From cd837ea69dc4afe69d48c0f33b3011a8023996d1 Mon Sep 17 00:00:00 2001
From: Paul Buetow <paul@buetow.org>
Date: Sun, 10 May 2026 10:57:44 +0300
Subject: navidrome: move /data PVC from NFS to local-path on r1

SQLite over NFS causes two problems: file-lock races on rolling
restarts (fixed with Recreate strategy but underlying fragility
remains), and 19s image-cache init at startup due to stunnel TLS
latency on every cache read.

Replace navidrome-data-pv/pvc (static hostPath over NFS at
/data/nfs/k3svolumes/navidrome/data) with a dynamic local-path PVC
provisioned on r1 (/var/lib/rancher/k3s/storage). Pin the deployment
to r1 via nodeSelector so the local PV is always reachable.

Existing DB and cache migrated: navidrome.db (23 MB), image/background/
plugin caches (~118 MB) copied via a migration pod before first start.
Result: startupTime=41ms (was ~20s), Image cache init=29ms (was ~19s).
Music PVC stays on NFS (200 GB library, unchanged).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---
 f3s/navidrome/helm-chart/templates/deployment.yaml |  8 ++++-
 .../helm-chart/templates/persistent-volume.yaml    | 34 ++++++++++++----------
 2 files changed, 25 insertions(+), 17 deletions(-)

diff --git a/f3s/navidrome/helm-chart/templates/deployment.yaml b/f3s/navidrome/helm-chart/templates/deployment.yaml
index 5a3862e..d85bb47 100644
--- a/f3s/navidrome/helm-chart/templates/deployment.yaml
+++ b/f3s/navidrome/helm-chart/templates/deployment.yaml
@@ -18,6 +18,12 @@ spec:
       labels:
         app: navidrome
     spec:
+      # Pin to r1 so the local-path PVC (which lives on r1's disk at
+      # /var/lib/rancher/k3s/storage) is always reachable. Without this, if
+      # the pod reschedules to a different node it would not find the data
+      # volume and Navidrome would start with an empty DB.
+      nodeSelector:
+        kubernetes.io/hostname: r1.lan.buetow.org
       containers:
       - name: navidrome
         image: deluan/navidrome:latest
@@ -34,7 +40,7 @@ spec:
           httpGet:
             path: /ping
             port: 4533
-          # Allow up to 5 min for cold start (NFS cache warm-up, scan, etc.)
+          # Allow up to 5 min for cold start (DB open, first scan, etc.)
           initialDelaySeconds: 15
           periodSeconds: 5
           timeoutSeconds: 3
diff --git a/f3s/navidrome/helm-chart/templates/persistent-volume.yaml b/f3s/navidrome/helm-chart/templates/persistent-volume.yaml
index 960bee3..3ca2867 100644
--- a/f3s/navidrome/helm-chart/templates/persistent-volume.yaml
+++ b/f3s/navidrome/helm-chart/templates/persistent-volume.yaml
@@ -1,31 +1,33 @@
-apiVersion: v1
-kind: PersistentVolume
-metadata:
-  name: navidrome-data-pv
-spec:
-  capacity:
-    storage: 10Gi
-  volumeMode: Filesystem
-  accessModes:
-    - ReadWriteOnce
-  persistentVolumeReclaimPolicy: Retain
-  hostPath:
-    path: /data/nfs/k3svolumes/navidrome/data
-    type: Directory
----
+# navidrome-data: local-path PVC (dynamic provisioning).
+# The SQLite DB, image cache, and artwork cache must NOT live on NFS because:
+#   1. SQLite file-lock semantics are fragile over NFS: under concurrent
+#      access (e.g. rolling restart) the new pod blocks on fcntl() until
+#      the old pod fully releases its locks.
+#   2. Every image/background cache read traverses the stunnel TLS tunnel,
+#      adding latency to every UI thumbnail and a 19 s "Image cache" init
+#      at startup.
+# The local-path storageClass provisions a directory on the node's own disk
+# (/var/lib/rancher/k3s/storage), eliminating both problems. The deployment
+# pins the pod to r1 via nodeSelector so the local PV is always reachable.
+# Trade-off: if r1 is down and the pod reschedules, it starts with a fresh
+# DB (loses play counts / scrobble queue) — acceptable for a home server.
 apiVersion: v1
 kind: PersistentVolumeClaim
 metadata:
   name: navidrome-data-pvc
   namespace: services
 spec:
-  storageClassName: ""
+  # local-path is the k3s default dynamic provisioner; WaitForFirstConsumer
+  # means the PV is created on the node where the pod is first scheduled.
+  storageClassName: local-path
   accessModes:
     - ReadWriteOnce
   resources:
     requests:
       storage: 10Gi
 ---
+# navidrome-music: keep on NFS — 200 GB library shared across nodes.
+# This is a static PV backed by hostPath at the NFS mount point.
 apiVersion: v1
 kind: PersistentVolume
 metadata:
-- 
cgit v1.2.3