2021 March Release Plan

[scarlett2018]

Release Manager

@Binyang2014

Endgame

Feature freeze: TBD Code freeze: 4.6 Scrum demo date: TBD Bug Bash date: 4.16 Release date & retrospective date: 4.26

Test Plan:

TBD

Top level themes (work item break down needed)

• Marketplace v1 backlog - @debuggy / @yiyione / @TobeyQin 2 weeks P0 May need to delay one week Test Owner: @suiguoxin @debuggy @hzy46

  Marketplace Mar. Release Plan #233

• x-plan - @Binyang2014 TBD Test Owner: @suiguoxin Done

Alert-Manager Test Owner: @Binyang2014 Test done

• [ ] send alert to user when job failed #5337 @suiguoxin P1 (defer)
• [x] Add alert & auto-fix for GPU perf issue #5342 #5383 P0 Test cases: https://github.com/microsoft/pai/pull/5383#issuecomment-812286658
• [x] Add kill-long-running-job email templates #5384 Test cases: https://github.com/microsoft/pai/pull/5384#issuecomment-812290954

Rest Server Test Owner: @yiyione Done

• [x] support sort by completionTime in get the list jobs API #5347 @suiguoxin
  • [x] API change #5375 P0 test cases: https://github.com/microsoft/pai/pull/5375#issuecomment-812292773
  • [x] application of this API in cluster-utlization #5376 P0

Deployment Test Owner: @Binyang2014

• [x] add / remove nodes with layout.yaml #5321 #5167 @Starmys P0 Test Done
• [x] webportal package build issue #5378 @suiguoxin P0 Test cases: https://github.com/microsoft/pai/pull/5378#issue-592689628
• [x] K8s API server's cert need renew each year #5334 P0 Test cases: https://github.com/microsoft/pai/issues/5334#issuecomment-815412607

Documents

• [x] Doc for renew API server doc @yiyione
• [x] Document for config.yaml @Starmys
• [x] Document for new submission page, user manual @debuggy
• [x] Add remove nodes
• [x] Doc Nvidia driver version mismatch

Other backlogs

Use case & best practice

• Use case and best practice summary - @hzy46 / @TobeyQin / @suiguoxin
  • OpenPAI Advantage
  • OpenPAI Best Practice -- P0 topics:
    • Cluster setup and onboarding
    • Utilization weekly report
    • Storage
    • How to debug
    • Leverage low-priority resources
    • AutoML
• ~Job profiling @hzy46 P1~ Not in PAI release
• HiveD user experience support - tbd @yangou1988 VC view page(design review in this release) P1
• HiveD convert old test cases and propose new test cases @hzy46 P1
• Dataset: integrate data prerequisite into marketplace and job submission page @hzy46 #5345 TBD P1