[YUNIKORN-3121] Add REST Endpoint for Scheduling Order Visibility

[mitdesai]

What is this PR for?

Added a new Rest Endpoint for getting the scheduling order - These come in handy when debugging issues in scheduling.

What type of PR is it?

• - Bug Fix
• - Improvement
• - Feature
• - Documentation
• - Hot Fix
• - Refactoring

Todos

• - Task

What is the Jira issue?

https://issues.apache.org/jira/browse/YUNIKORN-3121

How should this be tested?

Screenshots (if appropriate)

Questions:

• - The licenses files need update.
• - There is breaking changes for older versions.
• - It needs documentation.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 81.83%. Comparing base (e41a426) to head (28895fa).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1042      +/-   ##
==========================================
+ Coverage   81.76%   81.83%   +0.06%     
==========================================
  Files         103      103              
  Lines       13535    13557      +22     
==========================================
+ Hits        11067    11094      +27     
+ Misses       2208     2203       -5     
  Partials      260      260              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[pbacsko]

@mitdesai please take care of the linter problem. You can check locally by running "make lint".

[manirajv06]

Yes, it helps.

I propose orderLog or something like that similar to allocationLog to capture the order of events. We can come up with list of constant or message to be logged for each important step. As and when each step pass through, log the corresponding constant/message to the orderLog from the originating place as opposed to simulating the code given in this PR as it does not require us to update both the places. Finally, it can be dumped in the state dump. Not really sure this is something needs to be exposed through REST API.

In addition, we can also capture these kind of cases as well going forward.

[mitdesai]

I like the idea of in-place updates and maintaining the orderLog. But I still feel that it should be a separate endpoint even though we put this information in the state dump.

State dumps are huge. And when we are debugging a scenario where some pods are not getting scheduled and we just need to know the order of what is getting scheduled, it is quicker and less expensive to get selective information.

[manirajv06]

State dump is the place where we keep all these diagnostic info in general. It is just a matter of looking for "orderLog" string in the dump :)

[manirajv06]

Map containing applications does not guarantee specific order. Since we are looking for scheduling order at any given moment, I would rather log into data structure which strictly maintains order based on the insertion order so that it can consumed as is in the state dump. I can see Queue.TryAllocate as the best place to push the application into the above mentioned slice.

[mitdesai]

Ah, Good Catch. Not sure how I missed that. Let me fix it.