From 6c8c0dd9e4b8a469f24cb14b078cf7b2d21fe60f Mon Sep 17 00:00:00 2001 From: Ben Vincent Date: Sat, 15 Aug 2026 13:38:59 +1000 Subject: [PATCH] Deploy jellyfin-ha as a true-HA StatefulSet under a new media project (#237) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit ## Why Deploy the jellyfin-ha fork as a genuine high-availability service rather than a single replica, so its two headline capabilities can actually be exercised: the Redis-coordinated distributed transcoding (a surviving pod resumes the HLS segments of a pod that dies mid-stream) and the experimental PostgreSQL main database (which is what lets more than one replica share the same library). It lands in its own `jellyfin` namespace under a new `media` ArgoCD project. ## How **Workload — StatefulSet, 2 replicas.** The Deployment becomes a StatefulSet so each pod has a stable name. The fork's Redis transcode lease keys ownership on `JELLYFIN_INSTANCE_ID`, which is set from the downward-API pod name, giving each replica a unique, stable lease identity. Soft `podAntiAffinity` spreads the two pods across nodes and a `PodDisruptionBudget` keeps `minAvailable: 1` through drains and rollouts. **Main database — CloudNativePG.** A CNPG trio in-namespace mirrors the litellm pattern: a 3-instance `Cluster`, a PgBouncer `Pooler`, and Ceph RGW (barman) S3 backups to a dedicated `cnpg-jellyfin` bucket owned by a `cephrgw` `ObjectStoreUser`. An init container writes `/config/config/database.xml` selecting the fork's `Jellyfin-PostgreSQL` provider, and the connection string is composed from the CNPG-generated `jellyfin-postgres-app` secret (username / password / dbname) pointed at the pooler service — the password is never rendered into the manifest. Library-item metadata therefore moves off SQLite; metadata images, plugins, subtitles and config XML stay on `/config`. **Storage.** - `/config` is now a shared `ReadWriteMany` cephfs PVC (raid5, retain) so every replica reads/writes the same metadata and config. - `/config/transcodes` — the fork's real transcode temp path — is a shared RWX PVC (raid5, delete). This is the load-bearing fix: takeover reads the dead pod's in-flight `.ts`/`.m3u8` segments off shared storage, so per-pod scratch would silently break it. - `/cache` is per-pod via a `volumeClaimTemplate` (RWO). - The media library stays a fresh, empty RWX PVC mounted read-only; populating it is out of scope. **Hardware transcoding.** The container requests the `gpu.intel.com/i915` Intel device-plugin resource (which pins the pod to a GPU-labelled node and injects the DRI render node — no `/dev/dri` hostPath or privileged container) plus the render/video supplemental groups. VA-API hardware acceleration is now on by default: the `inject-config` init container seeds `/config/config/encoding.xml` with `HardwareAccelerationType` `vaapi`, `EnableHardwareEncoding`, the injected render node (`/dev/dri/renderD128`) and h264/hevc hardware decode, so transcodes use the iGPU on first boot with no manual admin-UI step. Both seed files (`database.xml`, `encoding.xml`) are written only when absent, so later admin changes persisted to the shared RWX `/config` are never clobbered on restart. **Networking.** The Gateway/HTTPRoute (traefik-internal, `jellyfin.k8s.syd1.au.unkin.net`) is unchanged; the Service gains `sessionAffinity: ClientIP` to keep a client pinned to one replica and reduce transcode-session churn. **Redis.** The in-namespace single-replica Redis stays as the transcode lease store. ## Follow-up UDP auto-discovery is disabled, but scheduled library scans still run on every replica (redundant scans). Single-scanner leader election is a planned follow-up pending a fork feature, tracked separately. --------- Co-authored-by: unkin-agent Co-authored-by: Ben Vincent Co-authored-by: Ben Vin Reviewed-on: https://git.unkin.net/unkin/argocd-apps/pulls/237 Co-authored-by: Ben Vincent Co-committed-by: Ben Vincent --- apps/base/jellyfin/cephrgw-config-backup.yaml | 32 +++ apps/base/jellyfin/cnpg_backup.yaml | 45 ++++ apps/base/jellyfin/cnpg_cluster.yaml | 126 +++++++++ apps/base/jellyfin/cnpg_pooler.yaml | 36 +++ apps/base/jellyfin/gateway.yaml | 37 +++ apps/base/jellyfin/httproute.yaml | 49 ++++ apps/base/jellyfin/kustomization.yaml | 27 ++ apps/base/jellyfin/namespace.yaml | 5 + apps/base/jellyfin/pdb.yaml | 13 + apps/base/jellyfin/pv-media-movies.yaml | 30 +++ apps/base/jellyfin/pv-media-tv.yaml | 30 +++ apps/base/jellyfin/pvc-config.yaml | 17 ++ apps/base/jellyfin/pvc-media-movies.yaml | 24 ++ apps/base/jellyfin/pvc-media-tv.yaml | 24 ++ apps/base/jellyfin/pvc-transcode.yaml | 23 ++ apps/base/jellyfin/redis-deployment.yaml | 66 +++++ apps/base/jellyfin/redis-pvc.yaml | 19 ++ apps/base/jellyfin/redis-service.yaml | 17 ++ apps/base/jellyfin/schedule.yaml | 58 +++++ apps/base/jellyfin/service.yaml | 18 ++ apps/base/jellyfin/statefulset.yaml | 245 ++++++++++++++++++ apps/base/jellyfin/vaultauth.yaml | 20 ++ apps/base/jellyfin/vaultstaticsecret.yaml | 24 ++ .../au-syd1/jellyfin/kustomization.yaml | 6 + argocd/applicationsets/kustomization.yaml | 1 + argocd/applicationsets/media.yaml | 31 +++ argocd/projects/kustomization.yaml | 1 + argocd/projects/media.yaml | 23 ++ 28 files changed, 1047 insertions(+) create mode 100644 apps/base/jellyfin/cephrgw-config-backup.yaml create mode 100644 apps/base/jellyfin/cnpg_backup.yaml create mode 100644 apps/base/jellyfin/cnpg_cluster.yaml create mode 100644 apps/base/jellyfin/cnpg_pooler.yaml create mode 100644 apps/base/jellyfin/gateway.yaml create mode 100644 apps/base/jellyfin/httproute.yaml create mode 100644 apps/base/jellyfin/kustomization.yaml create mode 100644 apps/base/jellyfin/namespace.yaml create mode 100644 apps/base/jellyfin/pdb.yaml create mode 100644 apps/base/jellyfin/pv-media-movies.yaml create mode 100644 apps/base/jellyfin/pv-media-tv.yaml create mode 100644 apps/base/jellyfin/pvc-config.yaml create mode 100644 apps/base/jellyfin/pvc-media-movies.yaml create mode 100644 apps/base/jellyfin/pvc-media-tv.yaml create mode 100644 apps/base/jellyfin/pvc-transcode.yaml create mode 100644 apps/base/jellyfin/redis-deployment.yaml create mode 100644 apps/base/jellyfin/redis-pvc.yaml create mode 100644 apps/base/jellyfin/redis-service.yaml create mode 100644 apps/base/jellyfin/schedule.yaml create mode 100644 apps/base/jellyfin/service.yaml create mode 100644 apps/base/jellyfin/statefulset.yaml create mode 100644 apps/base/jellyfin/vaultauth.yaml create mode 100644 apps/base/jellyfin/vaultstaticsecret.yaml create mode 100644 apps/overlays/au-syd1/jellyfin/kustomization.yaml create mode 100644 argocd/applicationsets/media.yaml create mode 100644 argocd/projects/media.yaml diff --git a/apps/base/jellyfin/cephrgw-config-backup.yaml b/apps/base/jellyfin/cephrgw-config-backup.yaml new file mode 100644 index 0000000..14ad7ed --- /dev/null +++ b/apps/base/jellyfin/cephrgw-config-backup.yaml @@ -0,0 +1,32 @@ +--- +# Second Ceph RGW (S3) bucket owned by the existing jellyfin backup user +# (cnpg-jellyfin-backup, defined in cnpg_backup.yaml) — one user, two buckets: +# the CNPG barman bucket plus this one, which k8up uses to hold restic backups +# of the jellyfin-config PVC. The BucketAccess emits read-write S3 creds into a +# Secret the k8up Schedule consumes. +apiVersion: ceph.unkin.net/v1alpha1 +kind: Bucket +metadata: + name: jellyfin-config-backup + namespace: jellyfin +spec: + placementTarget: ec + bucketName: jellyfin-config-backup + ownerRef: cnpg-jellyfin-backup + versioning: false + tags: + app: jellyfin + purpose: config-backup + retainOnDelete: true +--- +apiVersion: ceph.unkin.net/v1alpha1 +kind: BucketAccess +metadata: + name: jellyfin-config-backup + namespace: jellyfin +spec: + bucketRef: jellyfin-config-backup + level: read-write + # Operator writes AWS_ACCESS_KEY_ID / AWS_SECRET_ACCESS_KEY (+ S3_ENDPOINT, + # BUCKET_NAME) into this Secret; the k8up Schedule reads the access keys. + secretName: jellyfin-config-backup-s3 diff --git a/apps/base/jellyfin/cnpg_backup.yaml b/apps/base/jellyfin/cnpg_backup.yaml new file mode 100644 index 0000000..723c82c --- /dev/null +++ b/apps/base/jellyfin/cnpg_backup.yaml @@ -0,0 +1,45 @@ +--- +# Ceph RGW (S3) backup target for the jellyfin CNPG cluster, provisioned by the +# in-estate cephrgw-operator: one dedicated bucket + owner user. CNPG reads the +# S3 credential Secret from its own namespace. +apiVersion: ceph.unkin.net/v1alpha1 +kind: ObjectStoreUser +metadata: + name: cnpg-jellyfin-backup + namespace: jellyfin +spec: + displayName: "CNPG backup owner (jellyfin)" + uid: cnpg-jellyfin-backup + maxBuckets: 5 + secretName: cnpg-jellyfin-backup-s3 + retainOnDelete: true +--- +apiVersion: ceph.unkin.net/v1alpha1 +kind: Bucket +metadata: + name: cnpg-jellyfin + namespace: jellyfin +spec: + placementTarget: ec + bucketName: cnpg-jellyfin + ownerRef: cnpg-jellyfin-backup + versioning: false + tags: + app: jellyfin + purpose: cnpg-backup + retainOnDelete: true +--- +# Nightly base backup on top of always-on WAL archiving. Scheduled off-peak and +# staggered from the other CNPG clusters (6-field cron, seconds first). +apiVersion: postgresql.cnpg.io/v1 +kind: ScheduledBackup +metadata: + name: cnpg-jellyfin-nightly + namespace: jellyfin +spec: + schedule: "0 35 3 * * *" + immediate: false + backupOwnerReference: self + method: barmanObjectStore + cluster: + name: jellyfin-postgres diff --git a/apps/base/jellyfin/cnpg_cluster.yaml b/apps/base/jellyfin/cnpg_cluster.yaml new file mode 100644 index 0000000..46a2476 --- /dev/null +++ b/apps/base/jellyfin/cnpg_cluster.yaml @@ -0,0 +1,126 @@ +--- +# Main Jellyfin database. The jellyfin-ha fork's experimental EF Core provider +# moves the entire Jellyfin DB (incl. library items) off SQLite into PostgreSQL, +# which is what makes a shared-nothing multi-replica deployment possible. No +# bootstrap secret is