..
2024-07-18 08:13:42 -04:00
2024-05-20 17:24:10 -06:00
2024-07-18 08:13:42 -04:00
2024-07-20 10:33:43 -06:00
Jellyfin Software Media System
This is a helm chart for Jellyfin
Prerequisites
- Kubernetes 1.19+
- Helm 3+
TL;DR;
Configuration
The following tables lists the configurable parameters of the Jellyfin chart and their default values.
| Parameter | Description | Default |
|---|---|---|
image.repository |
Image repository | docker.io/jellyfin/jellyfin |
image.tag |
Image tag. Possible values listed here. | latest |
image.pullPolicy |
Image pull policy | IfNotPresent |
enableDLNA |
Enable DLNA for jellyfin | false |
Service.type |
Kubernetes service type for the jellyfin GUI | ClusterIP |
Service.port |
Kubernetes port where the jellyfin GUI is exposed | 8096 |
Service.annotations |
Service annotations for the jellyfin GUI | {} |
Service.labels |
Custom labels | {} |
Service.loadBalancerIP |
Loadbalance IP for the jellyfin GUI | {} |
Service.loadBalancerSourceRanges |
List of IP CIDRs allowed access to load balancer (if supported) | None |
ingress.enabled |
Enables Ingress | false |
ingress.annotations |
Ingress annotations | {} |
ingress.labels |
Custom labels | {} |
ingress.path |
Ingress path | / |
ingress.hosts |
Ingress accepted hostnames | chart-example.local |
ingress.tls |
Ingress TLS configuration | [] |
persistence.config.enabled |
Use persistent volume to store configuration data | false |
persistence.config.size |
Size of persistent volume claim | 1Gi |
persistence.config.existingClaim |
Use an existing PVC to persist data | nil |
persistence.config.storageClass |
Type of persistent volume claim | - |
persistence.config.accessMode |
Persistence access mode | ReadWriteOnce |
persistence.media.enabled |
Use persistent volume to store configuration data | true |
persistence.media.size |
Size of persistent volume claim | 10Gi |
persistence.media.existingClaim |
Use an existing PVC to persist data | nil |
persistence.media.storageClass |
Type of persistent volume claim | - |
persistence.media.accessMode |
Persistence access mode | ReadWriteOnce |
persistence.extraExistingClaimMounts |
Optionally add multiple existing claims | [] |
resources |
CPU/Memory resource requests/limits | {} |
nodeSelector |
Node labels for pod assignment | {} |
tolerations |
Toleration labels for pod assignment | [] |
affinity |
Affinity settings for pod assignment | {} |
livenessProbe.enabled |
Enable livenessProbe | true |
livenessProbe.failureThreshold |
Number of consecutive failures before restarting the pod | 3 |
livenessProbe.initialDelaySeconds |
Number of seconds after the container has started before liveness probes begin | 10 |
livenessProbe.periodSeconds |
How often to perform the probe | 10 |
livenessProbe.successThreshold |
Minimum consecutive successes for the probe to be considered successful after having failed | 1 |
livenessProbe.timeoutSeconds |
Number of seconds after which the probe times out | 1 |
readinessProbe.enabled |
Enable readinessProbe | true |
readinessProbe.failureThreshold |
Number of consecutive failures before restarting the pod | 3 |
readinessProbe.initialDelaySeconds |
Number of seconds after the container has started before liveness probes begin | 10 |
readinessProbe.periodSeconds |
How often to perform the probe | 10 |
readinessProbe.successThreshold |
Minimum consecutive successes for the probe to be considered successful after having failed | 1 |
readinessProbe.timeoutSeconds |
Number of seconds after which the probe times out | 1 |
Specify each parameter using the --set key=value[,key=value] argument to helm install.
Read through the values.yaml file. It has several suggested values.