logs) timezone should refers to PrimeHub System timezone setting #ch9128

Model Deployment

  • Implementation Design and CRD definition #ch6358
  • Model deployment controller implementation (basic) #ch8574
  • Metrics Collecting #ch7067
  • Resource constraints #ch8806
  • Change underlying implementation from seldondeployment to deployment. #ch9039

Admission

  • Admission Webhook allows resource validator to apply quota check only on group #ch9029

Spawner

  • Let NFS pods can be spread to all the nodes evenly. #ch7974
  • [Bugfix] Safe mode with repo2docker might broken after jupyterlab 2.x #ch8549
  • [Bugfix] Users switching causes Jupyter page redirecting issue after #ch8513

Image builder

  • Swap the order of pip and conda install #ch8492

Miscellaneous

  • Refactor AWS dev environment #ch9044
  • Change keycloak helm chart repo to codecentric/keycloak #ch7333
  • Upgrade rook version to support k8s v1.16.x on on-premise cluster (rook v1.0.x) #ch8099

v2.5.3

What's New

  • [Bugfix] Users switching causes Jupyter page redirecting issue after #ch8513

v2.5.2

What's New

  • Fix CI build

v2.5.1

What's New

  • Fix rook upgrade issue.

v2.5.0

What's New

  • Kubernetes v1.16 supported.
  • Prometheus-Operator v8 supported.
  • New alpha feature: Scheduled job support.: [Optional] Setup PRIMEHUB_FEATURE_JOB_SUBMISSION flag to enable job scheduler feature
  • Define primehub-ready kuberentes in primehub sites and better installation experience by helm install.

Job Submission

  • Submit the scheduled Job by job Controller #ch7777
  • Schedules design document #ch7779
  • Use system timezone in scheduled job #ch7954
  • Refactor graphql setting from jobSubmission path in primehub helm chart #ch7154
  • [Bugfix] group name contains underscore will break job submission #ch7985

PrimeHub Console

  • Have an indicator on user portal and job submission page to tell users it is not GA feature #ch7479
  • Refactor admission webhook for graphql endpoint settings. #ch7325
  • Add job scheduler icon on the user portal #ch8252
  • [Bugfix] PrimeHub console landing page miss PrimeHub icon #ch8382
  • [Bugfix] Fix user cannot use floating point RAM. e.g. 1.5 G #ch7605

Spawner

  • Show more details when k8s api was gone. #ch7758
  • [Bugfix] Hub Spawner not working for admin spawning on behalf of user #ch8174

Image Builder

  • [Bugfix] image builder enable logic is wrong in dev-gce init script #ch7994

Maintenance notebook

  • Apply Istio authn/authz to admin-notebook #ch4805

Miscellaneous

  • Put our primehub version into helm app version #ch6735
  • Group provisioned resources report, and quota recommendation #ch7170
  • Modify the description of bootstrap generated group #ch8536
  • In http setup, set sslRequired to none for all keycloak realms #ch8506
  • Set USER env according to base image #ch8234
  • Replace rollme of hub deployment #ch8322
  • Dataset upload resources should not be overridden #ch8102
  • Define the process to acquire credential and configure the primehub private images. #ch7323
  • Allow installing primehub in namespace other than hub #ch6194
  • Unify dev-kind/dev-gce/dev-gcp non-helm installer parts #ch4520
  • [Bugfix] Fix installation problem in airgap #ch8477
  • [Bugfix] Fix resource validation in airgap issue. #ch8479
  • [Bugfix] primehub-group & primehub-gitsync svc selector is broken in 2.4 #ch8097
  • [Bugfix] The playground introspection is not working #ch7935
  • [Bugfix] Miss to touch metacontroller.yaml when upgrade primehub to 2.4 #ch7961
  • [Doc] Primehub Installation/configuration guide (by helm install) #ch6200
  • [Doc] Add Steps of set up license file in PrimeHub installation #ch8535
  • [Doc] Document to setup a ph-ready k8s for single node on ubuntu #ch7795
  • [Doc] Document to setup a ph-ready k8s on GKE #ch6201

Breaking Changes

  • Primehub will no longer support kubernetes v1.13.

Upgrade Notes

  • admission: Change admission ca signing process.

    To migrate old admission webhook settings, please use make primehub-upgarde to upgrade primehub.

  • console images: The environment variables PRIMEHUB_CONSOLE_DOCKER_USERNAME and PRIMEHUB_CONSOLE_DOCKER_PASSWORD are not required for production anymore. They are moved to dockerhub public repositories. But for development purpose, if these two variables are defined, primehub still use the original private repositories.

  • [Optional] setup PRIMEHUB_FEATURE_PROMETHEUS_OPERATOR_8 flag to upgrade prometheus-operator to version 8.

v2.4.1

What's New

  • [Bugfix] Replace rollme of hub deployment #ch8332
  • [Bugfix] Dataset upload resources cannot be overridden #ch8102
  • [Bugfix] primehub-group & primehub-gitsync svc selector is broken in 2.4 #ch8097
  • [Bugfix] Bug primehub console landing page miss primehub #ch8382
  • [Bugfix] Bug group name contains underscore will break #ch7985

v2.4.0

What's New

  • License Key Management(Beta) You can now obtain trial license and install PrimeHub and experience EE .
  • Make persistent home directory under user path : Make users be able to mount their persistence volume to /home/jovyan/user, by Enable Safe Mode in the spwaner advanced settings.
  • Support repo2docker image: Make images built by repo2docker be able to run on PrimeHub.
  • Upgrade default helm version from 2.11.0 to 2.16.1 - #ch3127
  • Keycloak v8 supported.
  • We rename our Custom Image feature to Image builder.

