koordinator-sh / website Goto Github PK

Currently, the user manuals of some features (including CPU Burst, Memory QoS, Memory Evict, ...) are archived under the blogs. It may trouble new users who want to learn or use these features.

User manuals to add:

• CPU Burst #94
• Memory QoS #100
• TBD: Memory Evict (the feature is easy to use, maybe we'd better add documentation about implementing it via our new QoS Manager)

shows the most recently released version in home page

Improve the homepage of the document to introduce the advantages of the project

https://github.com/koordinator-sh/website/blob/main/docs/user-manuals/multi-hierarchy-elastic-quota-management.md#used--runtime-revoke

Suggestions:

1. To delete over-resource-used pods, we need to set enableCheckParentQuota: true in koord-scheduler-config, otherwise there would be no pods revocation.
2. To enable deletion, we need to add the following config to koord-scheduler-role ClusterRole, otherwise we would get error messages like User "system:serviceaccount:koordinator-system:koord-scheduler" cannot create resource "pods/eviction" in API group "" in the namespace "xxx"

- apiGroups:
  - ""
  resources:
  - pods/eviction
  verbs:
  - create

Task list：
Status Article Translator PR

• Project i18n initialization @leason00 #89
• Home page @leason00 #89
• Getting Started
  • Introduction @leason00 #93
  • Installation @leason00 #93
• Architecture
  • Overview @LeoLiuYan #92
  • Resource Model @LeoLiuYan #92
  • Priority @LeoLiuYan #92
  • QoS @LeoLiuYan #92
• User Manuals
  • Colocation Profile @dingbb #102
  • Load Aware Scheduling @dingbb #102
  • Fine-grained CPU Orchestration
  • Resource Reservation
  • PodMigrationJob
  • Device Scheduling
  • GangScheduling @Ghostbaby #99
  • Multi Hierarchy Elastic Quota Management
  • CPU Suppress @Ghostbaby
  • CPU Burst @Ghostbaby
  • Memory QoS
• Design Details
  • Koordlet @leason00 #95
  • RuntimeProxy @leason00 #96
  • Enhanced Scheduler Extension
  • Load-aware Scheduling @leason00 #103
  • Fine-grained CPU orchestration @leason00 #111
  • Resource Reservation @leason00
  • PodMigrationJob @leason00
  • Descheduler Framework
  • Fine-grained device scheduling
  • GangScheduling @Ghostbaby #110
  • Multi Hierarchy Elastic Quota Management
• Best Practices
  • Colocation of Spark Jobs @wangfeng1983 #113

What is your problem or scenario?

We need more documentation on dependencies. If possible, provide an installation tool to check whether the dependencies are met.

Koordlet will throws an error when the kubelet readonly port is not open. But I didn't find the relevant info in the documentation.

To improve the doc quality, pls add a spell checker in the github workflow.

The current document is loosely tiled, and it is difficult to systematically obtain a certain aspect of content, such as what should be paid attention to in the colocation practice.

Now koordinator only supports kubernetes v1.18 ~ v1.24, but in the website documentation, it only shows support >= kubernetes v1.18. We should clarify supported Kubernetes versions.

https://koordinator.sh/docs/best-practices/colocation-of-spark-jobs

We should add sections about:

• design motivation, why do we need Colocation Profile?
• mechanism, add an architecture diagram to help understand how the Colocation Profile works
• more examples