From 7a1ac18bc84f8068725765e52b803073d3eba9c3 Mon Sep 17 00:00:00 2001
From: Kevin Hannon <kehannon@redhat.com>
Date: Thu, 5 Oct 2023 11:26:06 -0400
Subject: [PATCH] Fix panic if there are more terminating pods than active pods

Co-authored-by: Aldo Culquicondor <1299064+alculquicondor@users.noreply.github.com>
---
 pkg/controller/job/job_controller.go      |  2 +-
 pkg/controller/job/job_controller_test.go | 22 ++++++++++++++++++----
 2 files changed, 19 insertions(+), 5 deletions(-)

diff --git a/pkg/controller/job/job_controller.go b/pkg/controller/job/job_controller.go
index 34aae78e93d..01a35d3ee10 100644
--- a/pkg/controller/job/job_controller.go
+++ b/pkg/controller/job/job_controller.go
@@ -1490,7 +1490,7 @@ func (jm *Controller) manageJob(ctx context.Context, job *batch.Job, jobCtx *syn
 		}
 	}
 
-	rmAtLeast := active + terminating - wantActive
+	rmAtLeast := active - wantActive
 	if rmAtLeast < 0 {
 		rmAtLeast = 0
 	}
diff --git a/pkg/controller/job/job_controller_test.go b/pkg/controller/job/job_controller_test.go
index 79b637e8af6..c6669887fb8 100644
--- a/pkg/controller/job/job_controller_test.go
+++ b/pkg/controller/job/job_controller_test.go
@@ -357,9 +357,10 @@ func TestControllerSyncJob(t *testing.T) {
 			jobPodReplacementPolicy: true,
 			terminatingPods:         1,
 			expectedTerminating:     ptr.To[int32](1),
-			expectedPodPatches:      2,
-			expectedDeletions:       1,
-			expectedFailed:          1,
+			// Removes finalizer and deletes one failed pod
+			expectedPodPatches: 1,
+			expectedFailed:     1,
+			expectedActive:     1,
 		},
 		"WQ job: recreate pods when terminating or failed": {
 			parallelism:             1,
@@ -375,7 +376,20 @@ func TestControllerSyncJob(t *testing.T) {
 			expectedPodPatches:      2,
 			expectedFailed:          2,
 		},
-
+		"more terminating pods than parallelism": {
+			parallelism:             1,
+			completions:             1,
+			backoffLimit:            6,
+			activePods:              2,
+			failedPods:              0,
+			terminatingPods:         4,
+			podReplacementPolicy:    podReplacementPolicy(batch.Failed),
+			jobPodReplacementPolicy: true,
+			expectedTerminating:     ptr.To[int32](4),
+			expectedActive:          1,
+			expectedDeletions:       1,
+			expectedPodPatches:      1,
+		},
 		"too few active pods and active back-off": {
 			parallelism:  1,
 			completions:  1,