From d2f534fd822fda817fa4cd3d941852f6c5559f52 Mon Sep 17 00:00:00 2001
From: nicktrn <55853254+nicktrn@users.noreply.github.com>
Date: Mon, 15 Sep 2025 15:02:47 +0100
Subject: [PATCH] feat(supervisor): add optional memory limit overhead

---
 apps/supervisor/src/env.ts                        | 1 +
 apps/supervisor/src/workloadManager/kubernetes.ts | 7 ++++++-
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/apps/supervisor/src/env.ts b/apps/supervisor/src/env.ts
index 99cb16425b..6d6bbd27d6 100644
--- a/apps/supervisor/src/env.ts
+++ b/apps/supervisor/src/env.ts
@@ -89,6 +89,7 @@ const Env = z.object({
   KUBERNETES_CPU_REQUEST_RATIO: z.coerce.number().min(0).max(1).default(0.75), // Ratio of CPU limit, so 0.75 = 75% of CPU limit
   KUBERNETES_MEMORY_REQUEST_MIN_GB: z.coerce.number().min(0).default(0),
   KUBERNETES_MEMORY_REQUEST_RATIO: z.coerce.number().min(0).max(1).default(1), // Ratio of memory limit, so 1 = 100% of memory limit
+  KUBERNETES_MEMORY_OVERHEAD_GB: z.coerce.number().min(0).optional(), // Optional memory overhead to add to the limit in GB
 
   // Placement tags settings
   PLACEMENT_TAGS_ENABLED: BoolEnv.default(false),
diff --git a/apps/supervisor/src/workloadManager/kubernetes.ts b/apps/supervisor/src/workloadManager/kubernetes.ts
index 55281c56a6..0f5e89c80a 100644
--- a/apps/supervisor/src/workloadManager/kubernetes.ts
+++ b/apps/supervisor/src/workloadManager/kubernetes.ts
@@ -25,6 +25,7 @@ export class KubernetesWorkloadManager implements WorkloadManager {
   private readonly cpuRequestRatio = env.KUBERNETES_CPU_REQUEST_RATIO;
   private readonly memoryRequestMinGb = env.KUBERNETES_MEMORY_REQUEST_MIN_GB;
   private readonly memoryRequestRatio = env.KUBERNETES_MEMORY_REQUEST_RATIO;
+  private readonly memoryOverheadGb = env.KUBERNETES_MEMORY_OVERHEAD_GB;
 
   constructor(private opts: WorkloadManagerOptions) {
     this.k8s = createK8sApi();
@@ -319,9 +320,13 @@ export class KubernetesWorkloadManager implements WorkloadManager {
   }
 
   #getResourceLimitsForMachine(preset: MachinePreset): ResourceQuantities {
+    const memoryLimit = this.memoryOverheadGb
+      ? preset.memory + this.memoryOverheadGb
+      : preset.memory;
+
     return {
       cpu: `${preset.cpu}`,
-      memory: `${preset.memory}G`,
+      memory: `${memoryLimit}G`,
     };
   }