docs(troubleshooting): add inotify limits documentation and workaround
Add comprehensive troubleshooting documentation for inotify instance limit
errors that cause Jellyfin crashes in environments with many containers.
Changes:
- Add Troubleshooting section to README with inotify limits explanation
- Document proper solution: increase kernel limits via sysctl
- Provide workaround for managed clusters: DOTNET_USE_POLLING_FILE_WATCHER
- Add commented example in values.yaml for easy copy-paste
- Update Chart version from 2.5.0 to 2.5.1
- Add Artifact Hub changelog annotation
The proper solution is to increase inotify limits on Kubernetes nodes:
```bash
sysctl -w fs.inotify.max_user_instances=512
```
For managed Kubernetes where node access is restricted, users can use the
polling workaround (less efficient but functional):
```yaml
jellyfin:
env:
- name: DOTNET_USE_POLLING_FILE_WATCHER
value: "1"
```
This addresses the crash issue mentioned in #64.
Fixes: #64
Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Aleksei Sviridkin <f@lex.la>
This commit is contained in:
1 parent
2401a45bb1
commit
4a45dc998a
4 files changed
+92
-3
No files matched your search
@@ -1,6 +1,6 @@
|
||||
# jellyfin
|
||||
|
||||
  
|
||||
  
|
||||
|
||||
A Helm chart for Jellyfin Media Server
|
||||
|
||||
@@ -68,7 +68,7 @@ helm install my-jellyfin jellyfin/jellyfin -f values.yaml
|
||||
| jellyfin.args | list | `[]` | Additional arguments for the entrypoint command. |
|
||||
| jellyfin.command | list | `[]` | Custom command to use as container entrypoint. |
|
||||
| jellyfin.enableDLNA | bool | `false` | Enable DLNA. Requires host network. See: https://jellyfin.org/docs/general/networking/dlna.html |
|
||||
| jellyfin.env | list | `[]` | Additional environment variables for the container. |
|
||||
| jellyfin.env | list | `[]` | Additional environment variables for the container. Example: Workaround for inotify limits (see Troubleshooting section in README) |
|
||||
| livenessProbe | object | `{"initialDelaySeconds":10,"tcpSocket":{"port":"http"}}` | Configure liveness probe for Jellyfin. |
|
||||
| metrics | object | `{"enabled":false,"serviceMonitor":{"enabled":false,"interval":"30s","labels":{},"metricRelabelings":[],"namespace":"","path":"/metrics","port":8096,"relabelings":[],"scheme":"http","scrapeTimeout":"30s","targetLabels":[],"tlsConfig":{}}}` | Configuration for metrics collection and monitoring |
|
||||
| metrics.enabled | bool | `false` | Enable or disable metrics collection |
|
||||
@@ -160,3 +160,44 @@ extraVolumeMounts:
|
||||
- name: hwa
|
||||
mountPath: /dev/dri
|
||||
```
|
||||
|
||||
## Troubleshooting
|
||||
|
||||
### inotify Instance Limit Reached
|
||||
|
||||
**Problem:** Jellyfin crashes with error:
|
||||
```
|
||||
System.IO.IOException: The configured user limit (128) on the number of inotify instances has been reached
|
||||
```
|
||||
|
||||
**Root cause:** The Linux kernel has a limit on inotify instances (file system watchers) per user. Jellyfin uses inotify to monitor media libraries for changes.
|
||||
|
||||
**Proper solution (recommended):**
|
||||
|
||||
Increase the inotify limit on the Kubernetes nodes:
|
||||
|
||||
```bash
|
||||
# Temporary (until reboot)
|
||||
sysctl -w fs.inotify.max_user_instances=512
|
||||
|
||||
# Permanent
|
||||
echo "fs.inotify.max_user_instances=512" >> /etc/sysctl.conf
|
||||
sysctl -p
|
||||
```
|
||||
|
||||
Recommended values:
|
||||
- `fs.inotify.max_user_instances`: 512 or higher
|
||||
- `fs.inotify.max_user_watches`: 524288 or higher (if you have large media libraries)
|
||||
|
||||
**Workaround (if you cannot modify host settings):**
|
||||
|
||||
If you're running on a managed Kubernetes cluster where you cannot modify node-level settings, you can force Jellyfin to use polling instead of inotify. **Note: This is less efficient and may increase CPU usage and delay change detection.**
|
||||
|
||||
```yaml
|
||||
jellyfin:
|
||||
env:
|
||||
- name: DOTNET_USE_POLLING_FILE_WATCHER
|
||||
value: "1"
|
||||
```
|
||||
|
||||
This workaround disables inotify file watching in favor of periodic polling, which doesn't require inotify instances but is less efficient.
|
||||
Reference in new issue
Block a user