From 5430c9ac016c93a869e7a8e9f36cb6497869e0f8 Mon Sep 17 00:00:00 2001
From: jenkins <brad@bstein.dev>
Date: Sat, 22 Aug 2026 22:06:39 -0300
Subject: [PATCH] jellyfin: prepare titan-22 media host

---
 .../core/node-prefer-noschedule-cronjob.yaml  |  7 +-
 .../core/jellyfin-volume-policy-cronjob.yaml  | 67 +++++++++++++++++++
 .../longhorn/core/kustomization.yaml          |  1 +
 .../base/priorityclass/kustomization.yaml     |  1 +
 .../modules/base/priorityclass/media.yaml     |  8 +++
 5 files changed, 83 insertions(+), 1 deletion(-)
 create mode 100644 infrastructure/longhorn/core/jellyfin-volume-policy-cronjob.yaml
 create mode 100644 infrastructure/modules/base/priorityclass/media.yaml

diff --git a/infrastructure/core/node-prefer-noschedule-cronjob.yaml b/infrastructure/core/node-prefer-noschedule-cronjob.yaml
index e9c225c1..3e6c092d 100644
--- a/infrastructure/core/node-prefer-noschedule-cronjob.yaml
+++ b/infrastructure/core/node-prefer-noschedule-cronjob.yaml
@@ -61,7 +61,12 @@ spec:
                   clear_worker titan-22 amd64
 
                   if k get node titan-22 >/dev/null 2>&1; then
-                    k label node titan-22 atlas.bstein.dev/general-compute=last-resort --overwrite=true || true
+                    k label node titan-22 \
+                      atlas.bstein.dev/general-compute=last-resort \
+                      atlas.bstein.dev/media-primary=true \
+                      --overwrite=true || true
+                    # Keep spare capacity available without letting routine pods displace media service.
+                    k taint node titan-22 atlas.bstein.dev/media-primary=true:PreferNoSchedule --overwrite=true || true
                   fi
 
                   if k get node titan-23 >/dev/null 2>&1; then
diff --git a/infrastructure/longhorn/core/jellyfin-volume-policy-cronjob.yaml b/infrastructure/longhorn/core/jellyfin-volume-policy-cronjob.yaml
new file mode 100644
index 00000000..25044d44
--- /dev/null
+++ b/infrastructure/longhorn/core/jellyfin-volume-policy-cronjob.yaml
@@ -0,0 +1,67 @@
+# infrastructure/longhorn/core/jellyfin-volume-policy-cronjob.yaml
+apiVersion: batch/v1
+kind: CronJob
+metadata:
+  name: jellyfin-volume-policy
+  namespace: longhorn-system
+spec:
+  schedule: "* * * * *"
+  concurrencyPolicy: Forbid
+  startingDeadlineSeconds: 30
+  successfulJobsHistoryLimit: 1
+  failedJobsHistoryLimit: 3
+  jobTemplate:
+    spec:
+      activeDeadlineSeconds: 120
+      backoffLimit: 1
+      ttlSecondsAfterFinished: 600
+      template:
+        spec:
+          serviceAccountName: longhorn-service-account
+          restartPolicy: Never
+          containers:
+            - name: ensure
+              image: bitnami/kubectl@sha256:554ab88b1858e8424c55de37ad417b16f2a0e65d1607aa0f3fe3ce9b9f10b131
+              command:
+                - /usr/bin/env
+                - bash
+                - -ceu
+                - |
+                  ensure_volume() {
+                    local pvc_name="${1}"
+                    local replicas="${2}"
+                    local disk_selector="${3}"
+                    local pv_name
+                    local volume_name
+                    local current_replicas
+                    local current_selector
+
+                    pv_name="$(kubectl -n jellyfin get pvc "${pvc_name}" -o jsonpath='{.spec.volumeName}')"
+                    test -n "${pv_name}"
+                    volume_name="$(kubectl get pv "${pv_name}" -o jsonpath='{.spec.csi.volumeHandle}')"
+                    test -n "${volume_name}"
+                    current_replicas="$(kubectl -n longhorn-system get volumes.longhorn.io "${volume_name}" -o jsonpath='{.spec.numberOfReplicas}')"
+                    current_selector="$(kubectl -n longhorn-system get volumes.longhorn.io "${volume_name}" -o jsonpath='{.spec.diskSelector[0]}')"
+
+                    if [[ "${current_replicas}" != "${replicas}" || "${current_selector}" != "${disk_selector}" ]]; then
+                      kubectl -n longhorn-system patch volumes.longhorn.io "${volume_name}" \
+                        --type=merge \
+                        --patch="{\"spec\":{\"numberOfReplicas\":${replicas},\"diskSelector\":[\"${disk_selector}\"]}}"
+                    else
+                      echo "${pvc_name}: ${replicas} replicas remain on ${disk_selector}"
+                    fi
+                  }
+
+                  ensure_volume jellyfin-config-astreae 3 astreae
+                  ensure_volume jellyfin-media-asteria-new 2 asteria
+
+                  if kubectl -n longhorn-system get nodes.longhorn.io titan-22 >/dev/null 2>&1; then
+                    allow_scheduling="$(kubectl -n longhorn-system get nodes.longhorn.io titan-22 -o jsonpath='{.spec.allowScheduling}')"
+                    if [[ "${allow_scheduling}" != "false" ]]; then
+                      kubectl -n longhorn-system patch nodes.longhorn.io titan-22 \
+                        --type=merge \
+                        --patch='{"spec":{"allowScheduling":false}}'
+                    else
+                      echo "titan-22: Longhorn replica scheduling remains disabled"
+                    fi
+                  fi
diff --git a/infrastructure/longhorn/core/kustomization.yaml b/infrastructure/longhorn/core/kustomization.yaml
index 1e899d36..8f610d1b 100644
--- a/infrastructure/longhorn/core/kustomization.yaml
+++ b/infrastructure/longhorn/core/kustomization.yaml
@@ -12,6 +12,7 @@ resources:
   - longhorn-settings-ensure-job.yaml
   - longhorn-csi-toleration-ensure-job.yaml
   - longhorn-disk-tags-ensure-job.yaml
+  - jellyfin-volume-policy-cronjob.yaml
 
 configMapGenerator:
   - name: longhorn-settings-ensure-script
diff --git a/infrastructure/modules/base/priorityclass/kustomization.yaml b/infrastructure/modules/base/priorityclass/kustomization.yaml
index a9b27e71..d83776b7 100644
--- a/infrastructure/modules/base/priorityclass/kustomization.yaml
+++ b/infrastructure/modules/base/priorityclass/kustomization.yaml
@@ -4,4 +4,5 @@ kind: Kustomization
 resources:
   - scavenger.yaml
   - cassandra.yaml
+  - media.yaml
   - veles.yaml
diff --git a/infrastructure/modules/base/priorityclass/media.yaml b/infrastructure/modules/base/priorityclass/media.yaml
new file mode 100644
index 00000000..79c67ac8
--- /dev/null
+++ b/infrastructure/modules/base/priorityclass/media.yaml
@@ -0,0 +1,8 @@
+# infrastructure/modules/base/priorityclass/media.yaml
+apiVersion: scheduling.k8s.io/v1
+kind: PriorityClass
+metadata:
+  name: media-core
+value: 400
+globalDefault: false
+description: "For interactive media workloads; below stateful database services"