Spawner

  • "Used" metric of Group Resources on spawner should include resources used by jobs #ch6910
  • Update Safe Mode tips #ch7309
  • make primehub compatible with jupyter-repo2docker images #ch4866
  • Refactor the way to integrate jupyterhub subchart. #ch6198
  • Make persistent home directory optional in spawner #ch6765
  • [Bugfix] Exceeded quota then wait for 5 mins will make spawner page to be "403 : Forbidden" #ch6895

Jub Submission

  • UI bugs and enhancements #ch6572
  • The job duration should be computed since running stage #ch6741
  • The detail page show more info about image/instance type #ch5701
  • Change phjob phase naming: ready change to preparing #ch7134
  • [Bugfix] phjob controller make the primehub-controller crashloopbackoff #ch6581
  • [Bugfix] Controller got OOMKilled when jobs are more than ~1200 #ch6821
  • [Bugfix] Controller crashes because of nil pointer. Job in Ready phase should have starttime and in Final phase should have finished time #ch6586
  • [Bugfix] Job submission/Jupyterhub - datasets should have correct write permission #ch6143
  • [Doc] Job Submission Design doc #ch4540

License key management

  • [License MVP] Beta release of License Key Management #ch5145
  • [License MVP] Auto install pre-generated license in dev-gce & dev-gcp #ch6868
  • [License MVP] Make the jupyterhub, hub and graphql pod restarts while cm/secret changes #ch7340
  • [License MVP] Expose env variables in running job #ch6918

Image Builder

  • Add resource request/limit to custom image pod #ch6167
  • Size of ephemeral storage for image builder should be configurable #ch7680
  • Image builder feature default disabled. #ch6421
  • [Bugfix] admin can't build larger custom image #ch6412
  • [Doc] Rename Custom Image/Build Image with Image Builder #ch7051

Miscellaneous

  • [Bugfix] Grafana drilldown doesn't work after our URL rewrite #ch6248
  • [Bugfix] "Cleanup orphan PVC" in maint notebook shows non-orphan PVCs #ch7307
  • [Bugfix] job runs by root instead of jovyan #ch6589
  • [Bugfix] primehub install scripts broken #ch6651
  • [Bugfix] non-admin user can't login primehub with keycloak 8 #ch7716
  • [Bugfix] The 'Back to Portal' link is not back to user portal #ch5254
  • [Bugfix] Take gpu image when instance type is cpu #ch6504
  • [Doc] Primehub Architecture: write the architecture document #ch6106
  • [CI] remove maintenance notebook #ch4912
  • Refactor the primehub helm chart, move sub-charts to primehub chart #ch5514
  • Create group volume in primehub-console rather than hub. #ch3205
  • Bump Canner UI to 3.15.1.
  • E2e test enhancement.
  • Make primehub group nfs and git sync as part of primehub helm chart. #ch3409

Breaking Changes

  • The primehub chart value structure change. The primehub-console, primehub-prerequisites, admin-notebook is not used anymore.
  • Primehub Enterprise Edition now requires a license to unlock full features. Please install a license after upgrade.

Upgrade Notes

  • If you didn't prepare a license file, your cluster will active by default license.

  • Primehub chart value path changed. Please modify the configuration in $(bin/phenv --path)/helm_override/primehub.yaml

    Old path New path
    primehub-prerequisite.hub.startnotebook jupyterhub.primehub.startnotebook
    primehub-prerequisite.* not used
    primehub-console.keycloak primehub.keycloak
    primehub-console.keycloak.url not used. env KC_SCHEME and KC_DOMAIN instead
    primehub-console.ui.* console.*
    primehub-console.watcher.* watcher.*
    primehub-console.graphql.* graphql.*
    admin-notebook adminNotebook
    jupyterhub.custom.adminEndpoint not used.
    jupyterhub.custom.keycloakHost not used. env KC_DOMAIN instead
    jupyterhub.custom.keycloakScheme not used. env KC_SCHEME instead
    jupyterhub.custom.keycloakRealm not used. env KC_REALM instead
    jupyterhub.custom.primehubRolePrefix primehub.keycloak.rolePrefix
    jupyterhub.custom.* jupyterhub.primehub.*
    *.ingress ingress

    For ingress customization, we unify the ingress setting to ingress.* rather than <component>.ingress.* in each component.

  • groupvolume&gitsync: Change the primehub-groupvolume and primehub-gitsync installation from manually install to primehub helm chart.

    To migrate groupvolume and gitsync to primehub helm chart, please use make primehub-upgarde to upgrade primehub.

    If don't want to migrate, please add the following config into helm_override/primehub.yaml

    groupvolume:
  enabled: false
gitsync:
  enabled: false

  • groupvolume: Now, group volume (pvc and pv) is created when creating a group (with a shared volume) in admin UI. And may need to change the default storage class and group volume storage class helm settings.

    If you overwrite

    jupyterhub:
  custom:
    groupVolumeStorageClass: ""
    groupVolumeAnnotations:
      primehub-group-sc: standard

    or

    graphql:
  primehubGroupSC: standard

    Please move your overwritten settings into

    primehub:
  sharedVolumeStorageClass: ""
groupvolume:
  storageClass: standard

v2.3.0

What's New

  • Custom Image (Beta): Custom image feature is now in beta!
  • Job Submission (Alpha): Setup PRIMEHUB_FEATURE_JOB_SUBMISSION flag to enable job submission feature!
  • Document Site: Our new document site is now online!

