From e6dd36759f07e5967f9f74cbda9c77a59a7977b1 Mon Sep 17 00:00:00 2001
From: Oleg Guba <oleg.guba@gmail.com>
Date: Thu, 29 Feb 2024 20:43:50 -0800
Subject: [PATCH 1/2] [kubernetes/scheduler] use lockless diagnosis collection
 in findNodesThatPassFilters

---
 pkg/scheduler/schedule_one.go | 19 ++++++++++++++-----
 1 file changed, 14 insertions(+), 5 deletions(-)

diff --git a/pkg/scheduler/schedule_one.go b/pkg/scheduler/schedule_one.go
index ae4be17bb63..42fd13eb38d 100644
--- a/pkg/scheduler/schedule_one.go
+++ b/pkg/scheduler/schedule_one.go
@@ -595,10 +595,15 @@ func (sched *Scheduler) findNodesThatPassFilters(
 	}
 
 	errCh := parallelize.NewErrorChannel()
-	var statusesLock sync.Mutex
 	var feasibleNodesLen int32
 	ctx, cancel := context.WithCancel(ctx)
 	defer cancel()
+
+	type nodeStatus struct {
+		node   string
+		status *framework.Status
+	}
+	result := make([]*nodeStatus, len(feasibleNodes))
 	checkNode := func(i int) {
 		// We check the nodes starting from where we left off in the previous scheduling cycle,
 		// this is to make sure all nodes have the same chance of being examined across pods.
@@ -617,12 +622,16 @@ func (sched *Scheduler) findNodesThatPassFilters(
 				feasibleNodes[length-1] = nodeInfo
 			}
 		} else {
-			statusesLock.Lock()
-			diagnosis.NodeToStatusMap[nodeInfo.Node().Name] = status
-			diagnosis.AddPluginStatus(status)
-			statusesLock.Unlock()
+			result[i] = &nodeStatus{node: nodeInfo.Node().Name, status: status}
 		}
 	}
+	for _, item := range result {
+		if item == nil {
+			continue
+		}
+		diagnosis.NodeToStatusMap[item.node] = item.status
+		diagnosis.AddPluginStatus(item.status)
+	}
 
 	beginCheckNode := time.Now()
 	statusCode := framework.Success

From ba525460e0519a47a6550c64e93f914a489610b4 Mon Sep 17 00:00:00 2001
From: Oleg Guba <oleg.guba@gmail.com>
Date: Fri, 1 Mar 2024 02:06:17 -0800
Subject: [PATCH 2/2] change result size to numAllNodes

---
 pkg/scheduler/schedule_one.go | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/pkg/scheduler/schedule_one.go b/pkg/scheduler/schedule_one.go
index 42fd13eb38d..5b0032f312c 100644
--- a/pkg/scheduler/schedule_one.go
+++ b/pkg/scheduler/schedule_one.go
@@ -603,7 +603,7 @@ func (sched *Scheduler) findNodesThatPassFilters(
 		node   string
 		status *framework.Status
 	}
-	result := make([]*nodeStatus, len(feasibleNodes))
+	result := make([]*nodeStatus, numAllNodes)
 	checkNode := func(i int) {
 		// We check the nodes starting from where we left off in the previous scheduling cycle,
 		// this is to make sure all nodes have the same chance of being examined across pods.
@@ -625,13 +625,6 @@ func (sched *Scheduler) findNodesThatPassFilters(
 			result[i] = &nodeStatus{node: nodeInfo.Node().Name, status: status}
 		}
 	}
-	for _, item := range result {
-		if item == nil {
-			continue
-		}
-		diagnosis.NodeToStatusMap[item.node] = item.status
-		diagnosis.AddPluginStatus(item.status)
-	}
 
 	beginCheckNode := time.Now()
 	statusCode := framework.Success
@@ -646,6 +639,13 @@ func (sched *Scheduler) findNodesThatPassFilters(
 	// are found.
 	fwk.Parallelizer().Until(ctx, numAllNodes, checkNode, metrics.Filter)
 	feasibleNodes = feasibleNodes[:feasibleNodesLen]
+	for _, item := range result {
+		if item == nil {
+			continue
+		}
+		diagnosis.NodeToStatusMap[item.node] = item.status
+		diagnosis.AddPluginStatus(item.status)
+	}
 	if err := errCh.ReceiveError(); err != nil {
 		statusCode = framework.Error
 		return feasibleNodes, err