Release Manager

@hzy46

Endgame

Feature freeze: TBD
Code freeze: 2.24
Scrum demo date: 2.25
Bug Bash date: TBD
Release date & retrospective date: 2.28

Test Plan:

TBD

Work Items

Job submission page new UI

• P0 side bar refine
• P0 basic info + task role
• P0 More info (advanced mode)
• P1 SKU (hived scheduler logic)
• P1 secrets function (including image auth)
• P1 ssh function
• P2 data (team storage)
• P2 save as template

Job protocol update

• P0 support data via extending prerequisite in job protocol extend prerequisite field in job protocol #5145
  • cmd runtime plugin modification
  • let runtime plugin parse prerequsite
  • change validation in webportal and rest-server
    • make uri optional
  • change openpai protocol
• P1 Host prerequisite in marketplace

Marketplace 2021 Feb. Release Plan

Dockerhub pull policy #5219

• Docker image pull frequency limit in dockerhub
  • P0 use ansible notebook to change docker daemon config (solve job pull and service start problems)
  • P0 start a new cache registry service to cache images from dockerhub
  • P1 change service yaml config when starting service to use mirror registry
  • leave an interface for external registry
• Test:
  • submit a job using dockerhub image
    • cache registry log shows that it receives a pull image request from the job container
  • batch submit 100+ simple jobs
    • cache registry log records all the pull image requests
    • check the worker node rate limit and it does not reach the limit

DB controller

• make DBController tolerant to wrong framework request make DBController tolerant to wrong framework request #4889 When user's job config is too large, OpenPAI job will be in Waiting or Stopping forever  #5093
• Remove sensitive Info in database

GPU Utilization Statistics

• create CronJob to send GPU utilization report @suiguoxin send regular GPU utilization report with CronJob #5281
• add gpuhours info and sort by gpuhours in the report @suiguoxin [alert-manager] add gpu*hours info in cluster-utilization cronjob #5294

Let runtime plugin access a "job application token" @suiguoxin

• generate a job application token for every job and mount it to initContainers generate an application token for every job #5270
• save job specific token in sperate namespace #5292 same job tokens in a seperated namespace
• Disable it by default in this release disable job specific token generation in rest-server #5312

save SSH publish keys on user profile page #5274

• webportal @Lijiaoa Add ssh public keys on user-profile page #5223
• openpai-runtime @suiguoxin
• Mount user's extension in runtime container create a user secret for every job to save user extension #5310

openpai runtime

• design discuss: [Proposal] openpai-runtime interface openpai-runtime#12 @Binyang2014

log experience

• Add ".log" extension to downloaded log file @Binyang2014 Add ".log" suffix for user-stdout/user-stderr/user-all log files #5272
• Support read log file behind the customized gateway @Binyang2014 remove log url WEBPORTAL_URL prefix #5271

Deployment

• P2 verbose mode for deployment scripts @Starmys

Agile CI and nightly-build

P1 Agile CI and split heavy tests with nightly-deployment #5173 @yiyione

• Setup the nightly-deployment and test
• Build & publish nightly tag image

Fault Document

@mydmdm

Document

• update add&remove node doc @suiguoxin
• update aad doc @suiguoxin

Bug Fix

• Memory leak in framework watcher of database controller Memory leak in framework watcher of database controller #5300
• Set correct launchTime in rest-server @suiguoxin fix get jobs api launchedTime field issue #5307

Backlog

HiveD scheduler

• P2 Cell as sku in hived scheduler. @abuccts (backbone support (config) for "Cell", submission form supports for Cell) @hzy46
  ETA: design: in progress
• P2 UX for HiveD @yangou1988 @hzhua

Show cluster-level info #5254

• Show cluster-level information for admin in webportal @TobeyQin @hzy46

dependabot alert

• P1 add / remove node with with layout.yaml Add / Remove Nodes with layout.yaml #5167 @suiguoxin

Need triage

• Rest-Server Job API Perf Tests and Improvements Rest-Server Job API Perf Tests and Improvements #5027
• Python 2/3.5 Deprecation Upgrade Python 2 to Python 3 #5042
• P2 dev-box can be inside master node: doc, testing
• Refine Alert System [Monitor] Alert System Refine #4810 P2 items
• P2 too many tokens created Too many identical tokens have been created. #5123
• P2 Group management & VC request management for user and admin @yiyione Group & VC request management for user and admin #4949
• P2 vscode bug @yiyione
• P2 code coverage concern when developing feature
• P1 (design) Submit job ui should show protocol prerequisites parts 2020 Oct. ~ Nov. Release Plan openpaimarketplace#73 @debuggy @yiyione @TobeyQin

Technical Investigations

• Cluster API - blog @suiguoxin
• KubeVela @SwordFaith
• Open Application Model
• Crossplane