Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

YUNIKORN-184. Update YuniKorn-Core Design Documentation #158

Merged
merged 1 commit into from Jun 12, 2020

Conversation

wangdatan
Copy link
Contributor

Updated initial draft.


### Components:

```
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can we use a chart for this?
This is a bit hard to read.

|
|Write Ops +----------------+
+-------------+ V ++Scheduler |
|ConfigWacher + +---------------+ Allocate || And |
Copy link
Contributor

@yangwwei yangwwei May 22, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Guess we are missing PartitionManager, placement rules etc here.
And the rest server.

|
+-------------------+-------------------------+
|--------+ +------+ +----------+ +----------+ |
||Node | |Queue | |Allocation| |Requests | | +
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Missing Application?


#### Scheduler Cache

Caches all data related to scheduler state, such as used resources of each queues, nodes, allocations. Relationship between allocations and nodes, etc.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Missing application

@TravisBuddy
Copy link

Hey @wangdatan,
Your changes look good to me!

View build log

TravisBuddy Request Identifier: a71ce1a0-9c86-11ea-8472-279b0b07e116

@codecov-commenter
Copy link

Codecov Report

Merging #158 into master will not change coverage.
The diff coverage is n/a.

Impacted file tree graph

@@           Coverage Diff           @@
##           master     #158   +/-   ##
=======================================
  Coverage   58.62%   58.62%           
=======================================
  Files          62       62           
  Lines        6096     6096           
=======================================
  Hits         3574     3574           
  Misses       2381     2381           
  Partials      141      141           

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update fea0d73...87161c9. Read the comment docs.

```

+---------------+ +--------------+
|K8s Shim | |YARN Shim |
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is it a good idea to mention the YARN shim here, if we don't have it implemented yet? I think it would be better to display only the K8s one and add a note that it can be extended with others, such as a YARN Shim

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

May be we can give a reference to SI here than shim
And cover more that in respective shim's as each shim may differ in impl


**Allocation by application**

When it goes to Application, see (`scheduler_application.go: func (sa *SchedulingApplication) tryAllocate`), It first sort the pending resource requests belong to the application (based on requests' priority). And based on the selected request, and configured node-sorting policy, it sorts nodes belong to the partition and try to allocate resources on the sorted nodes.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think it worths to mention somewhere what are the available sorting policies and how to configure and create a custom one. (or a link to a configuration page, when we can add this info)

+-------------+ V ++Scheduler |
|ConfigWacher + +---------------+ Allocate || And |
+-------------+ |Scheduler Cache| <-----------------| |
+----------> +---------------+ Preempt ++Preemptor |
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Its better we keep config out of core area? as anyway it will be a CRD etc later.
So it may confuse if we put it more tightly coupled.

Copy link
Contributor

@kingamarton kingamarton left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe it fits more in the shim part, but it would be useful to have a small description about the relationship between different entities, like pod, app, task

@yangwwei
Copy link
Contributor

yangwwei commented Jun 12, 2020

based on the discussion, let's merge this PR first. Later on, we can do more improvements based on this. cc @kingamarton @sunilgovind feel free to create follow up JIRAs to improve this doc.

@yangwwei yangwwei merged commit 42763f5 into apache:master Jun 12, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
6 participants