From 071217992b7acfde64279db66ad53425a7d62954 Mon Sep 17 00:00:00 2001
From: tginer <tginer@redhat.com>
Date: Thu, 20 Aug 2026 12:17:13 +0200
Subject: [PATCH] fix: disable leader election for single replica deployment

This change avoids unnecessary leader-election overhead, removing currently lease renewal failures upon API downtime.

Signed-off-by: tginer <tginer@redhat.com>
---
 deployments/gpu-operator/templates/operator.yaml | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/deployments/gpu-operator/templates/operator.yaml b/deployments/gpu-operator/templates/operator.yaml
index 484f83e088..c6a67cc0bc 100644
--- a/deployments/gpu-operator/templates/operator.yaml
+++ b/deployments/gpu-operator/templates/operator.yaml
@@ -8,7 +8,7 @@ metadata:
     app.kubernetes.io/component: "gpu-operator"
     nvidia.com/gpu-driver-upgrade-drain.skip: "true"
 spec:
-  replicas: 1
+  replicas: {{ .Values.operator.replicas | default 1 }}
   selector:
     matchLabels:
       app.kubernetes.io/component: "gpu-operator"
@@ -39,7 +39,9 @@ spec:
         imagePullPolicy: {{ .Values.operator.imagePullPolicy }}
         command: ["gpu-operator"]
         args:
+        {{- if gt (int (.Values.operator.replicas | default 1)) 1 }}
         - --leader-elect
+        {{- end }}
       {{- if .Values.operator.logging.develMode }}
         - --zap-devel
       {{- else }}