Stop Jellyfin for a straightforward full backup unless your snapshot method can capture application state consistently while the service is writing.
The trade-off is downtime versus consistency. A stopped-service archive is easy to reason about because database, configuration, and metadata stop changing during the copy. A live backup can be valid when the database backup mechanism or filesystem snapshot creates a coherent point in time, but an ordinary recursive copy during active writes is harder to trust.
Stopped-Service Copies Are the Simple Safe Baseline
Stopping Jellyfin briefly prevents new database and metadata writes while the backup tool walks the config tree. This removes many consistency questions for small home servers.
A live file copy can miss WAL-backed transactions; transaction-aware SQLite backups avoid copying an open database as if it were an ordinary static file.
Schedule the pause during a quiet window, confirm the process is stopped, copy the persistent state, and start the service again. Measure the outage so you know the real operational cost.
Live Backups Need a Consistent Snapshot Mechanism
A filesystem snapshot can freeze the view of many files at one instant even while the live service continues afterward. That is different from slowly copying changing files one by one.
Active datasets change during backup, so churn matters when a capture spans time.
If you use ZFS, Btrfs, or a database-aware backup, document what consistency guarantee it provides. Do not call a plain live file copy equivalent without testing.
Database Integrity Matters More Than Backup Completion
A backup job can return success while the captured database is not a usable recovery point. Verification must inspect the database and the application state together.
A trustworthy recovery set should avoid uncontrolled database copies during active writes; SQLite consistency depends on preserving a coherent database state.
Restore the backup to a disposable path and run an integrity check before depending on it. The persistent app-data layout makes this test easier because state is separated from the replaceable container.
Choose the Method From Your Recovery Objective
A household that can tolerate a two-minute maintenance window may gain little from complex live-backup machinery. A server with strict uptime goals may justify snapshots, but only if restores remain predictable.
A proper disaster-recovery test measures whether the chosen backup actually returns the service to a usable state.
Compare backup downtime, restore time, and failure complexity. Use the simplest method that meets the household recovery objective and survives a real restore rehearsal.
Support & Tips
More to Read

Why Does Jellyfin Run Hot or Noisy When Nobody Is Streaming?
Idle heat usually means background work or a shared-host workload, so identify the active process and scheduled task before changing cooling or hardware.

When Should You Rebuild Instead of Repairing Jellyfin?
Choose rebuild over repair when runtime drift is the problem and persistent state is backed up; do not โrebuildโ by deleting the only good...

How Much Free Storage Should Jellyfin Keep for Background Jobs?
No universal free-space percentage fits Jellyfin; measure persistent growth and temporary peaks separately, then reserve margin above both.

