ssThis story will track Upgrade testing. 

Steps:

• Create an OCP cluster (version to be upgraded)
• Install cert-manager
• Install Kueue Operator and Operand (w/ Jobs, Pods, Deployments and StatefulSets)
• Create the Resources (ResourceFlavor, ClusterQueue and LocalQueue)
• Create some pods and logs
  • Check that workloads were created
• Upgrade OCP 
  • Check the Resources
  • Pods and Jobs probably won't be kept once we're not supporting for now seamless upgrade
• Create some pods and logs (using the Resources created before the upgrade)
  • Check that workloads are still being accepted
• Run make test-e2e

Tests:

• OCP 4.18 Kueue 1.3 -> OCP 4.19
  • 418_upgrade_tests.txt
• OCP 4.19 Kueue 1.3 -> OCP 4.20
  • 419_upgrade_tests.txt
  • 419_upgrade_tests-rerun_failed.txt- initial run had 4 failures. All tests passed on the second attempt, confirming no functional regressions were introduced.
  • Besides proposed tests I also manually verified VisibilityOnDemand (using the 2 versions of API) and LocalQueue Default after upgrading. Both tests have Passed.
• OCP 4.20 Kueue 1.3 -> OCP 4.21

• • LocalQueueDefaulting and VisibilityOnDemanded - tested
  • 420_upgrade_tests.txt
  • 420_upgrade_tests-rerun_failed.txt
• OCP 4.21 Kueue upgrade from 1.2 to 1.3
  • Partial Uninstall 
    • FBC created
    • Kueue 1.2 installed
      • Obs: Kueue 1.2 tests are currently passing on OCP4.21 - here
    • Resource Flavor, ClusterQueue, LocalQueue, Pods and Jobs were created. 
    • Kueue was partially uninstalled
      • Delete all operand instances for this operator - not checked
    • All resources were kept (Resource Flavor, ClusterQueue, LocalQueue, Workloads  and Kueue/Cluster)
    • Kueue 1.3 installed
      • Installation finished successfully (Operand is the same)
    • New pods/jobs were created using the same resources
    • Running e2e tests to check if everything still works partial_uninstall.txt
  • Full Uninstall 
    • FBC created
    • Kueue 1.2 installed
    • All resources were kept (Resource Flavor, ClusterQueue, LocalQueue and Workloads)
      • Kueue/Cluster was delete as expected
    • Kueue 1.3 installed
    • New pods and jobs created
    • Ran e2e tests
      • full_uninstall.txt
      • full_uninstall-rerun_failed.txt 
• OCP 4.21 Kueue 1.3 Full Uninstall 
  • The difference now from Partial to Full Uninstall it's just the Operand.
    • When we uninstall Kueue partially and install it again, Operand is available (with the previous config) once installation ends;
    • When we uninstall Kueue fully and install it again, Operand is not available. We need to install it again.
      • On both, resources (ResourceFlavor, ClusterQueue, LocalQueue and Workloads) previously created are kept.
  • 421_full_uninstall.txt
  • 421_full_uninstall-rerun_failed.txt