From 7fa3bc9c342812e876d99dc6beb7a0cdf1dae396 Mon Sep 17 00:00:00 2001
From: jenkins <brad@bstein.dev>
Date: Sat, 3 Oct 2026 01:04:58 -0500
Subject: [PATCH] core: quarantine workers with sustained runtime storage
 stalls

---
 infrastructure/core/kustomization.yaml    |  1 +
 infrastructure/core/node-maintenance.yaml | 22 ++++++++++++++++++++++
 2 files changed, 23 insertions(+)
 create mode 100644 infrastructure/core/node-maintenance.yaml

diff --git a/infrastructure/core/kustomization.yaml b/infrastructure/core/kustomization.yaml
index 68972357..cbc9d5be 100644
--- a/infrastructure/core/kustomization.yaml
+++ b/infrastructure/core/kustomization.yaml
@@ -2,6 +2,7 @@
 apiVersion: kustomize.config.k8s.io/v1beta1
 kind: Kustomization
 resources:
+  - node-maintenance.yaml
   - ../modules/base
   - ../modules/profiles/atlas-ha
   - node-prefer-noschedule-serviceaccount.yaml
diff --git a/infrastructure/core/node-maintenance.yaml b/infrastructure/core/node-maintenance.yaml
new file mode 100644
index 00000000..62d19297
--- /dev/null
+++ b/infrastructure/core/node-maintenance.yaml
@@ -0,0 +1,22 @@
+# infrastructure/core/node-maintenance.yaml
+# Block new application placement while the measured runtime I/O faults remain.
+# Existing workloads and storage DaemonSets are not drained by these resources.
+apiVersion: v1
+kind: Node
+metadata:
+  name: titan-14
+  annotations:
+    kustomize.toolkit.fluxcd.io/prune: disabled
+    atlas.bstein.dev/maintenance-reason: "Runtime USB flash saturation; replace or repair media and verify I/O before uncordoning."
+spec:
+  unschedulable: true
+---
+apiVersion: v1
+kind: Node
+metadata:
+  name: titan-18
+  annotations:
+    kustomize.toolkit.fluxcd.io/prune: disabled
+    atlas.bstein.dev/maintenance-reason: "Sustained runtime I/O stalls prevent Longhorn engine startup; reduce load and verify storage before uncordoning."
+spec:
+  unschedulable: true