From 1e5918aa031c111b7f4c221b06fb9bb4f00a918e Mon Sep 17 00:00:00 2001
From: Jerry Chan <jerryachen@tencent.com>
Date: Tue, 26 Sep 2023 21:48:09 +0800
Subject: [PATCH] Fixed the issue where node was not deleted when evicting pods
 timed out.

---
 .../actuation/group_deletion_scheduler.go          | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/cluster-autoscaler/core/scaledown/actuation/group_deletion_scheduler.go b/cluster-autoscaler/core/scaledown/actuation/group_deletion_scheduler.go
index e8aa9676932e..ed6582fce133 100644
--- a/cluster-autoscaler/core/scaledown/actuation/group_deletion_scheduler.go
+++ b/cluster-autoscaler/core/scaledown/actuation/group_deletion_scheduler.go
@@ -89,8 +89,18 @@ func (ds *GroupDeletionScheduler) ScheduleDeletion(nodeInfo *framework.NodeInfo,
 
 	nodeDeleteResult := ds.prepareNodeForDeletion(nodeInfo, drain)
 	if nodeDeleteResult.Err != nil {
-		ds.AbortNodeDeletion(nodeInfo.Node(), nodeGroup.Id(), drain, "prepareNodeForDeletion failed", nodeDeleteResult)
-		return
+		isTimeout := true
+		for _, podEvictionResult := range nodeDeleteResult.PodEvictionResults {
+			// If there is any error message, it means it is not a pure timeout and node deletion is aborted.
+			if podEvictionResult.Err != nil {
+				isTimeout = false
+				break
+			}
+		}
+		if !isTimeout {
+			ds.AbortNodeDeletion(nodeInfo.Node(), nodeGroup.Id(), drain, "prepareNodeForDeletion failed", nodeDeleteResult)
+			return
+		}
 	}
 
 	ds.addToBatcher(nodeInfo, nodeGroup, batchSize, drain, opts.ZeroOrMaxNodeScaling)