Changes

Upgrade Notes

  • Jupyterhub: Change refresh_user call frequency from 300 seconds to every time when Jupyterhub API called. This behavior will consume more Keycloak resources. If Keycloak is unbearable in production environment, we can use jupyterhub.custom.auth_refresh_age in helm_override/primehub.yaml to configure the frequency of calling refresh_user.

Spawner

  • [UI] Spawner UI update according to window size #ch5245
  • [UI] Memory in group resource should display in "xxxGB" format - #ch5073
  • [UI] Simplify group selection dropdown - #ch5158

Custom Image

  • Put jobs in updated_at desc order - #ch5325
  • [UI] Improve Packages textarea - #ch5326
  • [UI] Add Build Images page on Admin-UI - #ch3075
  • [Doc] Update AdminUI with Build Images page - #ch4005

Job Submission

  • PHJob Controller - Terminate all pods once the phjob reaches activeDeadlineSeconds - #ch5033
  • PHJob Controller - advance spawn logic - #ch4180
  • PHJob Controller - Allow to delete pod when job complete after a given time - #ch5032
  • PHJob controller - Requeue if the reason of failed is due to admission - #ch4542
  • [Bug] PHJob controller - spawner logic bug - #ch5230
  • [Bug] Job Submission - always failed on demo a #ch5970
  • [Bug] Job page is cached so that the UI is not shown currectly. #ch5884

Bugfix

  • [Bug] Fix dev-gce use wrong image for calico controller #ch6235
  • [Bug] Fix resources-validation-webhook can not allow username content '@' on demo environment #ch5774
  • [Bug] Fix OTP message in account creation process #ch5951
  • [Bug] Fix maintenance notebook cleanup orphan pvc is broken #ch5822
  • [Bug] Fix user can't logout properly in primehub-console #ch5800
  • [Bug] Fix admin bit removal didn't sync to jupyterhub - #ch5827

Security

  • [Bug] Fix primehub-console insecure cookies - #ch5637

v2.2.2

Bugfix

  • [Bug] Resources-validation-webhook can not allow username content '@' - #ch5774
  • [Bug] Backport new Primehub theme version to 2.2 - #ch5951
  • [Bug] Move Primehub-admission image to docker hub - #ch5323

v2.2.0

What's New

  • Custom Image: Admin now can build images with custom packages which include APT, pip and conda.

  • Admission: We use admission to check resources now. Please reference admission document for more detail.

  • Dataset Upload (Beta): For a dataset with pv type, it can run a dataset upload server which allow resumable upgrade and disable upload progress.

Breaking Changes

Action Required

  • To enable Custom Image, you need to set PRIMEHUB_FEATURE_CUSTOM_IMAGE=true and it requires new environment variables:
    • PRIMEHUB_CONTROLLER_CUSTOM_IMAGE_REGISTRY_ENDPOINT: Registry endpoint (e.g. https://gcr.io)
    • PRIMEHUB_CONTROLLER_CUSTOM_IMAGE_REGISTRY_USERNAME: Registry username
    • PRIMEHUB_CONTROLLER_CUSTOM_IMAGE_REGISTRY_PASSWORD: Registry password
    • PRIMEHUB_CONTROLLER_CUSTOM_IMAGE_REPO_PREFIX: Prefix for image. (e.g. 'gcr.io/infuseai' for 'gcr.io/infuseai/primehub-controller:be20ee69c5')

Features

  • [Dataset Upload v1] Dataset Upload V1 - Beta - #ch5084
  • [Dataset Upload v1 Make dataset upload endpoint configurable in primehub console - #ch4132
  • [Admission Hook GA] Make ENABLE_ADMISSION default - #ch2476
  • [Spawner v1] Spawner style issue when server fails to start - #ch5004
  • [Spawner v1] The 'Start Notebook' button should always in the bottom - #ch4636
  • [Spawner v1] JH api endpoint and rendering for group resource usage stats - #ch3595
  • [Spawner v1] Update spawner page to apply the new design and show dataset info - #ch3598
  • [Spawner v1] Spawner Progress page redesign - #ch1550
  • [Custom Image v1] [Custom image v1] Allow to disable custom image by default - #ch4890
  • [Job Submission: MVP] PhJob Controller - implement requeue logic on ready state timeout - #ch4473
  • [Job Submission: MVP] Modify resource validation logic in jupyter and admission - #ch4466
  • [Job Submission: MVP] PHJob Controller - running pod failed reason should copy to phjob reason. - #ch4469
  • [Job Submission: MVP] PhJob Controller - helm integration - #ch4475
  • [Job Submission: MVP] PHJob Controller - implement job cancellation - #ch4472
  • [Job Submission: MVP] PhJob Controller - make working directory mount to a ephemeral storage - #ch4476
  • [Job Submission: MVP] Survey resource constraint solutions - #ch3588
  • [Job Submission: MVP] [cp1] PHJob Controller - basic workload - #ch3790
  • [Job Submission: MVP] [cp1] Provide sample PHJob definition - #ch3591
  • [Custom Image v1] Custom image v1 add feature flag to console - #ch5384
  • [Custom Image v1] Make primehub-controller to be helm managed - #ch4157
  • [Custom Image v1] Implement an controller to watch ImageSpec and create the follow-up jobs - #ch3077
  • [Custom Image v1] Implement the job log tailing API - #ch3787
  • [Workspace v1] Workspace: manage workspace and create resources in different workspace - #ch1002
  • All pods should have cpu/memory limits - #ch3882
  • Make primehub keycloak theme as the default installation. - #ch3241
  • Allow to bootstrap by service account instead of master account. - #ch4443
  • Change Primehub-keycloak-theme registry repo form gitlab to dockerhub - #ch4424
  • Disable kubernetes service links in pods - #ch416
  • Revisit monitor scripts to see if alert manager is now adequate - #ch3463
  • Gitsync daemonset should have toleration - #ch4013
  • Check 401 service account token and fix them - #ch3920