given, so CNPG generates the jellyfin-postgres-app secret +# (username/password/dbname) that the StatefulSet composes its DSN from. +apiVersion: postgresql.cnpg.io/v1 +kind: Cluster +metadata: + name: jellyfin-postgres + namespace: jellyfin +spec: + # Exclude the operator-managed data PVCs (jellyfin-postgres-N) from the + # jellyfin-config k8up Schedule (skipWithoutAnnotation is false cluster-wide, + # so unannotated PVCs are swept in). Postgres has its own barmanObjectStore + # backup below; restic must not touch the raw RWO data volumes. + inheritedMetadata: + annotations: + k8up.io/backup: "false" + affinity: + podAntiAffinityType: preferred + backup: + retentionPolicy: 30d + barmanObjectStore: + # Dedicated per-cluster Ceph RGW bucket (cephrgw-operator provisions it). + destinationPath: s3://cnpg-jellyfin + endpointURL: https://s3.ceph.unkin.net + endpointCA: + name: vault-ca-cert + key: ca.crt + s3Credentials: + accessKeyId: + name: cnpg-jellyfin-backup-s3 + key: AWS_ACCESS_KEY_ID + secretAccessKey: + name: cnpg-jellyfin-backup-s3 + key: AWS_SECRET_ACCESS_KEY + serverName: jellyfin + data: + compression: bzip2 + jobs: 2 + wal: + compression: zstd + maxParallel: 2 + bootstrap: + initdb: + database: jellyfin + encoding: UTF8 + localeCType: C + localeCollate: C + owner: jellyfin + enablePDB: true + enableSuperuserAccess: false + failoverDelay: 0 + # PG 17 — accepted by the fork's Npgsql/EF Core provider (needs PG14+); the + # provider generates its own migrations on first start. + imageName: ghcr.io/cloudnative-pg/postgresql:17-system-trixie + instances: 3 + logLevel: info + maxSyncReplicas: 0 + minSyncReplicas: 0 + monitoring: + customQueriesConfigMap: + - key: queries + name: cnpg-default-monitoring + disableDefaultQueries: false + enablePodMonitor: false + postgresql: + parameters: + archive_mode: "on" + archive_timeout: 5min + dynamic_shared_memory_type: posix + effective_cache_size: 256MB + full_page_writes: "on" + log_destination: csvlog + log_directory: /controller/log + log_filename: postgres + log_rotation_age: "0" + log_rotation_size: "0" + log_truncate_on_rotation: "false" + logging_collector: "on" + max_connections: "200" + max_parallel_workers: "16" + max_replication_slots: "16" + max_worker_processes: "16" + shared_buffers: 128MB + shared_memory_type: mmap + ssl_max_protocol_version: TLSv1.3 + ssl_min_protocol_version: TLSv1.3 + wal_keep_size: 256MB + wal_level: logical + wal_log_hints: "on" + wal_receiver_timeout: 5s + wal_sender_timeout: 5s + syncReplicaElectionConstraint: + enabled: false + primaryUpdateMethod: restart + primaryUpdateStrategy: unsupervised + probes: + liveness: + isolationCheck: + connectionTimeout: 1000 + enabled: true + requestTimeout: 1000 + replicationSlots: + highAvailability: + enabled: true + slotPrefix: _cnpg_ + synchronizeReplicas: + enabled: true + updateInterval: 30 + resources: + limits: + cpu: "1" + memory: 1Gi + requests: + cpu: 50m + memory: 512Mi + smartShutdownTimeout: 180 + startDelay: 3600 + stopDelay: 1800 + storage: + resizeInUseVolumes: true + size: 10Gi + storageClass: cephrbd-fast-delete + switchoverDelay: 3600 diff --git a/apps/base/jellyfin/cnpg_pooler.yaml b/apps/base/jellyfin/cnpg_pooler.yaml new file mode 100644 index 0000000..50f9263 --- /dev/null +++ b/apps/base/jellyfin/cnpg_pooler.yaml @@ -0,0 +1,36 @@ +--- +# PgBouncer pooler in front of the jellyfin-postgres cluster. Jellyfin connects +# here (jellyfin-postgres-pooler:5432) rather than the -rw service so EF Core's +# connection churn is absorbed by the pool. +apiVersion: postgresql.cnpg.io/v1 +kind: Pooler +metadata: + name: jellyfin-postgres-pooler + namespace: jellyfin +spec: + cluster: + name: jellyfin-postgres + instances: 2 + pgbouncer: + parameters: + default_pool_size: "50" + max_client_conn: "200" + paused: false + poolMode: session + template: + metadata: + labels: + app: jellyfin-pooler + spec: + affinity: + podAntiAffinity: + requiredDuringSchedulingIgnoredDuringExecution: + - labelSelector: + matchExpressions: + - key: app + operator: In + values: + - jellyfin-pooler + topologyKey: kubernetes.io/hostname + containers: [] + type: rw diff --git a/apps/base/jellyfin/gateway.yaml b/apps/base/jellyfin/gateway.yaml new file mode 100644 index 0000000..a491519 --- /dev/null +++ b/apps/base/jellyfin/gateway.yaml @@ -0,0 +1,37 @@ +--- +apiVersion: gateway.networking.k8s.io/v1 +kind: Gateway +metadata: + labels: + traefik.io/instance: internal + annotations: + cert-manager.io/cluster-issuer: vault-issuer + cert-manager.io/common-name: jellyfin.k8s.syd1.au.unkin.net + cert-manager.io/private-key-size: "4096" + external-dns.alpha.kubernetes.io/hostname: jellyfin.k8s.syd1.au.unkin.net + external-dns.alpha.kubernetes.io/target: 198.18.200.4 + name: jellyfin + namespace: jellyfin +spec: + gatewayClassName: traefik-internal + listeners: + - allowedRoutes: + namespaces: + from: Same + hostname: jellyfin.k8s.syd1.au.unkin.net + name: http + port: 80 + protocol: HTTP + - allowedRoutes: + namespaces: + from: Same + hostname: jellyfin.k8s.syd1.au.unkin.net + name: https + port: 443 + protocol: HTTPS + tls: + certificateRefs: + - group: "" + kind: Secret + name: jellyfin-tls + mode: Terminate diff --git a/apps/base/jellyfin/httproute.yaml b/apps/base/jellyfin/httproute.yaml new file mode 100644 index 0000000..1778dcd --- /dev/null +++ b/apps/base/jellyfin/httproute.yaml @@ -0,0 +1,49 @@ +--- +apiVersion: gateway.networking.k8s.io/v1 +kind: HTTPRoute +metadata: + name: http-redirect + namespace: jellyfin +spec: + hostnames: + - jellyfin.k8s.syd1.au.unkin.net + parentRefs: + - group: gateway.networking.k8s.io + kind: Gateway + name: jellyfin + sectionName: http + rules: + - filters: + - type: RequestRedirect + requestRedirect: + scheme: https + statusCode: 301 + matches: + - path: + type: PathPrefix + value: / +--- +apiVersion: gateway.networking.k8s.io/v1 +kind: HTTPRoute +metadata: + name: jellyfin-route + namespace: jellyfin +spec: + hostnames: + - jellyfin.k8s.syd1.au.unkin.net + parentRefs: + - group: gateway.networking.k8s.io + kind: Gateway + name: jellyfin + sectionName: https + rules: + - backendRefs: + - group: "" + kind: Service + name: jellyfin + port: 8096 + weight: 1 + matches: + - path: + type: PathPrefix + value: / diff --git a/apps/base/jellyfin/kustomization.yaml b/apps/base/jellyfin/kustomization.yaml new file mode 100644 index 0000000..1374887 --- /dev/null +++ b/apps/base/jellyfin/kustomization.yaml @@ -0,0 +1,27 @@ +--- +apiVersion: kustomize.config.k8s.io/v1beta1 +kind: Kustomization + +resources: + - namespace.yaml + - cnpg_cluster.yaml + - cnpg_pooler.yaml + - cnpg_backup.yaml + - cephrgw-config-backup.yaml + - vaultauth.yaml + - vaultstaticsecret.yaml + - schedule.yaml + - pvc-config.yaml + - pvc-transcode.yaml + - pv-media-tv.yaml + - pv-media-movies.yaml + - pvc-media-tv.yaml + - pvc-media-movies.yaml + - statefulset.yaml + - pdb.yaml + - service.yaml + - redis-deployment.yaml + - redis-pvc.yaml + - redis-service.yaml + - gateway.yaml + - httproute.yaml diff --git a/apps/base/jellyfin/namespace.yaml b/apps/base/jellyfin/namespace.yaml new file mode 100644 index 0000000..3122e11 --- /dev/null +++ b/apps/base/jellyfin/namespace.yaml @@ -0,0 +1,5 @@ +--- +apiVersion: v1 +kind: Namespace +metadata: + name: jellyfin diff --git a/apps/base/jellyfin/pdb.yaml b/apps/base/jellyfin/pdb.yaml new file mode 100644 index 0000000..4b21b5b --- /dev/null +++ b/apps/base/jellyfin/pdb.yaml @@ -0,0 +1,13 @@ +--- +# Keep at least one Jellyfin replica serving through voluntary disruptions +# (node drains, rollouts) so active streams can fail over rather than drop. +apiVersion: policy/v1 +kind: PodDisruptionBudget +metadata: + name: jellyfin + namespace: jellyfin +spec: + minAvailable: 1 + selector: + matchLabels: + app: jellyfin diff --git a/apps/base/jellyfin/pv-media-movies.yaml b/apps/base/jellyfin/pv-media-movies.yaml new file mode 100644 index 0000000..b9ddfcc --- /dev/null +++ b/apps/base/jellyfin/pv-media-movies.yaml @@ -0,0 +1,30 @@ +--- +# Static PV for the shared MOVIES CephFS subvolume. Same rootPath as arrstack's +# movies PV so radarr writes and jellyfin reads the identical library tree; each +# namespace gets its own PV (unique name + volumeHandle) pinned by claimRef. +apiVersion: v1 +kind: PersistentVolume +metadata: + name: jellyfin-media-movies +spec: + capacity: + storage: 1Ti + accessModes: + - ReadWriteMany + persistentVolumeReclaimPolicy: Retain + storageClassName: "" + volumeMode: Filesystem + claimRef: + namespace: jellyfin + name: jellyfin-media-movies + csi: + driver: cephfs.csi.ceph.com + volumeHandle: jellyfin-media-movies-static + nodeStageSecretRef: + name: csi-cephfs-secret + namespace: csi-cephfs + volumeAttributes: + staticVolume: "true" + clusterID: cephfs_csi_ssd_ec_4_1 + fsName: cephfs + rootPath: /volumes/csi_ssd_ec_4_1/media-movies/e95d8ace-c736-465a-acc3-0c3e46dcede9 diff --git a/apps/base/jellyfin/pv-media-tv.yaml b/apps/base/jellyfin/pv-media-tv.yaml new file mode 100644 index 0000000..a3df747 --- /dev/null +++ b/apps/base/jellyfin/pv-media-tv.yaml @@ -0,0 +1,30 @@ +--- +# Static PV for the shared TV CephFS subvolume. Same rootPath as arrstack's +# TV PV so sonarr writes and jellyfin reads the identical library tree; each +# namespace gets its own PV (unique name + volumeHandle) pinned by claimRef. +apiVersion: v1 +kind: PersistentVolume +metadata: + name: jellyfin-media-tv +spec: + capacity: + storage: 1Ti + accessModes: + - ReadWriteMany + persistentVolumeReclaimPolicy: Retain + storageClassName: "" + volumeMode: Filesystem + claimRef: + namespace: jellyfin + name: jellyfin-media-tv + csi: + driver: cephfs.csi.ceph.com + volumeHandle: jellyfin-media-tv-static + nodeStageSecretRef: + name: csi-cephfs-secret + namespace: csi-cephfs + volumeAttributes: + staticVolume: "true" + clusterID: cephfs_csi_ssd_ec_4_1 + fsName: cephfs + rootPath: /volumes/csi_ssd_ec_4_1/media-tv/4692957d-f5df-4f72-b9c9-56e4ee6d1333 diff --git a/apps/base/jellyfin/pvc-config.yaml b/apps/base/jellyfin/pvc-config.yaml new file mode 100644 index 0000000..53e6a4f --- /dev/null +++ b/apps/base/jellyfin/pvc-config.yaml @@ -0,0 +1,17 @@ +--- +# Jellyfin config: metadata images, plugins, subtitles and config XML. Shared +# ReadWriteMany across replicas (all pods read/write the same library metadata); +# the main library DB now lives in PostgreSQL, not here. Retain — this is state. +apiVersion: v1 +kind: PersistentVolumeClaim +metadata: + name: jellyfin-config + namespace: jellyfin +spec: + accessModes: + - ReadWriteMany + resources: + requests: + storage: 20Gi + storageClassName: cephfs-raid5-retain + volumeMode: Filesystem diff --git a/apps/base/jellyfin/pvc-media-movies.yaml b/apps/base/jellyfin/pvc-media-movies.yaml new file mode 100644 index 0000000..4fe15f6 --- /dev/null +++ b/apps/base/jellyfin/pvc-media-movies.yaml @@ -0,0 +1,24 @@ +--- +# Movie library, shared read-many across replicas. Statically bound to the +# jellyfin-media-movies PV (shared CephFS subvolume also used by arrstack/radarr). +# storageClassName "" + volumeName disables dynamic provisioning and binds the +# pre-created static PV. +apiVersion: v1 +kind: PersistentVolumeClaim +metadata: + name: jellyfin-media-movies + namespace: jellyfin + annotations: + # Exclude from the jellyfin-config k8up Schedule (skipWithoutAnnotation is + # false cluster-wide, so unannotated PVCs are swept in). Only jellyfin-config + # is backed up; the media library is not restic-backup material. + k8up.io/backup: "false" +spec: + accessModes: + - ReadWriteMany + resources: + requests: + storage: 1Ti + storageClassName: "" + volumeName: jellyfin-media-movies + volumeMode: Filesystem diff --git a/apps/base/jellyfin/pvc-media-tv.yaml b/apps/base/jellyfin/pvc-media-tv.yaml new file mode 100644 index 0000000..6593bc5 --- /dev/null +++ b/apps/base/jellyfin/pvc-media-tv.yaml @@ -0,0 +1,24 @@ +--- +# TV library, shared read-many across replicas. Statically bound to the +# jellyfin-media-tv PV (shared CephFS subvolume also used by arrstack/sonarr). +# storageClassName "" + volumeName disables dynamic provisioning and binds the +# pre-created static PV. +apiVersion: v1 +kind: PersistentVolumeClaim +metadata: + name: jellyfin-media-tv + namespace: jellyfin + annotations: + # Exclude from the jellyfin-config k8up Schedule (skipWithoutAnnotation is + # false cluster-wide, so unannotated PVCs are swept in). Only jellyfin-config + # is backed up; the media library is not restic-backup material. + k8up.io/backup: "false" +spec: + accessModes: + - ReadWriteMany + resources: + requests: + storage: 1Ti + storageClassName: "" + volumeName: jellyfin-media-tv + volumeMode: Filesystem diff --git a/apps/base/jellyfin/pvc-transcode.yaml b/apps/base/jellyfin/pvc-transcode.yaml new file mode 100644 index 0000000..01750ad --- /dev/null +++ b/apps/base/jellyfin/pvc-transcode.yaml @@ -0,0 +1,23 @@ +--- +# Shared transcode scratch. ReadWriteMany is the hard requirement for the HA +# fork: a taking-over pod must read the in-flight HLS segments written by the +# pod it replaces. Scratch data (delete reclaim); raid5 avoids the raid6 +# double-parity write penalty on the many small HLS segment writes. +apiVersion: v1 +kind: PersistentVolumeClaim +metadata: + name: jellyfin-transcode + namespace: jellyfin + annotations: + # Exclude from the jellyfin-config k8up Schedule (skipWithoutAnnotation is + # false cluster-wide, so unannotated PVCs are swept in). Transcode is RWX + # scratch — nothing to back up. + k8up.io/backup: "false" +spec: + accessModes: + - ReadWriteMany + resources: + requests: + storage: 100Gi + storageClassName: cephfs-raid5-delete + volumeMode: Filesystem diff --git a/apps/base/jellyfin/redis-deployment.yaml b/apps/base/jellyfin/redis-deployment.yaml new file mode 100644 index 0000000..eeae15b --- /dev/null +++ b/apps/base/jellyfin/redis-deployment.yaml @@ -0,0 +1,66 @@ +--- +apiVersion: apps/v1 +kind: Deployment +metadata: + name: redis + namespace: jellyfin +spec: + replicas: 1 + selector: + matchLabels: + app: redis + strategy: + type: Recreate + template: + metadata: + labels: + app: redis + spec: + containers: + - name: redis + image: redis:7-alpine + imagePullPolicy: IfNotPresent + command: + - redis-server + - --save + - "20" + - "1" + ports: + - containerPort: 6379 + name: redis + protocol: TCP + livenessProbe: + exec: + command: + - redis-cli + - ping + failureThreshold: 3 + initialDelaySeconds: 30 + periodSeconds: 30 + successThreshold: 1 + timeoutSeconds: 5 + readinessProbe: + exec: + command: + - redis-cli + - ping + failureThreshold: 3 + initialDelaySeconds: 5 + periodSeconds: 10 + successThreshold: 1 + timeoutSeconds: 5 + resources: + limits: + cpu: 500m + memory: 512Mi + requests: + cpu: 50m + memory: 128Mi + volumeMounts: + - mountPath: /data + name: data + restartPolicy: Always + volumes: + - name: data + persistentVolumeClaim: + claimName: jellyfin-redis-data diff --git a/apps/base/jellyfin/redis-pvc.yaml b/apps/base/jellyfin/redis-pvc.yaml new file mode 100644 index 0000000..a6717ad --- /dev/null +++ b/apps/base/jellyfin/redis-pvc.yaml @@ -0,0 +1,19 @@ +--- +apiVersion: v1 +kind: PersistentVolumeClaim +metadata: + name: jellyfin-redis-data + namespace: jellyfin + annotations: + # Exclude from the jellyfin-config k8up Schedule (skipWithoutAnnotation is + # false cluster-wide, so unannotated PVCs are swept in). Redis holds only + # ephemeral transcode-lease state; RWO would also fail to mount while in use. + k8up.io/backup: "false" +spec: + accessModes: + - ReadWriteOnce + resources: + requests: + storage: 5Gi + storageClassName: cephrbd-fast-delete + volumeMode: Filesystem diff --git a/apps/base/jellyfin/redis-service.yaml b/apps/base/jellyfin/redis-service.yaml new file mode 100644 index 0000000..6e0e34f --- /dev/null +++ b/apps/base/jellyfin/redis-service.yaml @@ -0,0 +1,17 @@ +--- +apiVersion: v1 +kind: Service +metadata: + name: redis + namespace: jellyfin +spec: + internalTrafficPolicy: Cluster + ports: + - name: redis + port: 6379 + protocol: TCP + targetPort: redis + selector: + app: redis + sessionAffinity: None + type: ClusterIP diff --git a/apps/base/jellyfin/schedule.yaml b/apps/base/jellyfin/schedule.yaml new file mode 100644 index 0000000..6043a18 --- /dev/null +++ b/apps/base/jellyfin/schedule.yaml @@ -0,0 +1,58 @@ +--- +# k8up Schedule: restic backups of the jellyfin-config PVC (library metadata, +# plugins, config XML) to the dedicated Ceph RGW config-backup bucket. S3 creds +# come from the cephrgw BucketAccess Secret (jellyfin-config-backup-s3); the +# restic repo password comes from Vault via the jellyfin-k8up-restic Secret. +# +# s3.ceph.unkin.net presents the internal unkin.net CA, which the k8up/restic +# image does not trust by default, so the reflected vault-ca-cert Secret is +# mounted into every job pod and pointed at via backend.tlsOptions.caCert. +apiVersion: k8up.io/v1 +kind: Schedule +metadata: + name: jellyfin-config + namespace: jellyfin +spec: + backend: + repoPasswordSecretRef: + name: jellyfin-k8up-restic + key: password + s3: + endpoint: https://s3.ceph.unkin.net + bucket: jellyfin-config-backup + accessKeyIDSecretRef: + name: jellyfin-config-backup-s3 + key: AWS_ACCESS_KEY_ID + secretAccessKeySecretRef: + name: jellyfin-config-backup-s3 + key: AWS_SECRET_ACCESS_KEY + tlsOptions: + caCert: /etc/k8up/ca/ca.crt + volumeMounts: + - name: vault-ca + mountPath: /etc/k8up/ca + readOnly: true + backup: + schedule: "0 2 * * *" + failedJobsHistoryLimit: 3 + successfulJobsHistoryLimit: 3 + volumes: + - name: vault-ca + secret: + secretName: vault-ca-cert + prune: + schedule: "0 3 * * 0" + retention: + keepDaily: 14 + keepWeekly: 8 + keepMonthly: 12 + volumes: + - name: vault-ca + secret: + secretName: vault-ca-cert + check: + schedule: "0 4 * * 0" + volumes: + - name: vault-ca + secret: + secretName: vault-ca-cert diff --git a/apps/base/jellyfin/service.yaml b/apps/base/jellyfin/service.yaml new file mode 100644 index 0000000..a8c810e --- /dev/null +++ b/apps/base/jellyfin/service.yaml @@ -0,0 +1,18 @@ +--- +apiVersion: v1 +kind: Service +metadata: + name: jellyfin + namespace: jellyfin +spec: + internalTrafficPolicy: Cluster + ports: + - name: http + port: 8096 + protocol: TCP + targetPort: http + selector: + app: jellyfin + # Pin each client to one replica to reduce transcode-session churn/takeover. + sessionAffinity: ClientIP + type: ClusterIP diff --git a/apps/base/jellyfin/statefulset.yaml b/apps/base/jellyfin/statefulset.yaml new file mode 100644 index 0000000..11fdbd4 --- /dev/null +++ b/apps/base/jellyfin/statefulset.yaml @@ -0,0 +1,245 @@ +--- +apiVersion: apps/v1 +kind: StatefulSet +metadata: + name: jellyfin + namespace: jellyfin +spec: + # HA: two replicas coordinate transcode session ownership through Redis and + # resume each other's HLS segments off the shared RWX transcode PVC. Stable + # pod names (jellyfin-0/1) are the lease owner identity, hence StatefulSet. + replicas: 2 + serviceName: jellyfin + podManagementPolicy: Parallel + updateStrategy: + type: RollingUpdate + selector: + matchLabels: + app: jellyfin + template: + metadata: + labels: + app: jellyfin + spec: + securityContext: + # Group-write the shared RWX volumes and grant the render/video groups so + # the runAsUser 1000 process can open the Intel DRI render node injected + # by the device plugin. + fsGroup: 1000 + supplementalGroups: + - 44 + - 105 + - 109 + seccompProfile: + type: RuntimeDefault + affinity: + # Spread the two replicas across nodes for node-level HA. Soft so a + # single-GPU-node cluster still schedules both (i915 has 4 shared slots). + podAntiAffinity: + preferredDuringSchedulingIgnoredDuringExecution: + - weight: 100 + podAffinityTerm: + labelSelector: + matchLabels: + app: jellyfin + topologyKey: kubernetes.io/hostname + initContainers: + # Seed the fork's PostgreSQL provider (database.xml) and Intel iGPU + # hardware transcode settings (encoding.xml) before Jellyfin starts. + # Runs as root to chown into the shared config volume; mirrors the fork + # Helm chart's inject-db-config. Each file is written only when absent so + # admin changes persisted to the shared RWX /config survive pod restarts. + - name: inject-config + image: busybox:1.37.0 + command: + - sh + - -c + - | + mkdir -p /config/config + chown 1000:1000 /config/config + chmod 775 /config/config + if [ ! -e /config/config/database.xml ]; then + cat > /config/config/database.xml << 'DBEOF' + + + Jellyfin-PostgreSQL + NoLock + + DBEOF + chown 1000:1000 /config/config/database.xml + chmod 664 /config/config/database.xml + fi + # VAAPI on the Intel render node the device plugin injects + # (/dev/dri/renderD128 — ffmpeg's default DRM node, reachable via + # the render/video supplementalGroups). Without this the attached + # iGPU is idle and every transcode runs in software. Omitted + # elements fall back to the fork's EncodingOptions defaults. + if [ ! -e /config/config/encoding.xml ]; then + cat > /config/config/encoding.xml << 'ENCEOF' + + + -1 + vaapi + /dev/dri/renderD128 + true + true + false + false + false + false + false + + h264 + hevc + vc1 + vp9 + + + ENCEOF + chown 1000:1000 /config/config/encoding.xml + chmod 664 /config/config/encoding.xml + fi + resources: + requests: + cpu: 10m + memory: 32Mi + limits: + cpu: 100m + memory: 64Mi + volumeMounts: + - name: config + mountPath: /config + containers: + - name: jellyfin + image: artifactapi.k8s.syd1.au.unkin.net/docker-internal/jellyfin-ha:v0.1.0 + imagePullPolicy: IfNotPresent + ports: + - name: http + containerPort: 8096 + protocol: TCP + env: + # Pod identity for the Redis transcode lease owner. The fork reads + # JELLYFIN_INSTANCE_ID (falling back to MachineName); the stable + # StatefulSet pod name gives each replica a unique lease identity so + # takeover can target a dead replica. JELLYFIN_HA_POD_NAME is set for + # parity with the fork Helm chart (nothing currently reads it). + - name: JELLYFIN_INSTANCE_ID + valueFrom: + fieldRef: + fieldPath: metadata.name + - name: JELLYFIN_HA_POD_NAME + valueFrom: + fieldRef: + fieldPath: metadata.name + # Multiple replicas must not each answer UDP auto-discovery. + - name: JELLYFIN_Network__AutoDiscovery + value: "false" + # Config dir must differ from the data root (Jellyfin sanity check). + - name: JELLYFIN_CONFIG_DIR + value: /config/config + # Distributed transcode session store (jellyfin-ha additions). + - name: Jellyfin__TranscodeStore__RedisConnectionString + value: "redis:6379,abortConnect=false" + - name: Jellyfin__TranscodeStore__LeaseDurationSeconds + value: "30" + # PostgreSQL main DB via the CNPG-generated app secret, routed through + # the PgBouncer pooler. Composed with $(VAR) expansion so the password + # is never rendered into the manifest; CNPG passwords are URL-safe. + - name: PGUSER + valueFrom: + secretKeyRef: + name: jellyfin-postgres-app + key: username + - name: PGPASSWORD + valueFrom: + secretKeyRef: + name: jellyfin-postgres-app + key: password + - name: PGDB + valueFrom: + secretKeyRef: + name: jellyfin-postgres-app + key: dbname + - name: POSTGRES_CONNECTION_STRING + value: "postgresql://$(PGUSER):$(PGPASSWORD)@jellyfin-postgres-pooler:5432/$(PGDB)" + - name: DATABASE_URL + value: "postgresql://$(PGUSER):$(PGPASSWORD)@jellyfin-postgres-pooler:5432/$(PGDB)" + livenessProbe: + httpGet: + path: /health + port: http + initialDelaySeconds: 30 + periodSeconds: 30 + timeoutSeconds: 5 + failureThreshold: 3 + readinessProbe: + httpGet: + path: /health + port: http + initialDelaySeconds: 10 + periodSeconds: 10 + timeoutSeconds: 5 + failureThreshold: 3 + resources: + requests: + cpu: "1" + memory: 1Gi + gpu.intel.com/i915: "1" + limits: + cpu: "4" + memory: 6Gi + # Intel iGPU (QSV/VA-API) slot. Requesting it pins the pod to a + # GPU-labelled node and injects /dev/dri/renderD* automatically, so + # no /dev/dri hostPath or privileged container is needed. VA-API is + # pre-enabled via the seeded encoding.xml (see inject-config), so + # transcodes use the iGPU on first boot with no manual UI step. + gpu.intel.com/i915: "1" + securityContext: + runAsUser: 1000 + runAsGroup: 1000 + volumeMounts: + - name: config + mountPath: /config + - name: transcode + # Fork's real transcode temp path. RWX so a surviving pod reads the + # in-flight .ts/.m3u8 segments of the pod it takes over. A per-pod + # volume here silently breaks HA takeover. + mountPath: /config/transcodes + - name: cache + mountPath: /cache + - name: media-tv + mountPath: /media/tv + readOnly: true + - name: media-movies + mountPath: /media/movies + readOnly: true + volumes: + - name: config + persistentVolumeClaim: + claimName: jellyfin-config + - name: transcode + persistentVolumeClaim: + claimName: jellyfin-transcode + - name: media-tv + persistentVolumeClaim: + claimName: jellyfin-media-tv + - name: media-movies + persistentVolumeClaim: + claimName: jellyfin-media-movies + volumeClaimTemplates: + # Per-pod scratch cache — RWO, disposable, one PVC per replica. + - metadata: + name: cache + annotations: + # Exclude the per-pod cache PVCs from the jellyfin-config k8up Schedule + # (skipWithoutAnnotation is false cluster-wide). Cache is disposable and + # RWO — it would also fail to mount into the backup pod while in use. + k8up.io/backup: "false" + spec: + accessModes: + - ReadWriteOnce + resources: + requests: + storage: 30Gi + storageClassName: cephrbd-fast-delete + volumeMode: Filesystem diff --git a/apps/base/jellyfin/vaultauth.yaml b/apps/base/jellyfin/vaultauth.yaml new file mode 100644 index 0000000..9fbc2c2 --- /dev/null +++ b/apps/base/jellyfin/vaultauth.yaml @@ -0,0 +1,20 @@ +--- +apiVersion: secrets.hashicorp.com/v1beta1 +kind: VaultAuth +metadata: + name: default + namespace: jellyfin + annotations: + argocd.argoproj.io/sync-wave: "0" +spec: + allowedNamespaces: + - jellyfin + kubernetes: + audiences: + - vault + role: default + serviceAccount: default + tokenExpirationSeconds: 600 + method: kubernetes + mount: k8s/au/syd1 + vaultConnectionRef: vso-system/default diff --git a/apps/base/jellyfin/vaultstaticsecret.yaml b/apps/base/jellyfin/vaultstaticsecret.yaml new file mode 100644 index 0000000..9de7ef2 --- /dev/null +++ b/apps/base/jellyfin/vaultstaticsecret.yaml @@ -0,0 +1,24 @@ +--- +# restic repository password for the k8up jellyfin-config backups. Seeded at +# kv/kubernetes/namespace/jellyfin/default/k8up-restic (key: password); the +# default k8s role's templated policy already grants read here, so no +# terraform-vault change is needed. VSO syncs it into the jellyfin-k8up-restic +# Secret that the Schedule references via backend.repoPasswordSecretRef. +apiVersion: secrets.hashicorp.com/v1beta1 +kind: VaultStaticSecret +metadata: + name: jellyfin-k8up-restic + namespace: jellyfin + annotations: + argocd.argoproj.io/sync-wave: "0" +spec: + destination: + create: true + name: jellyfin-k8up-restic + overwrite: true + hmacSecretData: true + mount: kv + path: kubernetes/namespace/jellyfin/default/k8up-restic + refreshAfter: 5m + type: kv-v2 + vaultAuthRef: default diff --git a/apps/overlays/au-syd1/jellyfin/kustomization.yaml b/apps/overlays/au-syd1/jellyfin/kustomization.yaml new file mode 100644 index 0000000..d8d5df6 --- /dev/null +++ b/apps/overlays/au-syd1/jellyfin/kustomization.yaml @@ -0,0 +1,6 @@ +--- +apiVersion: kustomize.config.k8s.io/v1beta1 +kind: Kustomization + +resources: + - ../../../base/jellyfin diff --git a/argocd/applicationsets/kustomization.yaml b/argocd/applicationsets/kustomization.yaml index b54cc35..c95dedd 100644 --- a/argocd/applicationsets/kustomization.yaml +++ b/argocd/applicationsets/kustomization.yaml @@ -5,6 +5,7 @@ kind: Kustomization resources: - aitooling.yaml - logging.yaml + - media.yaml - observability.yaml - platform.yaml - storage.yaml diff --git a/argocd/applicationsets/media.yaml b/argocd/applicationsets/media.yaml new file mode 100644 index 0000000..9fe99a5 --- /dev/null +++ b/argocd/applicationsets/media.yaml @@ -0,0 +1,31 @@ +--- +apiVersion: argoproj.io/v1alpha1 +kind: ApplicationSet +metadata: + name: media-apps + namespace: argocd +spec: + generators: + - git: + repoURL: https://git.unkin.net/unkin/argocd-apps + revision: HEAD + directories: + - path: apps/overlays/*/jellyfin + template: + metadata: + name: 'media-{{path[3]}}' + spec: + project: media + source: + repoURL: https://git.unkin.net/unkin/argocd-apps + targetRevision: HEAD + path: '{{path}}' + destination: + server: https://kubernetes.default.svc + namespace: '{{path[3]}}' + syncPolicy: + automated: + prune: true + selfHeal: true + syncOptions: + - ServerSideApply=true diff --git a/argocd/projects/kustomization.yaml b/argocd/projects/kustomization.yaml index b54cc35..c95dedd 100644 --- a/argocd/projects/kustomization.yaml +++ b/argocd/projects/kustomization.yaml @@ -5,6 +5,7 @@ kind: Kustomization resources: - aitooling.yaml - logging.yaml + - media.yaml - observability.yaml - platform.yaml - storage.yaml diff --git a/argocd/projects/media.yaml b/argocd/projects/media.yaml new file mode 100644 index 0000000..4b3f29f --- /dev/null +++ b/argocd/projects/media.yaml @@ -0,0 +1,23 @@ +--- +apiVersion: argoproj.io/v1alpha1 +kind: AppProject +metadata: + name: media + namespace: argocd +spec: + description: Media services + sourceRepos: + - https://git.unkin.net/unkin/argocd-apps + destinations: + - namespace: 'jellyfin' + server: https://kubernetes.default.svc + - namespace: 'arrstack' + server: https://kubernetes.default.svc + clusterResourceWhitelist: + - group: '' + kind: Namespace + - group: '' + kind: PersistentVolume + namespaceResourceWhitelist: + - group: '*' + kind: '*'