From 4ab43ca04361e8698ef7242f7c3287e4f242047e Mon Sep 17 00:00:00 2001
From: dvirlabs <dvirlabs@gmail.com>
Date: Mon, 2 Mar 2026 03:43:35 +0200
Subject: [PATCH] Add toleration to make sure the apps run on the gpu worker

---
 manifests/ollama/values.yaml     | 7 +++++++
 manifests/open-webui/values.yaml | 7 +++++++
 2 files changed, 14 insertions(+)

diff --git a/manifests/ollama/values.yaml b/manifests/ollama/values.yaml
index 1b5eb59..0309e4a 100644
--- a/manifests/ollama/values.yaml
+++ b/manifests/ollama/values.yaml
@@ -20,6 +20,13 @@ resources:
 nodeSelector:
   kubernetes.io/hostname: k3s-worker-5-gpu
 
+# Tolerate GPU node taint
+tolerations:
+  - key: gpu
+    operator: Equal
+    value: "true"
+    effect: NoSchedule
+
 persistentVolume:
   enabled: true
   size: 30Gi
diff --git a/manifests/open-webui/values.yaml b/manifests/open-webui/values.yaml
index 1f64395..a2b595f 100644
--- a/manifests/open-webui/values.yaml
+++ b/manifests/open-webui/values.yaml
@@ -23,3 +23,10 @@ resources:
 # Schedule on GPU worker node (same as Ollama for low latency)
 nodeSelector:
   kubernetes.io/hostname: k3s-worker-5-gpu
+
+# Tolerate GPU node taint
+tolerations:
+  - key: gpu
+    operator: Equal
+    value: "true"
+    effect: NoSchedule