Security

Bugfix

  • [Bug] graphql might not response to client anymore - #ch4759
  • [Bug] When user logout and login again immediately, he will see error message. - #ch3045
  • [Bug] The newer JupyterLab won’t download image files - #ch4036
  • [Bug] Grafana logout button is broken - #ch4692
  • [Bug] When user choose a group with no dataset in spawner UI, the information in datasets section won't update - #ch4831
  • [Bug] Change script path in kc_everyone_group_default_limit.sh - #ch4452
  • [Bug] Admin dashboard discard button doesn't work - #ch4392
  • [Bug] Unable to remove logo from primehub admin console - #ch4668
  • [Bug] New spawner page group usage bug - #ch4601
  • [Bug] Fix login dialog in the primehub theme when admin enables oauth providers - #ch3266
  • [Bug] Fix OTP page - #ch5540
  • [Bug] Fix Airgap rek init config - #ch5276

Third Party

Document

  • [Doc] Create PrimeHub Manual, Admin UI Manual in Markdown format - #ch3757
  • [Doc] Update PrimeHub Manual with new progress bar in spawner - #ch3741

v2.1.4

Bugfix

  • [Bug] fix zh-TW OTP message in primehub-keycloak-theme - #ch5540
  • [Bug] primehub-console insecure cookies - #ch5637

v2.1.3

Bugfix

  • [Bug] Fix missing images in airgap environment - #ch5276

v2.1.2

Bugfix

  • [Bug] When user logout and login again immediately, he will see error message. - #ch3045

v2.1.1

Feature

  • make dataset upload beta - #ch5084

Bugfix

  • When user logout and login again immediately, he will see error message. - #ch3045
  • unable to remove logo from primehub admin console - #ch4668
  • admin dashboard discard button doesn't work - #ch4392
  • fix grafana dashboard logout issue - #ch4692

v2.1.0

What's New

The main themes of this release are:

  • User Portal (GA): A single user portal for user to access all the components and the admin console.
  • GPU/CPU Image selection: Admin can provide different image name for GPU/CPU
  • Dataset PVC Creation: Support to create dataset PVC from admin console.
  • Meta Chart: Since the 2.0 release, the meta chart is enabled by default. In release 2.1, the bootstrap script is run in the cluster and triggered when running helm install and upgrade.

Breaking Changes

N/A

Action Required

  • Because PRIMEHUB_FEATURE_USER_PORTAL is enabled by default, there are some action required:
    • For default installation, the PRIMEHUB_SCHEME, PRIMEHUB_DOMAIN, KC_SCHEME, KC_DOMAIN is required environment now. For definition, please reference configuration document
    • modify keycloak client grafana-proxy redirect URIs to https://<PRIMEHUB_DOMAIN>/grafana/*
  • Image has new spec attributes type and urlForGpu now, those two new attrs are used for seamlessly picking different image by type in spawner. For migrating existing images, please run modules/upgrade/optional/add_gpu_cpu_type_to_images.sh
  • Set a new environment variable PRIMEHUB_STORAGE_CLASS. For rke installation, please set it to rook-block. For dev-gcp, set it to standard. For dev-gce, set it to rook-block. For dev-kind set it to local-path. It is the storage class setting for all RWO pvc usage.

Features

  • Make User portal as default landing page #ch2574
  • admin-console should have a nav link back to portal #ch2859
  • Add Grafana to default list of user portal #ch2981
  • Adjust the application path #ch3495
  • Make the banner consistence in the user portal and admin console #ch3498
  • Unify primehub-console's navbar style #ch3498
  • unified ingress: deprecate PH_DOMAIN #ch1727
  • unified ingress: grafana #ch1725
  • update checklist for PrimeHub installation #ch2937
  • Tweaks for User Portal #ch2674
  • remove deprecated variables #ch3772
  • [Dataset] Admin Dashboard provides UI to create dataset pv-type pvc. #ch1794
  • Move dataset upload chart to metachart. #ch3065
  • In meta chart, use Job to perform idempotent upgrade #ch1238
  • Add images type (gpu/cpu/universal) on images list page #ch3661
  • Update image type wording both to universal #ch3393
  • conditional spawner options for image and GPU/CPU #ch2642
  • admin-ui support new ux-behavior to set up GPU/CPU in 1 image crd #ch2638
  • set maintenance notebook files to read only #ch1920
  • Make primehub-admission helm managed #ch1735
  • parameterize primehub-admission by helm-chart #ch1749
  • Refactor admission flask server #ch1909
  • crd openapi schema #ch2370
  • Support validation schema defined in the CRD (Image, InstanceType and Dataset) #ch2370
  • Script to check if the service account token valid #ch3764 Workspace
  • Provide a tool to detect and fix orphan mount & rbdmap #ch3460
  • Create ImageSpec and ImageSpecJob CRD #ch3807
  • [Custom image v1] Evaluate Tekton pipelines for image building #ch3076
  • [Custom image v1] Evaluate if metacontroller satisfy the operator needs #ch3086
  • [Custom image v1] Setup Google container registry for storing built images #ch3083
  • Design Training Job Submission MVP #ch2724
  • Workspace: Admin Console UI Changes #ch1278

