From b380a29baadafc7c23ea9cbf26ce62d8ae896e98 Mon Sep 17 00:00:00 2001
From: Manikandan R <manirajv06@gmail.com>
Date: Fri, 23 Aug 2024 11:15:18 +0530
Subject: [PATCH 1/2] [YUNIKORN-2808] E2E test
 Verify_preemption_on_priority_queue test is flaky

---
 pkg/scheduler/objects/preemption.go | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/pkg/scheduler/objects/preemption.go b/pkg/scheduler/objects/preemption.go
index 3184658a5..677a82d09 100644
--- a/pkg/scheduler/objects/preemption.go
+++ b/pkg/scheduler/objects/preemption.go
@@ -176,6 +176,11 @@ func (p *Preemptor) initWorkingState() {
 	p.allocationsByNode = allocationsByNode
 	p.queueByAlloc = queueByAlloc
 	p.nodeAvailableMap = nodeAvailableMap
+	log.Log(log.SchedPreemption).Info("Preemption triggered based on working state", zap.String("ask", p.ask.GetAllocationKey()),
+		zap.String("ask queue path", p.queuePath),
+		zap.Int("allocationsByNode", len(p.allocationsByNode)),
+		zap.Int("queueByAlloc", len(p.queueByAlloc)),
+		zap.Int("nodeAvailableMap", len(p.nodeAvailableMap)))
 }
 
 // checkPreemptionQueueGuarantees verifies that it's possible to free enough resources to fit the given ask
@@ -198,6 +203,8 @@ func (p *Preemptor) checkPreemptionQueueGuarantees() bool {
 			snapshot.RemoveAllocation(alloc.GetAllocatedResource())
 			remaining := currentQueue.GetRemainingGuaranteedResource()
 			if remaining != nil && resources.StrictlyGreaterThanOrEquals(remaining, resources.Zero) {
+				log.Log(log.SchedPreemption).Info("Preemption Queue Guarantees check has passed", zap.String("ask", p.ask.GetAllocationKey()),
+					zap.String("ask queue path", p.queuePath))
 				return true
 			}
 		}
@@ -530,6 +537,10 @@ func (p *Preemptor) tryNodes() (string, []*Allocation, bool) {
 		}
 		// identify which victims and in which order should be tried
 		if idx, victims := p.calculateVictimsByNode(nodeAvailable, allocations); victims != nil {
+			log.Log(log.SchedPreemption).Info("Node wise Potential victims collected for preemption", zap.String("ask", p.ask.GetAllocationKey()),
+				zap.String("ask queue path", p.queuePath),
+				zap.String("node", nodeID),
+				zap.Int("victims count", len(victims)))
 			victimsByNode[nodeID] = victims
 			keys := make([]string, 0)
 			for _, victim := range victims {
@@ -569,6 +580,10 @@ func (p *Preemptor) TryPreemption() (*AllocationResult, bool) {
 		// no preemption possible
 		return nil, false
 	}
+	log.Log(log.SchedPreemption).Info("Node chosen for preemption", zap.String("ask", p.ask.GetAllocationKey()),
+		zap.String("ask queue path", p.queuePath),
+		zap.String("node", nodeID),
+		zap.Int("victims count", len(victims)))
 
 	// look for additional victims in case we have not yet made enough capacity in the queue
 	extraVictims, ok := p.calculateAdditionalVictims(victims)
@@ -576,6 +591,10 @@ func (p *Preemptor) TryPreemption() (*AllocationResult, bool) {
 		// not enough resources were preempted
 		return nil, false
 	}
+	log.Log(log.SchedPreemption).Info("Additional victims chosen for preemption", zap.String("ask", p.ask.GetAllocationKey()),
+		zap.String("ask queue path", p.queuePath),
+		zap.String("node", nodeID),
+		zap.Int("additional victims count", len(extraVictims)))
 	victims = append(victims, extraVictims...)
 	if len(victims) == 0 {
 		return nil, false

From 0cb6b73324695d3a5157545b1d047fb0171b3235 Mon Sep 17 00:00:00 2001
From: Manikandan R <manirajv06@gmail.com>
Date: Fri, 23 Aug 2024 11:58:14 +0530
Subject: [PATCH 2/2] Run the tests for many versions

---
 .github/workflows/pre-commit.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/pre-commit.yml b/.github/workflows/pre-commit.yml
index 0c5d1fb4b..fa86ecc0d 100644
--- a/.github/workflows/pre-commit.yml
+++ b/.github/workflows/pre-commit.yml
@@ -44,7 +44,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        k8s: [v1.31.0]
+        k8s: [v1.31.0, v1.30.0, v1.29.4, v1.28.9, v1.27.13, v1.26.15, v1.25.16, v1.24.17]
         plugin: [""]
     steps:
       - name: Checkout yunikorn-k8shim source code