[YUNIKORN-1597] e2e tests: Recover assigned pods in Bound state

[pbacsko]

What is this PR for?

The PR contains an e2e test which validates that YK can recover properly with Running/Pending placeholders.

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-1597

How should this be tested?

Screenshots (if appropriate)

Questions:

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

[craigcondit]

I'm -1 on this approach as it doesn't allow for running on anything but a Kind cluster. Our tests should be cluster-agnostic, and adapt to the environment they are run in so that they can be used on real clusters as well.

[craigcondit]

Alternate approach: Query the cluster for nodes with available capacity, and construct pods with nodeAffinity specified so that they are only schedulable on those nodes.

[pbacsko]

I'm -1 on this approach as it doesn't allow for running on anything but a Kind cluster. Our tests should be cluster-agnostic, and adapt to the environment they are run in so that they can be used on real clusters as well.

Yeah, most tests are agnostic, but there are some (maybe just the one which I modified) which already depends on having 3 nodes (1 control + 2 worker). Perhaps the existing tests should be fixed, too in a follow-up JIRA.

Alternate approach: Query the cluster for nodes with available capacity, and construct pods with nodeAffinity specified so that they are only schedulable on those nodes.

Hm, interesting. I'll try this.

[codecov]

Codecov Report

Merging #544 (30beb9b) into master (20116e5) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #544   +/-   ##
=======================================
  Coverage   69.97%   69.97%           
=======================================
  Files          46       46           
  Lines        7881     7881           
=======================================
  Hits         5515     5515           
  Misses       2163     2163           
  Partials      203      203           

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[pbacsko]

Test failure seems to be unrelated.

[manirajv06]

@pbacsko PR has introduced new methods to set/remove node labels etc. There was a need to add taint/untaint node methods as well to remove the dependency on k8s test framework. Hence used some methods from this PR to above said PR. So, Please rebase your wc and take it forward.

[craigcondit]

+1 LGTM.