Security

  • k8s CVE-2019-11253 mitigation and advisories #ch3808

Bugfix

  • [BUG] bootstrap image incompatible with the latest code (GPU, CPU feature) #ch3676
  • [BUG] memory resource literal should convert to bytes #ch3956
  • [Bug] 'feature/ch2436/provide-scripts-to-delete-orphan-user-group' break the ci job 'build:infra-upgrade-test-previous' #ch3701
  • [Bug] Attributes in everyone group is removed after editing "Default User Volume Capacity" #ch3007
  • [Bug] Default USER_PORTAL cause CI fail in infra-upgrade-test-previous #ch3934
  • [Bug] If numeric password is not allowed in bootstrap #ch3339
  • [Bug] Resource validator should check all container resource's usage #ch3414
  • [Bug] The make init generated .env incorrect. #ch3213
  • [Bug] User can't go to login page after logout from JupyterHub #ch2786
  • [Bug] When admin updates group permissions for some instances & images, hub won't auto update for users #ch3041
  • [Bug] When instance type set memory with fraction, cannot spawn #ch4082
  • [Bug] When user choose a cpu or gpu instance, then choose the other instance of same type, the image list shouldn't be reset #ch3422
  • [Bug] feature/ch2574/make-user-portal-default break the ci job 'build:infra-upgrade-test-latest' #ch3700
  • [Bug] graphql-server should terminate when oidc refresh token fails #ch2358
  • [Bug] kernel-gateway cannot see GPU device #ch3641
  • [Bug] the volume size of pv dataset is incorrect in demo.b #ch3629
  • [Bug] when user create an instance type without overcommitting values, and click the edit button again, it shouldn't show overcommitting enabled. #ch3649
  • [Bug][CI][Kind] no instance type and image can be found #ch3991
  • cloud provisioner - quota on everyone group #ch1957
  • user portal broken after group removal #ch3003
  • Reproduce the group volume migration issue and find the solution #ch3012
  • investigate dev-gcp's use of load balancer #ch3933

Third Party

  • Upgrade cert-manager #ch2673
  • Upgrade Grafana dashboard to 6.3.0 and support unified ingress #ch1725
  • followup node-problem-detector #ch2088
  • [gpu] a preview image contains nvdashboard plugin (predefine layout for UX) #ch3768
  • Survey nvdashboard jupyter plugin #ch2220

Document

  • build docs and provide as help site as part of primehub installation #ch1746
  • open source license files #ch2291
  • [Doc] Update ZD PrimeHub Manual with conditional spawner scenario #ch3389
  • [Doc] Update dataset page on ZD with two new features #ch2900
  • Beta Features Policy (eg Dataset Upload) #ch3816
  • Polish PrimeHub-Site to release-ready #ch2947
  • Spec: Hub UI can show resources usage for users' group before launch #ch391
  • Spec: design the architecture to integrate user portal, workspace, and application. #ch2976
  • Write design document for dataset upload #ch2503
  • Write document for primehub-ee metachart #ch2875

v2.0.0

Breaking Changes

  • PrimeHub is managed by a single meta chart

Action Required

  • [Downtime Required] Migrate to metachart. Please follow the instructions in the migration document to v2.0
  • Dataset PVC Please configure primehubGroupSC at etc/helm_override/primehub.yaml. For example, set standard storage-class for GKE: 
    primehub-console:
  graphql:
    primehubGroupSC: standard
    The primehub-console needs a storage-class to create dataset pvc.

Change starndard to other pvc storage class if needed.

  • [Optional] setup PRIMEHUB_FEATURE_ENABLE_KERNEL_GATEWAY flag to enable jupyter separated kernel feature
  • [Optional] setup PRIMEHUB_FEATURE_DATASET_UPLOAD flag to enable dataset upload feature

Deprecated

  • Configuration in ${primehub}/etc. Configuration files should locate in ~/.primehub/config/$(kubectl config current-context) instead of ${primehub}/etc. The preferred path can also by retrieved by the command ${primehub}/bin/phenv --path. For more information, please see customization document

Features

  • [admin] Admin Dashboard provides UI to create dataset pv-type pvc. #ch1794
  • [rke] Configure csr-approval feature enabled by default #ch1641
  • [primehub] Make the meta chart the default installation #ch2096
    • [dev] Use metachart in dev-gce and dev-gcp installation. #ch2874
    • [airgap] Make sure metachart can be installed in airgap environment #ch2876
  • [primehub] Upgrade primehub-console to v3.5.2 (backport)

Bugfix

  • [admin] Attributes in everyone group is removed after editing "Default User Volume Capacity" #ch3007
  • [hub] Bring scheduled-maintenance anti affinity back #ch2667
  • [hub] Refresh user state after the administrator reconfigred settings #ch2676
  • [hub] When admin updates group permissions for some instances & images, hub won't auto update for users #ch3041 (backport)

Documents

  • [primehub] Update PrimeHub 2 migration documents to avoid lost role binding #ch3023
  • [hub] Write a design document for persistence storage for notebooks #ch1119
  • [admin] Write design document for graphql #ch1116

v1.8.2

Security

  • [CVE-2019-9512] Upgrade nginx-ingress chart to 1.17.1 (app version 0.25.1)

v1.8.1

Bugfix

  • Fix the issue that user can't launch instance when its CPU/memory requests are disabled. #ch2475
  • Fix group sorting error in spawner page. #ch2471

v1.8.0

Action Required

  • [Optional] Disable image and instance type customization Set env READ_ONLY_ON_INSTANCE_TYPE_AND_IMAGE=true will disable image and instance type customization.

  • [Optional] For external keycloak installation, install keycloak-tools to enable keycloak idempotent upgrades.

    kubectl apply -f modules/bootstrap/keycloak-tools.yaml

  • [Optional] Apply pod's image mutation

    Label any namespaces which want to apply pod's image replcaing. For example:

    kubectl label ns hub primehub.io/image-mutation-webhook=enabled

    In an airgap environment, that should be all namespaces. This version, it keeps both PRIMEHUB_AIRGAPPED and image-mutation for backward compatible.

  • [Optional] Portal Default Cover

    Since 1.8.0, we add serveral default cover images for portal link, you can override ui.portalConfig setup to use it:

    • Gitlab: /admin/default-covers/gitlab.png
    • Mattermost: /admin/default-covers/mattermost.png
    • Jupyter: /admin/default-covers/jupyter.png default portal setup in modules/charts/primehub-console/values.yml:
    portalConfig: |
  services:
    # Service portal setup example:
    # - name: Example
    #   uri: "https://example.com"
    #   image: "/admin/default-covers/default.png"
    #   adminOnly: true | false
    #- name: Gitlab
    #  uri: "https://gitlab.com/infuseai"
    #  image: "/admin/default-covers/gitlab.png"
    - name: JupyterHub
      uri: "/hub"
      image: "/admin/default-covers/jupyter.png"
    - name: Support
      uri: "https://infuseai.zendesk.com/hc/en-us"
      image: "/admin/default-covers/support.png"
    - name: JupyterHub Admin
      uri: "/hub/admin"
      adminOnly: true
      image: "/admin/default-covers/jh-admin.png"
    - name: Admin Dashboard
      uri: "/admin/cms"
      image: "/admin/default-covers/admin-ui.png"
      adminOnly: true
    - name: Maintenance Notebook
      uri: "/admin/maintenance"
      image: "/admin/default-covers/notebook.png"
      adminOnly: true
  welcomeMessage: >

Features

  • Add new icons in user protal. #ch1860
  • Implement an operator to control dataset upload server. #ch1796
  • Spawner UI will order instanceTypes & images options in alphabetical order. #ch1742
  • Use PrimeHub logo in JupyterHub navbar and can be used to redirect to user portal. #ch1740
  • Support feature toggling infrastructure to provide an alternative to maintaining multiple source-code branches. #ch1675
  • Add new template condition for readonly image and instanceType. #ch1595
  • Design document for JupyterHub integration. #ch1114
  • User won't be stuck in spawning page when resources is not enough. #ch1020
  • Support idempotent bootstrap and upgrade for external keycloak setup.

Bugfix

  • Fix out of range issue when resizing user volume. #ch2125
  • Fix missing space issue in "connect existing groups" button. #ch2070
  • Fix deprecated fields in z2jh to avoid warning and provide forward compatibility. #ch231
  • Fix the issue that group loading too long if there are too many users. !453
  • Fix unexpected error when user without any dataset in launch group. !445
  • Fix user storage capacity unit. !440
  • Fix typo of terraform load module. !437

v1.7.0

  • Fix missing change image.txt !421
  • Bump admin-notebook version with user and group cleanup !420
  • Speed up OSD backfill !419
  • Fix jupyterhub chart uses wrong image for airgap !418
  • Remove skip deps for old helmfile !417
  • Fix ci docker version !415
  • Write design document for CRDs !414
  • Refine primehub version provides way !413
  • Support admission hook
  • Rewrite template logics !411
  • K8S ca signer checker !410
  • Use feature flag to support old convention for group volume !407
  • Update notebook image tag and move some fixed config into admin notebook image !406
  • Add ansible playbook for migrate aia bluestore !405
  • Add User portal mvp admin shows regular user !404
  • Unify admin hub ingress !402
  • Fix: launchGrouponly should work for each type !401

Breaking Changes

  • Replace admin-ui with primehub-console
  • Unify domain name

Action Required

  • [Downtime Required] Migrate primehub-groupvolume.

    After upgrade v1.7.0 primehub-groupvolume, need to run migration script to migrate exist project-volume and dataset-volume.

    After migrating primehub-groupvolume until running the migration process, user pod can not access the data in project and dataset volume.

    For the production environment, it's better to schedule downtime upgrade with customer, or we need to let customer known user will encounter data access error during the upgrade process.

    • RKE environment will run rke_migration.sh script for migration or follow the migration process provided in READMD.md.
    • GKE environment will run gke_migration.sh script for migration. (See modules/primehub-groupvolume-nfs/README.md for more info.)

  • Migrate admin-ui to primehub-console

    1. Add the environment variables for image pull secret to .env

      PRIMEHUB_CONSOLE_DOCKER_USERNAME=
PRIMEHUB_CONSOLE_DOCKER_PASSWORD=

      The value can be found from

      kubectl -n primehub get secret gitlab-registry -o jsonpath='{.data.\.dockerconfigjson}' | base64 --decode | jq .

    2. Add the helm override yaml.

      Old sample (admin-ui.yaml):

      primehub:
  keycloakUrl: https://id.example.com/auth
  adminLocale: en
  adminUrl: https://example.com/admin
  appPrefix: /admin
  crdNamespace: hub

      Should migrate to new one (primehub-console.yaml):

      keycloak:
  url: https://id.example.com/auth

ui:
  locale: en
  url: https://example.com/admin
  appPrefix: /admin

graphql:
  crdNamespace: hub

      Before install primehub-console, you should remove old admin-ui release:

      helm del --purge admin-ui

  • [Optional] User portal page

    1. PRIMEHUB_FEAURE_USER_PORTAL=true will enable universal landing page for PrimeHub users.
    2. You can setup items link and welcome messages on landing page in primhub-console.yaml:
    ...
ui:
  ...
  portalConfig: |
    services:
      - name: Google
        uri: "https://www.google.com"
        image: "https://www.google.com.tw/images/branding/googlelogo/2x/googlelogo_color_272x92dp.png"
      - name: GitLab
        uri: "https://gitlab.com"
        image: "https://shiring.github.io/tutorials/2017/09/04/GitLab_logo.png"
      - name: JupyterHub
        uri: "https://hub.aaron.dev.primehub.io"
        image: "https://cdn-images-1.medium.com/max/2400/1*m87_Htb_9Pstq0UcvNJ49w.png"
      - name: Admin Dashboard
        uri: "/cms"
        adminOnly: true
        image: "https://github.com/raw/InfuseAI/primehub/master/docs/media/logo.png"
      - name: Maintenance Notebook
        uri: "/maintenance"
        adminOnly: true
        image: "https://github.com/raw/InfuseAI/primehub/master/docs/media/logo.png"
    welcomeMessage: >
      <h1>Getting Started</h1>
      <div>End of Lines</div>

  • [Optional] Migrate to unified domain name

    1. Add PRIMEHUB_FEATURE_USER_PORTAL=true, PRIMEHUB_DOMAIN= and PRIMEHUB_SCHEME= to .env.
    2. Add new Valid Redirect URIs to all affected keycloak clients. (e.g. admin-ui, jupyterhub, maintenance-proxy)
    3. For more information, please see docs/design/user-portal.md

v1.6.2

  • Feature flag to support old convention for group volume [ch1257]
  • Support extra pod and node affinity [ch2667]
  • Upgrade admin-ui to v2.12.5
    • Fixed bug: Attributes in everyone group is removed after editing "Default User Volume Capacity" [ch3007]

Deprecated

  • To keep using legacy project group naming convention, do NOT run modules/upgrade/optional/kc_group_volume_migration.sh and need to add SUPPORT_OLD_GROUP_VOLUME_CONVENTION=true to your env file. This flag will be applied when you upgrade hub.

v1.6.0

  • Fixed Bug: metacontroller primehub-group nfs pvc deletion (reproduce) [ch378]
  • Fixed cpu limits configured in group quota are not effective [ch423]
  • Fixed User can't spawn pod because volume manager try to chown everything [ch728]
  • Implement the single upgrade script for PrimeHub [ch294]
  • JupyterLab 404 message patch [ch341]
  • Group volume size & options [ch361]
  • Support Image Pull Secret for Images [ch362]
  • Users in everyone group has no quota by default [ch497]
  • Helm chart for maintenance notebook environment. [ch557]
  • Upgrade nginx-ingress chart to 1.6.16 and increase client-header-buffer-size to 16k [ch708]
  • Implement dataset launchGroupOnly in JupyterHub spawner [ch710]
  • Support Maintenance notebook feature. [ch711]
  • Patch rke nginx-ingress configMap [ch713]
  • nvidia-gpu-device-plugin has too small memory limit [ch419]

Action Required

  • In order to make primehub-admin-notebook to work properly, You'll have to increase the client_header_buffer_size of your nginx-ingress up to 16k. (See modules/charts/nginx-ingress.yaml for example.).

    For rke, this change need patch nginx-ingress config map, please run the following command:

    kubectl -n ingress-nginx patch cm nginx-configuration --type merge -p '{"data":{"client-header-buffer-size":"16k"}}'

  • The system group (everyone) now has resource limit by default. In previous version, everyone group has unlimited resource limits. They are now set to 0 by default. If you want to adopt this behavior when upgrading, please source your env file and run modules/upgrade/optional/kc_everyone_group_default_limit.sh to apply.

  • Groups now have options to enable shared volume (project group) and set volume capacity on Admin UI. In previous version, project groups depend on naming convention that group name starts with Project-. To migrate existing project groups to new ones with options, please source your env file and run modules/upgrade/optional/kc_group_volume_migration.sh. The migration script will rename existing project groups to remove the prefix Project-, so please check if there is any group name conflict by running modules/upgrade/optional/kc_check_group_names_conflict.sh.

  • Dataset now has option Launch Group Only and will display Mount Root on Admin UI. To migrate existing datasets, please run modules/upgrade/optional/add_mount_root_to_datasets.sh.

Deprecated

  • Project group naming convention is deprecated. Use group options to enable shared volume instead of naming group starts with Project-.

v1.5.0

  • Minimum autoscaling support [ch73]
  • map keycloak superuser role to jupyterhub admin [ch65]
  • Improve JupyterLab UX on reconnect and server stop [ch42]
  • Reorg kubectl plugins [ch130]
  • Script to create/delete keycloak user for hub load testing [ch169]
  • Implement the single upgrade script for PrimeHub [ch294]
  • Planning for cuda10/nvidia410 and tf-2.0 tests [ch70] !302
  • Adjust default JupyterPageSlow alert rules [ch166] !304
  • Setup stable demo environment for potential clients [ch66]
  • Replace minikube with KIND in CI [ch68]
  • Add unit test to our CI [ch312]
  • Basic auto deploy demo hub [ch67]
  • Setup g suite oauth <-> b.demo [ch136]
  • In hub config, singleuser.nodeSelector is not working [ch145]
  • Remove Stop My Server button in control panel after server stopped [ch259]
  • JupyterLab UI is not rendering properly on JupyterHub 1.0 [ch128]
  • Group name should not be editable. bump admin-ui to v2.8.2 !294 canner-admin-ui#104

Action Required

  1. After install helm-managed, also run the upgrade make primehub-upgrade

  2. To use the kubectl plugins, set the $PRIMEHUB_HOME/bin in the PATH environment variables or copy binaries inside to where the env $PATH can resolve it. Here are the kubectl plugins provided

     # shortcut to run ceph command
 kubectl ceph
 # shortcut to run rbd command
 kubectl rbd
 # wrapper of kubectl scale
 kubectl stasis
 # force delete
 kubectl fdelete

  3. The nvidia memory limit changes. Please increase memory for nvidia device plugin

     kubectl apply -f modules/nvidia/device-plugin/nvidia-device-plugin.yaml

v1.4.0

  • Adjust group user volume capacity logic (min -> max) #457
  • Fix the Gitlab CI to upgrade the kubernetes version #456
  • Modify monitoring script for check_node_cordoned not always throw alert #455
  • Limit user process and other resources that might be harmful to node #454
  • Show primehub version in hub launch page #453
  • Initial PrimeHub chart phase1 #452
  • The docker registry should not be stored at the tmp folder #451
  • Add merge request template #449
  • KC password cannot be a number #448
  • Cpu limit issue on demo hub #447
  • [CVE-2019-9946] Upgrade kubernetes to 1.12.7 #446
  • [CVE-2019-10255] Upgrade JupyterHub to 0.9.6 #445
  • [CVE-2019-1002101] Upgrade kubectl binary to 1.12.7 #444
  • Full cluster shutdown script for rook 0.9 #443
  • Unifying user volume capacity keys #441
  • Full cluster shutdown for rook 0.9.x #440
  • Reorg kubectl plugins #439
  • Rsync script for gitsync operation #430

v1.3.0

Al hail Himalaya!

  • Set TZ env for user notebook #396
  • Make sure group volume works when default storage class is set #170 stage/qa
  • Install Tensorboard-enabled image to Demo Hub #259
  • Grafana dashboard for jupyterhub #271
  • Backup script for group volume #346
  • [CVE-2019-1002101] Upgrade kubetctl binary to 1.12.7 #444
  • Failed to install nvidia-driver in airgap #429
  • User quota bug in admin UI #194
  • Make disable_cfs_quota as default in kubelet #433
  • Improve keycloak backup/restore #395
  • Ceph metadata device on VG #401
  • Set rook-block sc default reclaim policy to Retain #432
  • Don't override the image in helm value for non-airgapped setting #417
  • Cleanup for modules/support #416
  • Fix dev-gce Makefile for new installer #414
  • Ensure dev-gce install in airgap #406
  • Automate dev-gce airgap environment setup #403
  • Install Tensorboard-enabled image to Demo Hub #259
  • Remove public keys from repo #415

Action Required

  • Migratie Keycloak client jupyterhub mapper by running the migration script

      ./modules/bootstrap/kc_migration.sh

  • Migrate group volume nfs metacontroller

      ./modules/primehub-groupvolume-nfs/upgrade.sh

  • upgrade kubectl to 1.12.7. copy ./bin/kubectl to $PATH

  • upgrade k8s to 1.12.7. Please following the steps

    1. edit rke configuration file cluster.yml

    2. Add a system_images.kubernetes to rancher/hyperkube:v1.12.7-rancher1 and don't change the kubernetes_version

       kubernetes_version: v1.12.4-rancher1-1    <--- leave v1.12.4. This is used for rke
 system_images:
   etcd: rancher/coreos-etcd:v3.2.24
   kubernetes: rancher/hyperkube:v1.12.7-rancher1    <-----
   alpine: rancher/rke-tools:v0.1.16
   nginx_proxy: rancher/rke-tools:v0.1.16

    3. do the rke up

v1.2.0

  • Group volume (metacontroller) does not work in dev-gce #383 type/bug
  • HOWTO document for add/remove nodes and change the controlplane node #382
  • EFK should only save hub & jupyter logs with log level ≥ warning #378
  • Dataset refresh should not require restarting server #354 type/bug
  • Upgrade rook 0.9.1 #344
  • Test rook bluestore layout for mixed devices #316
  • Survey how to remove a ceph node from cluster #386
  • Upgrade docker for security #404
  • Airgap image push script should stripe shasum #405
  • Dev-gce one command setup script #399
  • Full shutdown/restart script #361
  • Remove ansible templates from ansible "release builder" role #319
  • Refine helmfile process for downloaded charts in airgapped mode #297
  • Per-node upgrade helper #250
  • Survey how to add/remove/manage etcd by rke #385

v1.1.0

  • Jupyter-notebook setup have pip timeout issue when enable network policy #392
  • Upgrade kibana to 6.5.1 #391
  • Tensorboard start-notebook-d conflict with network policy #390
  • Remove not being referenced anymore files in installer/skeleton/ #388
  • Allow multiple login actions for newly created users #384
  • Lock EFK chart version #371
  • Add make install-required-ansible-roles back #366
  • Prometheus-node-exporter does not have resources requests/limits #353
  • Investigate better way to do nvidia.com/gpu: 0 #352
  • Default nodeSelector for fluentd #349
  • Fixed tensorboard serverextension is not correctly set up. #339
  • Correct namespace of primehub-groupvolume-nfs #338
  • Test upgrade path to kubernetes 1.12 #335
  • Show GPU utilization information from dashboard #318
  • Improve pod-level OOM behaviour #317
  • Installer should use downloaded helm binary to pull the remote helm chart #311
  • Keycloak image not overridden with airgap #310
  • Lock prometheus chart version #304
  • Custom search for KeyCloak events in Kibana #296
  • Calico should only throw log with level >= warning #295

v1.0.0

  • Update kubernetes to 1.10.11 for rke install. CVE-2018-1002105 (#300)